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));
2736 } else if(SvNVX(sv) == 0.0) {
2741 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2748 SvUPGRADE(sv, SVt_PV);
2751 s = SvGROW_mutable(sv, len + 1);
2754 return (char*)memcpy(s, tbuf, len + 1);
2760 assert(SvTYPE(sv) >= SVt_PVMG);
2761 /* This falls through to the report_uninit near the end of the
2763 } else if (SvTHINKFIRST(sv)) {
2768 if (flags & SV_SKIP_OVERLOAD)
2770 tmpstr = AMG_CALLun(sv,string);
2771 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2772 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2774 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2778 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2779 if (flags & SV_CONST_RETURN) {
2780 pv = (char *) SvPVX_const(tmpstr);
2782 pv = (flags & SV_MUTABLE_RETURN)
2783 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2786 *lp = SvCUR(tmpstr);
2788 pv = sv_2pv_flags(tmpstr, lp, flags);
2801 SV *const referent = SvRV(sv);
2805 retval = buffer = savepvn("NULLREF", len);
2806 } else if (SvTYPE(referent) == SVt_REGEXP) {
2807 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2812 /* If the regex is UTF-8 we want the containing scalar to
2813 have an UTF-8 flag too */
2819 if ((seen_evals = RX_SEEN_EVALS(re)))
2820 PL_reginterp_cnt += seen_evals;
2823 *lp = RX_WRAPLEN(re);
2825 return RX_WRAPPED(re);
2827 const char *const typestr = sv_reftype(referent, 0);
2828 const STRLEN typelen = strlen(typestr);
2829 UV addr = PTR2UV(referent);
2830 const char *stashname = NULL;
2831 STRLEN stashnamelen = 0; /* hush, gcc */
2832 const char *buffer_end;
2834 if (SvOBJECT(referent)) {
2835 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2838 stashname = HEK_KEY(name);
2839 stashnamelen = HEK_LEN(name);
2841 if (HEK_UTF8(name)) {
2847 stashname = "__ANON__";
2850 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2851 + 2 * sizeof(UV) + 2 /* )\0 */;
2853 len = typelen + 3 /* (0x */
2854 + 2 * sizeof(UV) + 2 /* )\0 */;
2857 Newx(buffer, len, char);
2858 buffer_end = retval = buffer + len;
2860 /* Working backwards */
2864 *--retval = PL_hexdigit[addr & 15];
2865 } while (addr >>= 4);
2871 memcpy(retval, typestr, typelen);
2875 retval -= stashnamelen;
2876 memcpy(retval, stashname, stashnamelen);
2878 /* retval may not neccesarily have reached the start of the
2880 assert (retval >= buffer);
2882 len = buffer_end - retval - 1; /* -1 for that \0 */
2890 if (SvREADONLY(sv) && !SvOK(sv)) {
2893 if (flags & SV_UNDEF_RETURNS_NULL)
2895 if (ckWARN(WARN_UNINITIALIZED))
2900 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2901 /* I'm assuming that if both IV and NV are equally valid then
2902 converting the IV is going to be more efficient */
2903 const U32 isUIOK = SvIsUV(sv);
2904 char buf[TYPE_CHARS(UV)];
2908 if (SvTYPE(sv) < SVt_PVIV)
2909 sv_upgrade(sv, SVt_PVIV);
2910 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2912 /* inlined from sv_setpvn */
2913 s = SvGROW_mutable(sv, len + 1);
2914 Move(ptr, s, len, char);
2918 else if (SvNOKp(sv)) {
2919 if (SvTYPE(sv) < SVt_PVNV)
2920 sv_upgrade(sv, SVt_PVNV);
2921 if (SvNVX(sv) == 0.0) {
2922 s = SvGROW_mutable(sv, 2);
2927 /* The +20 is pure guesswork. Configure test needed. --jhi */
2928 s = SvGROW_mutable(sv, NV_DIG + 20);
2929 /* some Xenix systems wipe out errno here */
2930 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2940 if (isGV_with_GP(sv)) {
2941 GV *const gv = MUTABLE_GV(sv);
2942 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2943 SV *const buffer = sv_newmortal();
2945 /* FAKE globs can get coerced, so need to turn this off temporarily
2948 gv_efullname3(buffer, gv, "*");
2949 SvFLAGS(gv) |= wasfake;
2951 if (SvPOK(buffer)) {
2953 *lp = SvCUR(buffer);
2955 return SvPVX(buffer);
2966 if (flags & SV_UNDEF_RETURNS_NULL)
2968 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2970 if (SvTYPE(sv) < SVt_PV)
2971 /* Typically the caller expects that sv_any is not NULL now. */
2972 sv_upgrade(sv, SVt_PV);
2976 const STRLEN len = s - SvPVX_const(sv);
2982 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2983 PTR2UV(sv),SvPVX_const(sv)));
2984 if (flags & SV_CONST_RETURN)
2985 return (char *)SvPVX_const(sv);
2986 if (flags & SV_MUTABLE_RETURN)
2987 return SvPVX_mutable(sv);
2992 =for apidoc sv_copypv
2994 Copies a stringified representation of the source SV into the
2995 destination SV. Automatically performs any necessary mg_get and
2996 coercion of numeric values into strings. Guaranteed to preserve
2997 UTF8 flag even from overloaded objects. Similar in nature to
2998 sv_2pv[_flags] but operates directly on an SV instead of just the
2999 string. Mostly uses sv_2pv_flags to do its work, except when that
3000 would lose the UTF-8'ness of the PV.
3006 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3009 const char * const s = SvPV_const(ssv,len);
3011 PERL_ARGS_ASSERT_SV_COPYPV;
3013 sv_setpvn(dsv,s,len);
3021 =for apidoc sv_2pvbyte
3023 Return a pointer to the byte-encoded representation of the SV, and set *lp
3024 to its length. May cause the SV to be downgraded from UTF-8 as a
3027 Usually accessed via the C<SvPVbyte> macro.
3033 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3035 PERL_ARGS_ASSERT_SV_2PVBYTE;
3037 sv_utf8_downgrade(sv,0);
3038 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3042 =for apidoc sv_2pvutf8
3044 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3045 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3047 Usually accessed via the C<SvPVutf8> macro.
3053 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3055 PERL_ARGS_ASSERT_SV_2PVUTF8;
3057 sv_utf8_upgrade(sv);
3058 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3063 =for apidoc sv_2bool
3065 This macro is only used by sv_true() or its macro equivalent, and only if
3066 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3067 It calls sv_2bool_flags with the SV_GMAGIC flag.
3069 =for apidoc sv_2bool_flags
3071 This function is only used by sv_true() and friends, and only if
3072 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3073 contain SV_GMAGIC, then it does an mg_get() first.
3080 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3084 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3086 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3092 SV * const tmpsv = AMG_CALLun(sv,bool_);
3093 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3094 return cBOOL(SvTRUE(tmpsv));
3096 return SvRV(sv) != 0;
3099 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3101 (*sv->sv_u.svu_pv > '0' ||
3102 Xpvtmp->xpv_cur > 1 ||
3103 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3110 return SvIVX(sv) != 0;
3113 return SvNVX(sv) != 0.0;
3115 if (isGV_with_GP(sv))
3125 =for apidoc sv_utf8_upgrade
3127 Converts the PV of an SV to its UTF-8-encoded form.
3128 Forces the SV to string form if it is not already.
3129 Will C<mg_get> on C<sv> if appropriate.
3130 Always sets the SvUTF8 flag to avoid future validity checks even
3131 if the whole string is the same in UTF-8 as not.
3132 Returns the number of bytes in the converted string
3134 This is not as a general purpose byte encoding to Unicode interface:
3135 use the Encode extension for that.
3137 =for apidoc sv_utf8_upgrade_nomg
3139 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3141 =for apidoc sv_utf8_upgrade_flags
3143 Converts the PV of an SV to its UTF-8-encoded form.
3144 Forces the SV to string form if it is not already.
3145 Always sets the SvUTF8 flag to avoid future validity checks even
3146 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3147 will C<mg_get> on C<sv> if appropriate, else not.
3148 Returns the number of bytes in the converted string
3149 C<sv_utf8_upgrade> and
3150 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3152 This is not as a general purpose byte encoding to Unicode interface:
3153 use the Encode extension for that.
3157 The grow version is currently not externally documented. It adds a parameter,
3158 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3159 have free after it upon return. This allows the caller to reserve extra space
3160 that it intends to fill, to avoid extra grows.
3162 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3163 which can be used to tell this function to not first check to see if there are
3164 any characters that are different in UTF-8 (variant characters) which would
3165 force it to allocate a new string to sv, but to assume there are. Typically
3166 this flag is used by a routine that has already parsed the string to find that
3167 there are such characters, and passes this information on so that the work
3168 doesn't have to be repeated.
3170 (One might think that the calling routine could pass in the position of the
3171 first such variant, so it wouldn't have to be found again. But that is not the
3172 case, because typically when the caller is likely to use this flag, it won't be
3173 calling this routine unless it finds something that won't fit into a byte.
3174 Otherwise it tries to not upgrade and just use bytes. But some things that
3175 do fit into a byte are variants in utf8, and the caller may not have been
3176 keeping track of these.)
3178 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3179 isn't guaranteed due to having other routines do the work in some input cases,
3180 or if the input is already flagged as being in utf8.
3182 The speed of this could perhaps be improved for many cases if someone wanted to
3183 write a fast function that counts the number of variant characters in a string,
3184 especially if it could return the position of the first one.
3189 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3193 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3195 if (sv == &PL_sv_undef)
3199 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3200 (void) sv_2pv_flags(sv,&len, flags);
3202 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3206 (void) SvPV_force(sv,len);
3211 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3216 sv_force_normal_flags(sv, 0);
3219 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3220 sv_recode_to_utf8(sv, PL_encoding);
3221 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3225 if (SvCUR(sv) == 0) {
3226 if (extra) SvGROW(sv, extra);
3227 } else { /* Assume Latin-1/EBCDIC */
3228 /* This function could be much more efficient if we
3229 * had a FLAG in SVs to signal if there are any variant
3230 * chars in the PV. Given that there isn't such a flag
3231 * make the loop as fast as possible (although there are certainly ways
3232 * to speed this up, eg. through vectorization) */
3233 U8 * s = (U8 *) SvPVX_const(sv);
3234 U8 * e = (U8 *) SvEND(sv);
3236 STRLEN two_byte_count = 0;
3238 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3240 /* See if really will need to convert to utf8. We mustn't rely on our
3241 * incoming SV being well formed and having a trailing '\0', as certain
3242 * code in pp_formline can send us partially built SVs. */
3246 if (NATIVE_IS_INVARIANT(ch)) continue;
3248 t--; /* t already incremented; re-point to first variant */
3253 /* utf8 conversion not needed because all are invariants. Mark as
3254 * UTF-8 even if no variant - saves scanning loop */
3260 /* Here, the string should be converted to utf8, either because of an
3261 * input flag (two_byte_count = 0), or because a character that
3262 * requires 2 bytes was found (two_byte_count = 1). t points either to
3263 * the beginning of the string (if we didn't examine anything), or to
3264 * the first variant. In either case, everything from s to t - 1 will
3265 * occupy only 1 byte each on output.
3267 * There are two main ways to convert. One is to create a new string
3268 * and go through the input starting from the beginning, appending each
3269 * converted value onto the new string as we go along. It's probably
3270 * best to allocate enough space in the string for the worst possible
3271 * case rather than possibly running out of space and having to
3272 * reallocate and then copy what we've done so far. Since everything
3273 * from s to t - 1 is invariant, the destination can be initialized
3274 * with these using a fast memory copy
3276 * The other way is to figure out exactly how big the string should be
3277 * by parsing the entire input. Then you don't have to make it big
3278 * enough to handle the worst possible case, and more importantly, if
3279 * the string you already have is large enough, you don't have to
3280 * allocate a new string, you can copy the last character in the input
3281 * string to the final position(s) that will be occupied by the
3282 * converted string and go backwards, stopping at t, since everything
3283 * before that is invariant.
3285 * There are advantages and disadvantages to each method.
3287 * In the first method, we can allocate a new string, do the memory
3288 * copy from the s to t - 1, and then proceed through the rest of the
3289 * string byte-by-byte.
3291 * In the second method, we proceed through the rest of the input
3292 * string just calculating how big the converted string will be. Then
3293 * there are two cases:
3294 * 1) if the string has enough extra space to handle the converted
3295 * value. We go backwards through the string, converting until we
3296 * get to the position we are at now, and then stop. If this
3297 * position is far enough along in the string, this method is
3298 * faster than the other method. If the memory copy were the same
3299 * speed as the byte-by-byte loop, that position would be about
3300 * half-way, as at the half-way mark, parsing to the end and back
3301 * is one complete string's parse, the same amount as starting
3302 * over and going all the way through. Actually, it would be
3303 * somewhat less than half-way, as it's faster to just count bytes
3304 * than to also copy, and we don't have the overhead of allocating
3305 * a new string, changing the scalar to use it, and freeing the
3306 * existing one. But if the memory copy is fast, the break-even
3307 * point is somewhere after half way. The counting loop could be
3308 * sped up by vectorization, etc, to move the break-even point
3309 * further towards the beginning.
3310 * 2) if the string doesn't have enough space to handle the converted
3311 * value. A new string will have to be allocated, and one might
3312 * as well, given that, start from the beginning doing the first
3313 * method. We've spent extra time parsing the string and in
3314 * exchange all we've gotten is that we know precisely how big to
3315 * make the new one. Perl is more optimized for time than space,
3316 * so this case is a loser.
3317 * So what I've decided to do is not use the 2nd method unless it is
3318 * guaranteed that a new string won't have to be allocated, assuming
3319 * the worst case. I also decided not to put any more conditions on it
3320 * than this, for now. It seems likely that, since the worst case is
3321 * twice as big as the unknown portion of the string (plus 1), we won't
3322 * be guaranteed enough space, causing us to go to the first method,
3323 * unless the string is short, or the first variant character is near
3324 * the end of it. In either of these cases, it seems best to use the
3325 * 2nd method. The only circumstance I can think of where this would
3326 * be really slower is if the string had once had much more data in it
3327 * than it does now, but there is still a substantial amount in it */
3330 STRLEN invariant_head = t - s;
3331 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3332 if (SvLEN(sv) < size) {
3334 /* Here, have decided to allocate a new string */
3339 Newx(dst, size, U8);
3341 /* If no known invariants at the beginning of the input string,
3342 * set so starts from there. Otherwise, can use memory copy to
3343 * get up to where we are now, and then start from here */
3345 if (invariant_head <= 0) {
3348 Copy(s, dst, invariant_head, char);
3349 d = dst + invariant_head;
3353 const UV uv = NATIVE8_TO_UNI(*t++);
3354 if (UNI_IS_INVARIANT(uv))
3355 *d++ = (U8)UNI_TO_NATIVE(uv);
3357 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3358 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3362 SvPV_free(sv); /* No longer using pre-existing string */
3363 SvPV_set(sv, (char*)dst);
3364 SvCUR_set(sv, d - dst);
3365 SvLEN_set(sv, size);
3368 /* Here, have decided to get the exact size of the string.
3369 * Currently this happens only when we know that there is
3370 * guaranteed enough space to fit the converted string, so
3371 * don't have to worry about growing. If two_byte_count is 0,
3372 * then t points to the first byte of the string which hasn't
3373 * been examined yet. Otherwise two_byte_count is 1, and t
3374 * points to the first byte in the string that will expand to
3375 * two. Depending on this, start examining at t or 1 after t.
3378 U8 *d = t + two_byte_count;
3381 /* Count up the remaining bytes that expand to two */
3384 const U8 chr = *d++;
3385 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3388 /* The string will expand by just the number of bytes that
3389 * occupy two positions. But we are one afterwards because of
3390 * the increment just above. This is the place to put the
3391 * trailing NUL, and to set the length before we decrement */
3393 d += two_byte_count;
3394 SvCUR_set(sv, d - s);
3398 /* Having decremented d, it points to the position to put the
3399 * very last byte of the expanded string. Go backwards through
3400 * the string, copying and expanding as we go, stopping when we
3401 * get to the part that is invariant the rest of the way down */
3405 const U8 ch = NATIVE8_TO_UNI(*e--);
3406 if (UNI_IS_INVARIANT(ch)) {
3407 *d-- = UNI_TO_NATIVE(ch);
3409 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3410 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3417 /* Mark as UTF-8 even if no variant - saves scanning loop */
3423 =for apidoc sv_utf8_downgrade
3425 Attempts to convert the PV of an SV from characters to bytes.
3426 If the PV contains a character that cannot fit
3427 in a byte, this conversion will fail;
3428 in this case, either returns false or, if C<fail_ok> is not
3431 This is not as a general purpose Unicode to byte encoding interface:
3432 use the Encode extension for that.
3438 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3442 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3444 if (SvPOKp(sv) && SvUTF8(sv)) {
3450 sv_force_normal_flags(sv, 0);
3452 s = (U8 *) SvPV(sv, len);
3453 if (!utf8_to_bytes(s, &len)) {
3458 Perl_croak(aTHX_ "Wide character in %s",
3461 Perl_croak(aTHX_ "Wide character");
3472 =for apidoc sv_utf8_encode
3474 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3475 flag off so that it looks like octets again.
3481 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3483 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3486 sv_force_normal_flags(sv, 0);
3488 if (SvREADONLY(sv)) {
3489 Perl_croak_no_modify(aTHX);
3491 (void) sv_utf8_upgrade(sv);
3496 =for apidoc sv_utf8_decode
3498 If the PV of the SV is an octet sequence in UTF-8
3499 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3500 so that it looks like a character. If the PV contains only single-byte
3501 characters, the C<SvUTF8> flag stays being off.
3502 Scans PV for validity and returns false if the PV is invalid UTF-8.
3508 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3510 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3516 /* The octets may have got themselves encoded - get them back as
3519 if (!sv_utf8_downgrade(sv, TRUE))
3522 /* it is actually just a matter of turning the utf8 flag on, but
3523 * we want to make sure everything inside is valid utf8 first.
3525 c = (const U8 *) SvPVX_const(sv);
3526 if (!is_utf8_string(c, SvCUR(sv)+1))
3528 e = (const U8 *) SvEND(sv);
3531 if (!UTF8_IS_INVARIANT(ch)) {
3541 =for apidoc sv_setsv
3543 Copies the contents of the source SV C<ssv> into the destination SV
3544 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3545 function if the source SV needs to be reused. Does not handle 'set' magic.
3546 Loosely speaking, it performs a copy-by-value, obliterating any previous
3547 content of the destination.
3549 You probably want to use one of the assortment of wrappers, such as
3550 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3551 C<SvSetMagicSV_nosteal>.
3553 =for apidoc sv_setsv_flags
3555 Copies the contents of the source SV C<ssv> into the destination SV
3556 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3557 function if the source SV needs to be reused. Does not handle 'set' magic.
3558 Loosely speaking, it performs a copy-by-value, obliterating any previous
3559 content of the destination.
3560 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3561 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3562 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3563 and C<sv_setsv_nomg> are implemented in terms of this function.
3565 You probably want to use one of the assortment of wrappers, such as
3566 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3567 C<SvSetMagicSV_nosteal>.
3569 This is the primary function for copying scalars, and most other
3570 copy-ish functions and macros use this underneath.
3576 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3578 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3579 HV *old_stash = NULL;
3581 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3583 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3584 const char * const name = GvNAME(sstr);
3585 const STRLEN len = GvNAMELEN(sstr);
3587 if (dtype >= SVt_PV) {
3593 SvUPGRADE(dstr, SVt_PVGV);
3594 (void)SvOK_off(dstr);
3595 /* FIXME - why are we doing this, then turning it off and on again
3597 isGV_with_GP_on(dstr);
3599 GvSTASH(dstr) = GvSTASH(sstr);
3601 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3602 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3603 SvFAKE_on(dstr); /* can coerce to non-glob */
3606 if(GvGP(MUTABLE_GV(sstr))) {
3607 /* If source has method cache entry, clear it */
3609 SvREFCNT_dec(GvCV(sstr));
3613 /* If source has a real method, then a method is
3615 else if(GvCV((const GV *)sstr)) {
3620 /* If dest already had a real method, that's a change as well */
3621 if(!mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)) {
3625 /* We don’t need to check the name of the destination if it was not a
3626 glob to begin with. */
3627 if(dtype == SVt_PVGV) {
3628 const char * const name = GvNAME((const GV *)dstr);
3629 if(strEQ(name,"ISA"))
3632 const STRLEN len = GvNAMELEN(dstr);
3633 if (len > 1 && name[len-2] == ':' && name[len-1] == ':') {
3636 /* Set aside the old stash, so we can reset isa caches on
3638 old_stash = GvHV(dstr);
3643 gp_free(MUTABLE_GV(dstr));
3644 isGV_with_GP_off(dstr);
3645 (void)SvOK_off(dstr);
3646 isGV_with_GP_on(dstr);
3647 GvINTRO_off(dstr); /* one-shot flag */
3648 GvGP(dstr) = gp_ref(GvGP(sstr));
3649 if (SvTAINTED(sstr))
3651 if (GvIMPORTED(dstr) != GVf_IMPORTED
3652 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3654 GvIMPORTED_on(dstr);
3657 if(mro_changes == 2) mro_isa_changed_in(GvSTASH(dstr));
3658 else if(mro_changes == 3) {
3659 HV * const stash = GvHV(dstr);
3660 if((stash && HvNAME(stash)) || (old_stash && HvNAME(old_stash)))
3662 stash && HvNAME(stash) ? stash : NULL,
3663 old_stash && HvNAME(old_stash) ? old_stash : NULL,
3667 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3672 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3674 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3676 const int intro = GvINTRO(dstr);
3679 const U32 stype = SvTYPE(sref);
3681 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3684 GvINTRO_off(dstr); /* one-shot flag */
3685 GvLINE(dstr) = CopLINE(PL_curcop);
3686 GvEGV(dstr) = MUTABLE_GV(dstr);
3691 location = (SV **) &GvCV(dstr);
3692 import_flag = GVf_IMPORTED_CV;
3695 location = (SV **) &GvHV(dstr);
3696 import_flag = GVf_IMPORTED_HV;
3699 location = (SV **) &GvAV(dstr);
3700 import_flag = GVf_IMPORTED_AV;
3703 location = (SV **) &GvIOp(dstr);
3706 location = (SV **) &GvFORM(dstr);
3709 location = &GvSV(dstr);
3710 import_flag = GVf_IMPORTED_SV;
3713 if (stype == SVt_PVCV) {
3714 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3715 if (GvCVGEN(dstr)) {
3716 SvREFCNT_dec(GvCV(dstr));
3718 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3721 SAVEGENERICSV(*location);
3725 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3726 CV* const cv = MUTABLE_CV(*location);
3728 if (!GvCVGEN((const GV *)dstr) &&
3729 (CvROOT(cv) || CvXSUB(cv)))
3731 /* Redefining a sub - warning is mandatory if
3732 it was a const and its value changed. */
3733 if (CvCONST(cv) && CvCONST((const CV *)sref)
3735 == cv_const_sv((const CV *)sref)) {
3737 /* They are 2 constant subroutines generated from
3738 the same constant. This probably means that
3739 they are really the "same" proxy subroutine
3740 instantiated in 2 places. Most likely this is
3741 when a constant is exported twice. Don't warn.
3744 else if (ckWARN(WARN_REDEFINE)
3746 && (!CvCONST((const CV *)sref)
3747 || sv_cmp(cv_const_sv(cv),
3748 cv_const_sv((const CV *)
3750 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3753 ? "Constant subroutine %s::%s redefined"
3754 : "Subroutine %s::%s redefined"),
3755 HvNAME_get(GvSTASH((const GV *)dstr)),
3756 GvENAME(MUTABLE_GV(dstr)));
3760 cv_ckproto_len(cv, (const GV *)dstr,
3761 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3762 SvPOK(sref) ? SvCUR(sref) : 0);
3764 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3765 GvASSUMECV_on(dstr);
3766 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3769 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3770 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3771 GvFLAGS(dstr) |= import_flag;
3773 if (stype == SVt_PVHV) {
3774 const char * const name = GvNAME((GV*)dstr);
3775 const STRLEN len = GvNAMELEN(dstr);
3777 len > 1 && name[len-2] == ':' && name[len-1] == ':'
3778 && (HvNAME(dref) || HvNAME(sref))
3781 HvNAME(sref) ? (HV *)sref : NULL,
3782 HvNAME(dref) ? (HV *)dref : NULL,
3787 else if (stype == SVt_PVAV && strEQ(GvNAME((GV*)dstr), "ISA")) {
3788 sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3789 mro_isa_changed_in(GvSTASH(dstr));
3794 if (SvTAINTED(sstr))
3800 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3803 register U32 sflags;
3805 register svtype stype;
3807 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3812 if (SvIS_FREED(dstr)) {
3813 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3814 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3816 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3818 sstr = &PL_sv_undef;
3819 if (SvIS_FREED(sstr)) {
3820 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3821 (void*)sstr, (void*)dstr);
3823 stype = SvTYPE(sstr);
3824 dtype = SvTYPE(dstr);
3826 (void)SvAMAGIC_off(dstr);
3829 /* need to nuke the magic */
3833 /* There's a lot of redundancy below but we're going for speed here */
3838 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3839 (void)SvOK_off(dstr);
3847 sv_upgrade(dstr, SVt_IV);
3851 sv_upgrade(dstr, SVt_PVIV);
3855 goto end_of_first_switch;
3857 (void)SvIOK_only(dstr);
3858 SvIV_set(dstr, SvIVX(sstr));
3861 /* SvTAINTED can only be true if the SV has taint magic, which in
3862 turn means that the SV type is PVMG (or greater). This is the
3863 case statement for SVt_IV, so this cannot be true (whatever gcov
3865 assert(!SvTAINTED(sstr));
3870 if (dtype < SVt_PV && dtype != SVt_IV)
3871 sv_upgrade(dstr, SVt_IV);
3879 sv_upgrade(dstr, SVt_NV);
3883 sv_upgrade(dstr, SVt_PVNV);
3887 goto end_of_first_switch;
3889 SvNV_set(dstr, SvNVX(sstr));
3890 (void)SvNOK_only(dstr);
3891 /* SvTAINTED can only be true if the SV has taint magic, which in
3892 turn means that the SV type is PVMG (or greater). This is the
3893 case statement for SVt_NV, so this cannot be true (whatever gcov
3895 assert(!SvTAINTED(sstr));
3901 #ifdef PERL_OLD_COPY_ON_WRITE
3902 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3903 if (dtype < SVt_PVIV)
3904 sv_upgrade(dstr, SVt_PVIV);
3911 sv_upgrade(dstr, SVt_PV);
3914 if (dtype < SVt_PVIV)
3915 sv_upgrade(dstr, SVt_PVIV);
3918 if (dtype < SVt_PVNV)
3919 sv_upgrade(dstr, SVt_PVNV);
3923 const char * const type = sv_reftype(sstr,0);
3925 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
3927 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3932 if (dtype < SVt_REGEXP)
3933 sv_upgrade(dstr, SVt_REGEXP);
3936 /* case SVt_BIND: */
3939 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
3940 glob_assign_glob(dstr, sstr, dtype);
3943 /* SvVALID means that this PVGV is playing at being an FBM. */
3947 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
3949 if (SvTYPE(sstr) != stype)
3950 stype = SvTYPE(sstr);
3951 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
3952 glob_assign_glob(dstr, sstr, dtype);
3956 if (stype == SVt_PVLV)
3957 SvUPGRADE(dstr, SVt_PVNV);
3959 SvUPGRADE(dstr, (svtype)stype);
3961 end_of_first_switch:
3963 /* dstr may have been upgraded. */
3964 dtype = SvTYPE(dstr);
3965 sflags = SvFLAGS(sstr);
3967 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
3968 /* Assigning to a subroutine sets the prototype. */
3971 const char *const ptr = SvPV_const(sstr, len);
3973 SvGROW(dstr, len + 1);
3974 Copy(ptr, SvPVX(dstr), len + 1, char);
3975 SvCUR_set(dstr, len);
3977 SvFLAGS(dstr) |= sflags & SVf_UTF8;
3981 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3982 const char * const type = sv_reftype(dstr,0);
3984 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
3986 Perl_croak(aTHX_ "Cannot copy to %s", type);
3987 } else if (sflags & SVf_ROK) {
3988 if (isGV_with_GP(dstr)
3989 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
3992 if (GvIMPORTED(dstr) != GVf_IMPORTED
3993 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3995 GvIMPORTED_on(dstr);
4000 glob_assign_glob(dstr, sstr, dtype);
4004 if (dtype >= SVt_PV) {
4005 if (isGV_with_GP(dstr)) {
4006 glob_assign_ref(dstr, sstr);
4009 if (SvPVX_const(dstr)) {
4015 (void)SvOK_off(dstr);
4016 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4017 SvFLAGS(dstr) |= sflags & SVf_ROK;
4018 assert(!(sflags & SVp_NOK));
4019 assert(!(sflags & SVp_IOK));
4020 assert(!(sflags & SVf_NOK));
4021 assert(!(sflags & SVf_IOK));
4023 else if (isGV_with_GP(dstr)) {
4024 if (!(sflags & SVf_OK)) {
4025 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4026 "Undefined value assigned to typeglob");
4029 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4030 if (dstr != (const SV *)gv) {
4031 const char * const name = GvNAME((const GV *)dstr);
4032 const STRLEN len = GvNAMELEN(dstr);
4033 HV *old_stash = NULL;
4034 bool reset_isa = FALSE;
4035 if (len > 1 && name[len-2] == ':' && name[len-1] == ':') {
4036 /* Set aside the old stash, so we can reset isa caches
4037 on its subclasses. */
4038 old_stash = GvHV(dstr);
4043 gp_free(MUTABLE_GV(dstr));
4044 GvGP(dstr) = gp_ref(GvGP(gv));
4047 HV * const stash = GvHV(dstr);
4049 (stash && HvNAME(stash))
4050 || (old_stash && HvNAME(old_stash))
4053 stash && HvNAME(stash) ? stash : NULL,
4054 old_stash && HvNAME(old_stash) ? old_stash : NULL,
4061 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4062 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4064 else if (sflags & SVp_POK) {
4068 * Check to see if we can just swipe the string. If so, it's a
4069 * possible small lose on short strings, but a big win on long ones.
4070 * It might even be a win on short strings if SvPVX_const(dstr)
4071 * has to be allocated and SvPVX_const(sstr) has to be freed.
4072 * Likewise if we can set up COW rather than doing an actual copy, we
4073 * drop to the else clause, as the swipe code and the COW setup code
4074 * have much in common.
4077 /* Whichever path we take through the next code, we want this true,
4078 and doing it now facilitates the COW check. */
4079 (void)SvPOK_only(dstr);
4082 /* If we're already COW then this clause is not true, and if COW
4083 is allowed then we drop down to the else and make dest COW
4084 with us. If caller hasn't said that we're allowed to COW
4085 shared hash keys then we don't do the COW setup, even if the
4086 source scalar is a shared hash key scalar. */
4087 (((flags & SV_COW_SHARED_HASH_KEYS)
4088 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4089 : 1 /* If making a COW copy is forbidden then the behaviour we
4090 desire is as if the source SV isn't actually already
4091 COW, even if it is. So we act as if the source flags
4092 are not COW, rather than actually testing them. */
4094 #ifndef PERL_OLD_COPY_ON_WRITE
4095 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4096 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4097 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4098 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4099 but in turn, it's somewhat dead code, never expected to go
4100 live, but more kept as a placeholder on how to do it better
4101 in a newer implementation. */
4102 /* If we are COW and dstr is a suitable target then we drop down
4103 into the else and make dest a COW of us. */
4104 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4109 (sflags & SVs_TEMP) && /* slated for free anyway? */
4110 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4111 (!(flags & SV_NOSTEAL)) &&
4112 /* and we're allowed to steal temps */
4113 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4114 SvLEN(sstr)) /* and really is a string */
4115 #ifdef PERL_OLD_COPY_ON_WRITE
4116 && ((flags & SV_COW_SHARED_HASH_KEYS)
4117 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4118 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4119 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4123 /* Failed the swipe test, and it's not a shared hash key either.
4124 Have to copy the string. */
4125 STRLEN len = SvCUR(sstr);
4126 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4127 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4128 SvCUR_set(dstr, len);
4129 *SvEND(dstr) = '\0';
4131 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4133 /* Either it's a shared hash key, or it's suitable for
4134 copy-on-write or we can swipe the string. */
4136 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4140 #ifdef PERL_OLD_COPY_ON_WRITE
4142 if ((sflags & (SVf_FAKE | SVf_READONLY))
4143 != (SVf_FAKE | SVf_READONLY)) {
4144 SvREADONLY_on(sstr);
4146 /* Make the source SV into a loop of 1.
4147 (about to become 2) */
4148 SV_COW_NEXT_SV_SET(sstr, sstr);
4152 /* Initial code is common. */
4153 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4158 /* making another shared SV. */
4159 STRLEN cur = SvCUR(sstr);
4160 STRLEN len = SvLEN(sstr);
4161 #ifdef PERL_OLD_COPY_ON_WRITE
4163 assert (SvTYPE(dstr) >= SVt_PVIV);
4164 /* SvIsCOW_normal */
4165 /* splice us in between source and next-after-source. */
4166 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4167 SV_COW_NEXT_SV_SET(sstr, dstr);
4168 SvPV_set(dstr, SvPVX_mutable(sstr));
4172 /* SvIsCOW_shared_hash */
4173 DEBUG_C(PerlIO_printf(Perl_debug_log,
4174 "Copy on write: Sharing hash\n"));
4176 assert (SvTYPE(dstr) >= SVt_PV);
4178 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4180 SvLEN_set(dstr, len);
4181 SvCUR_set(dstr, cur);
4182 SvREADONLY_on(dstr);
4186 { /* Passes the swipe test. */
4187 SvPV_set(dstr, SvPVX_mutable(sstr));
4188 SvLEN_set(dstr, SvLEN(sstr));
4189 SvCUR_set(dstr, SvCUR(sstr));
4192 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4193 SvPV_set(sstr, NULL);
4199 if (sflags & SVp_NOK) {
4200 SvNV_set(dstr, SvNVX(sstr));
4202 if (sflags & SVp_IOK) {
4203 SvIV_set(dstr, SvIVX(sstr));
4204 /* Must do this otherwise some other overloaded use of 0x80000000
4205 gets confused. I guess SVpbm_VALID */
4206 if (sflags & SVf_IVisUV)
4209 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4211 const MAGIC * const smg = SvVSTRING_mg(sstr);
4213 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4214 smg->mg_ptr, smg->mg_len);
4215 SvRMAGICAL_on(dstr);
4219 else if (sflags & (SVp_IOK|SVp_NOK)) {
4220 (void)SvOK_off(dstr);
4221 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4222 if (sflags & SVp_IOK) {
4223 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4224 SvIV_set(dstr, SvIVX(sstr));
4226 if (sflags & SVp_NOK) {
4227 SvNV_set(dstr, SvNVX(sstr));
4231 if (isGV_with_GP(sstr)) {
4232 /* This stringification rule for globs is spread in 3 places.
4233 This feels bad. FIXME. */
4234 const U32 wasfake = sflags & SVf_FAKE;
4236 /* FAKE globs can get coerced, so need to turn this off
4237 temporarily if it is on. */
4239 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4240 SvFLAGS(sstr) |= wasfake;
4243 (void)SvOK_off(dstr);
4245 if (SvTAINTED(sstr))
4250 =for apidoc sv_setsv_mg
4252 Like C<sv_setsv>, but also handles 'set' magic.
4258 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4260 PERL_ARGS_ASSERT_SV_SETSV_MG;
4262 sv_setsv(dstr,sstr);
4266 #ifdef PERL_OLD_COPY_ON_WRITE
4268 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4270 STRLEN cur = SvCUR(sstr);
4271 STRLEN len = SvLEN(sstr);
4272 register char *new_pv;
4274 PERL_ARGS_ASSERT_SV_SETSV_COW;
4277 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4278 (void*)sstr, (void*)dstr);
4285 if (SvTHINKFIRST(dstr))
4286 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4287 else if (SvPVX_const(dstr))
4288 Safefree(SvPVX_const(dstr));
4292 SvUPGRADE(dstr, SVt_PVIV);
4294 assert (SvPOK(sstr));
4295 assert (SvPOKp(sstr));
4296 assert (!SvIOK(sstr));
4297 assert (!SvIOKp(sstr));
4298 assert (!SvNOK(sstr));
4299 assert (!SvNOKp(sstr));
4301 if (SvIsCOW(sstr)) {
4303 if (SvLEN(sstr) == 0) {
4304 /* source is a COW shared hash key. */
4305 DEBUG_C(PerlIO_printf(Perl_debug_log,
4306 "Fast copy on write: Sharing hash\n"));
4307 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4310 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4312 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4313 SvUPGRADE(sstr, SVt_PVIV);
4314 SvREADONLY_on(sstr);
4316 DEBUG_C(PerlIO_printf(Perl_debug_log,
4317 "Fast copy on write: Converting sstr to COW\n"));
4318 SV_COW_NEXT_SV_SET(dstr, sstr);
4320 SV_COW_NEXT_SV_SET(sstr, dstr);
4321 new_pv = SvPVX_mutable(sstr);
4324 SvPV_set(dstr, new_pv);
4325 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4328 SvLEN_set(dstr, len);
4329 SvCUR_set(dstr, cur);
4338 =for apidoc sv_setpvn
4340 Copies a string into an SV. The C<len> parameter indicates the number of
4341 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4342 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4348 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4351 register char *dptr;
4353 PERL_ARGS_ASSERT_SV_SETPVN;
4355 SV_CHECK_THINKFIRST_COW_DROP(sv);
4361 /* len is STRLEN which is unsigned, need to copy to signed */
4364 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4366 SvUPGRADE(sv, SVt_PV);
4368 dptr = SvGROW(sv, len + 1);
4369 Move(ptr,dptr,len,char);
4372 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4377 =for apidoc sv_setpvn_mg
4379 Like C<sv_setpvn>, but also handles 'set' magic.
4385 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4387 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4389 sv_setpvn(sv,ptr,len);
4394 =for apidoc sv_setpv
4396 Copies a string into an SV. The string must be null-terminated. Does not
4397 handle 'set' magic. See C<sv_setpv_mg>.
4403 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4406 register STRLEN len;
4408 PERL_ARGS_ASSERT_SV_SETPV;
4410 SV_CHECK_THINKFIRST_COW_DROP(sv);
4416 SvUPGRADE(sv, SVt_PV);
4418 SvGROW(sv, len + 1);
4419 Move(ptr,SvPVX(sv),len+1,char);
4421 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4426 =for apidoc sv_setpv_mg
4428 Like C<sv_setpv>, but also handles 'set' magic.
4434 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4436 PERL_ARGS_ASSERT_SV_SETPV_MG;
4443 =for apidoc sv_usepvn_flags
4445 Tells an SV to use C<ptr> to find its string value. Normally the
4446 string is stored inside the SV but sv_usepvn allows the SV to use an
4447 outside string. The C<ptr> should point to memory that was allocated
4448 by C<malloc>. The string length, C<len>, must be supplied. By default
4449 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4450 so that pointer should not be freed or used by the programmer after
4451 giving it to sv_usepvn, and neither should any pointers from "behind"
4452 that pointer (e.g. ptr + 1) be used.
4454 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4455 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4456 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4457 C<len>, and already meets the requirements for storing in C<SvPVX>)
4463 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4468 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4470 SV_CHECK_THINKFIRST_COW_DROP(sv);
4471 SvUPGRADE(sv, SVt_PV);
4474 if (flags & SV_SMAGIC)
4478 if (SvPVX_const(sv))
4482 if (flags & SV_HAS_TRAILING_NUL)
4483 assert(ptr[len] == '\0');
4486 allocate = (flags & SV_HAS_TRAILING_NUL)
4488 #ifdef Perl_safesysmalloc_size
4491 PERL_STRLEN_ROUNDUP(len + 1);
4493 if (flags & SV_HAS_TRAILING_NUL) {
4494 /* It's long enough - do nothing.
4495 Specfically Perl_newCONSTSUB is relying on this. */
4498 /* Force a move to shake out bugs in callers. */
4499 char *new_ptr = (char*)safemalloc(allocate);
4500 Copy(ptr, new_ptr, len, char);
4501 PoisonFree(ptr,len,char);
4505 ptr = (char*) saferealloc (ptr, allocate);
4508 #ifdef Perl_safesysmalloc_size
4509 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4511 SvLEN_set(sv, allocate);
4515 if (!(flags & SV_HAS_TRAILING_NUL)) {
4518 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4520 if (flags & SV_SMAGIC)
4524 #ifdef PERL_OLD_COPY_ON_WRITE
4525 /* Need to do this *after* making the SV normal, as we need the buffer
4526 pointer to remain valid until after we've copied it. If we let go too early,
4527 another thread could invalidate it by unsharing last of the same hash key
4528 (which it can do by means other than releasing copy-on-write Svs)
4529 or by changing the other copy-on-write SVs in the loop. */
4531 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4533 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4535 { /* this SV was SvIsCOW_normal(sv) */
4536 /* we need to find the SV pointing to us. */
4537 SV *current = SV_COW_NEXT_SV(after);
4539 if (current == sv) {
4540 /* The SV we point to points back to us (there were only two of us
4542 Hence other SV is no longer copy on write either. */
4544 SvREADONLY_off(after);
4546 /* We need to follow the pointers around the loop. */
4548 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4551 /* don't loop forever if the structure is bust, and we have
4552 a pointer into a closed loop. */
4553 assert (current != after);
4554 assert (SvPVX_const(current) == pvx);
4556 /* Make the SV before us point to the SV after us. */
4557 SV_COW_NEXT_SV_SET(current, after);
4563 =for apidoc sv_force_normal_flags
4565 Undo various types of fakery on an SV: if the PV is a shared string, make
4566 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4567 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4568 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4569 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4570 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4571 set to some other value.) In addition, the C<flags> parameter gets passed to
4572 C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4573 with flags set to 0.
4579 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4583 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4585 #ifdef PERL_OLD_COPY_ON_WRITE
4586 if (SvREADONLY(sv)) {
4588 const char * const pvx = SvPVX_const(sv);
4589 const STRLEN len = SvLEN(sv);
4590 const STRLEN cur = SvCUR(sv);
4591 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4592 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4593 we'll fail an assertion. */
4594 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4597 PerlIO_printf(Perl_debug_log,
4598 "Copy on write: Force normal %ld\n",
4604 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4607 if (flags & SV_COW_DROP_PV) {
4608 /* OK, so we don't need to copy our buffer. */
4611 SvGROW(sv, cur + 1);
4612 Move(pvx,SvPVX(sv),cur,char);
4617 sv_release_COW(sv, pvx, next);
4619 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4625 else if (IN_PERL_RUNTIME)
4626 Perl_croak_no_modify(aTHX);
4629 if (SvREADONLY(sv)) {
4631 const char * const pvx = SvPVX_const(sv);
4632 const STRLEN len = SvCUR(sv);
4637 SvGROW(sv, len + 1);
4638 Move(pvx,SvPVX(sv),len,char);
4640 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4642 else if (IN_PERL_RUNTIME)
4643 Perl_croak_no_modify(aTHX);
4647 sv_unref_flags(sv, flags);
4648 else if (SvFAKE(sv) && isGV_with_GP(sv))
4650 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4651 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analagous
4652 to sv_unglob. We only need it here, so inline it. */
4653 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4654 SV *const temp = newSV_type(new_type);
4655 void *const temp_p = SvANY(sv);
4657 if (new_type == SVt_PVMG) {
4658 SvMAGIC_set(temp, SvMAGIC(sv));
4659 SvMAGIC_set(sv, NULL);
4660 SvSTASH_set(temp, SvSTASH(sv));
4661 SvSTASH_set(sv, NULL);
4663 SvCUR_set(temp, SvCUR(sv));
4664 /* Remember that SvPVX is in the head, not the body. */
4666 SvLEN_set(temp, SvLEN(sv));
4667 /* This signals "buffer is owned by someone else" in sv_clear,
4668 which is the least effort way to stop it freeing the buffer.
4670 SvLEN_set(sv, SvLEN(sv)+1);
4672 /* Their buffer is already owned by someone else. */
4673 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4674 SvLEN_set(temp, SvCUR(sv)+1);
4677 /* Now swap the rest of the bodies. */
4679 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4680 SvFLAGS(sv) |= new_type;
4681 SvANY(sv) = SvANY(temp);
4683 SvFLAGS(temp) &= ~(SVTYPEMASK);
4684 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4685 SvANY(temp) = temp_p;
4694 Efficient removal of characters from the beginning of the string buffer.
4695 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4696 the string buffer. The C<ptr> becomes the first character of the adjusted
4697 string. Uses the "OOK hack".
4698 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4699 refer to the same chunk of data.
4705 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4711 const U8 *real_start;
4715 PERL_ARGS_ASSERT_SV_CHOP;
4717 if (!ptr || !SvPOKp(sv))
4719 delta = ptr - SvPVX_const(sv);
4721 /* Nothing to do. */
4724 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4725 nothing uses the value of ptr any more. */
4726 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4727 if (ptr <= SvPVX_const(sv))
4728 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4729 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4730 SV_CHECK_THINKFIRST(sv);
4731 if (delta > max_delta)
4732 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4733 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4734 SvPVX_const(sv) + max_delta);
4737 if (!SvLEN(sv)) { /* make copy of shared string */
4738 const char *pvx = SvPVX_const(sv);
4739 const STRLEN len = SvCUR(sv);
4740 SvGROW(sv, len + 1);
4741 Move(pvx,SvPVX(sv),len,char);
4744 SvFLAGS(sv) |= SVf_OOK;
4747 SvOOK_offset(sv, old_delta);
4749 SvLEN_set(sv, SvLEN(sv) - delta);
4750 SvCUR_set(sv, SvCUR(sv) - delta);
4751 SvPV_set(sv, SvPVX(sv) + delta);
4753 p = (U8 *)SvPVX_const(sv);
4758 real_start = p - delta;
4762 if (delta < 0x100) {
4766 p -= sizeof(STRLEN);
4767 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4771 /* Fill the preceding buffer with sentinals to verify that no-one is
4773 while (p > real_start) {
4781 =for apidoc sv_catpvn
4783 Concatenates the string onto the end of the string which is in the SV. The
4784 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4785 status set, then the bytes appended should be valid UTF-8.
4786 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4788 =for apidoc sv_catpvn_flags
4790 Concatenates the string onto the end of the string which is in the SV. The
4791 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4792 status set, then the bytes appended should be valid UTF-8.
4793 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4794 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4795 in terms of this function.
4801 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4805 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4807 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4809 SvGROW(dsv, dlen + slen + 1);
4811 sstr = SvPVX_const(dsv);
4812 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4813 SvCUR_set(dsv, SvCUR(dsv) + slen);
4815 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4817 if (flags & SV_SMAGIC)
4822 =for apidoc sv_catsv
4824 Concatenates the string from SV C<ssv> onto the end of the string in
4825 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4826 not 'set' magic. See C<sv_catsv_mg>.
4828 =for apidoc sv_catsv_flags
4830 Concatenates the string from SV C<ssv> onto the end of the string in
4831 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4832 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4833 and C<sv_catsv_nomg> are implemented in terms of this function.
4838 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4842 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4846 const char *spv = SvPV_flags_const(ssv, slen, flags);
4848 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4849 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4850 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4851 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4852 dsv->sv_flags doesn't have that bit set.
4853 Andy Dougherty 12 Oct 2001
4855 const I32 sutf8 = DO_UTF8(ssv);
4858 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4860 dutf8 = DO_UTF8(dsv);
4862 if (dutf8 != sutf8) {
4864 /* Not modifying source SV, so taking a temporary copy. */
4865 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
4867 sv_utf8_upgrade(csv);
4868 spv = SvPV_const(csv, slen);
4871 /* Leave enough space for the cat that's about to happen */
4872 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
4874 sv_catpvn_nomg(dsv, spv, slen);
4877 if (flags & SV_SMAGIC)
4882 =for apidoc sv_catpv
4884 Concatenates the string onto the end of the string which is in the SV.
4885 If the SV has the UTF-8 status set, then the bytes appended should be
4886 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
4891 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
4894 register STRLEN len;
4898 PERL_ARGS_ASSERT_SV_CATPV;
4902 junk = SvPV_force(sv, tlen);
4904 SvGROW(sv, tlen + len + 1);
4906 ptr = SvPVX_const(sv);
4907 Move(ptr,SvPVX(sv)+tlen,len+1,char);
4908 SvCUR_set(sv, SvCUR(sv) + len);
4909 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4914 =for apidoc sv_catpv_flags
4916 Concatenates the string onto the end of the string which is in the SV.
4917 If the SV has the UTF-8 status set, then the bytes appended should
4918 be valid UTF-8. If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get>
4919 on the SVs if appropriate, else not.
4925 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, I32 flags)
4927 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
4928 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
4932 =for apidoc sv_catpv_mg
4934 Like C<sv_catpv>, but also handles 'set' magic.
4940 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4942 PERL_ARGS_ASSERT_SV_CATPV_MG;
4951 Creates a new SV. A non-zero C<len> parameter indicates the number of
4952 bytes of preallocated string space the SV should have. An extra byte for a
4953 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4954 space is allocated.) The reference count for the new SV is set to 1.
4956 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4957 parameter, I<x>, a debug aid which allowed callers to identify themselves.
4958 This aid has been superseded by a new build option, PERL_MEM_LOG (see
4959 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4960 modules supporting older perls.
4966 Perl_newSV(pTHX_ const STRLEN len)
4973 sv_upgrade(sv, SVt_PV);
4974 SvGROW(sv, len + 1);
4979 =for apidoc sv_magicext
4981 Adds magic to an SV, upgrading it if necessary. Applies the
4982 supplied vtable and returns a pointer to the magic added.
4984 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4985 In particular, you can add magic to SvREADONLY SVs, and add more than
4986 one instance of the same 'how'.
4988 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4989 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4990 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4991 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
4993 (This is now used as a subroutine by C<sv_magic>.)
4998 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
4999 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5004 PERL_ARGS_ASSERT_SV_MAGICEXT;
5006 SvUPGRADE(sv, SVt_PVMG);
5007 Newxz(mg, 1, MAGIC);
5008 mg->mg_moremagic = SvMAGIC(sv);
5009 SvMAGIC_set(sv, mg);
5011 /* Sometimes a magic contains a reference loop, where the sv and
5012 object refer to each other. To prevent a reference loop that
5013 would prevent such objects being freed, we look for such loops
5014 and if we find one we avoid incrementing the object refcount.
5016 Note we cannot do this to avoid self-tie loops as intervening RV must
5017 have its REFCNT incremented to keep it in existence.
5020 if (!obj || obj == sv ||
5021 how == PERL_MAGIC_arylen ||
5022 how == PERL_MAGIC_symtab ||
5023 (SvTYPE(obj) == SVt_PVGV &&
5024 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5025 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5026 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5031 mg->mg_obj = SvREFCNT_inc_simple(obj);
5032 mg->mg_flags |= MGf_REFCOUNTED;
5035 /* Normal self-ties simply pass a null object, and instead of
5036 using mg_obj directly, use the SvTIED_obj macro to produce a
5037 new RV as needed. For glob "self-ties", we are tieing the PVIO
5038 with an RV obj pointing to the glob containing the PVIO. In
5039 this case, to avoid a reference loop, we need to weaken the
5043 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5044 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5050 mg->mg_len = namlen;
5053 mg->mg_ptr = savepvn(name, namlen);
5054 else if (namlen == HEf_SVKEY) {
5055 /* Yes, this is casting away const. This is only for the case of
5056 HEf_SVKEY. I think we need to document this abberation of the
5057 constness of the API, rather than making name non-const, as
5058 that change propagating outwards a long way. */
5059 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5061 mg->mg_ptr = (char *) name;
5063 mg->mg_virtual = (MGVTBL *) vtable;
5067 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5072 =for apidoc sv_magic
5074 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
5075 then adds a new magic item of type C<how> to the head of the magic list.
5077 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5078 handling of the C<name> and C<namlen> arguments.
5080 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5081 to add more than one instance of the same 'how'.
5087 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5088 const char *const name, const I32 namlen)
5091 const MGVTBL *vtable;
5094 PERL_ARGS_ASSERT_SV_MAGIC;
5096 #ifdef PERL_OLD_COPY_ON_WRITE
5098 sv_force_normal_flags(sv, 0);
5100 if (SvREADONLY(sv)) {
5102 /* its okay to attach magic to shared strings; the subsequent
5103 * upgrade to PVMG will unshare the string */
5104 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5107 && how != PERL_MAGIC_regex_global
5108 && how != PERL_MAGIC_bm
5109 && how != PERL_MAGIC_fm
5110 && how != PERL_MAGIC_sv
5111 && how != PERL_MAGIC_backref
5114 Perl_croak_no_modify(aTHX);
5117 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5118 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5119 /* sv_magic() refuses to add a magic of the same 'how' as an
5122 if (how == PERL_MAGIC_taint) {
5124 /* Any scalar which already had taint magic on which someone
5125 (erroneously?) did SvIOK_on() or similar will now be
5126 incorrectly sporting public "OK" flags. */
5127 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5135 vtable = &PL_vtbl_sv;
5137 case PERL_MAGIC_overload:
5138 vtable = &PL_vtbl_amagic;
5140 case PERL_MAGIC_overload_elem:
5141 vtable = &PL_vtbl_amagicelem;
5143 case PERL_MAGIC_overload_table:
5144 vtable = &PL_vtbl_ovrld;
5147 vtable = &PL_vtbl_bm;
5149 case PERL_MAGIC_regdata:
5150 vtable = &PL_vtbl_regdata;
5152 case PERL_MAGIC_regdatum:
5153 vtable = &PL_vtbl_regdatum;
5155 case PERL_MAGIC_env:
5156 vtable = &PL_vtbl_env;
5159 vtable = &PL_vtbl_fm;
5161 case PERL_MAGIC_envelem:
5162 vtable = &PL_vtbl_envelem;
5164 case PERL_MAGIC_regex_global:
5165 vtable = &PL_vtbl_mglob;
5167 case PERL_MAGIC_isa:
5168 vtable = &PL_vtbl_isa;
5170 case PERL_MAGIC_isaelem:
5171 vtable = &PL_vtbl_isaelem;
5173 case PERL_MAGIC_nkeys:
5174 vtable = &PL_vtbl_nkeys;
5176 case PERL_MAGIC_dbfile:
5179 case PERL_MAGIC_dbline:
5180 vtable = &PL_vtbl_dbline;
5182 #ifdef USE_LOCALE_COLLATE
5183 case PERL_MAGIC_collxfrm:
5184 vtable = &PL_vtbl_collxfrm;
5186 #endif /* USE_LOCALE_COLLATE */
5187 case PERL_MAGIC_tied:
5188 vtable = &PL_vtbl_pack;
5190 case PERL_MAGIC_tiedelem:
5191 case PERL_MAGIC_tiedscalar:
5192 vtable = &PL_vtbl_packelem;
5195 vtable = &PL_vtbl_regexp;
5197 case PERL_MAGIC_sig:
5198 vtable = &PL_vtbl_sig;
5200 case PERL_MAGIC_sigelem:
5201 vtable = &PL_vtbl_sigelem;
5203 case PERL_MAGIC_taint:
5204 vtable = &PL_vtbl_taint;
5206 case PERL_MAGIC_uvar:
5207 vtable = &PL_vtbl_uvar;
5209 case PERL_MAGIC_vec:
5210 vtable = &PL_vtbl_vec;
5212 case PERL_MAGIC_arylen_p:
5213 case PERL_MAGIC_rhash:
5214 case PERL_MAGIC_symtab:
5215 case PERL_MAGIC_vstring:
5216 case PERL_MAGIC_checkcall:
5219 case PERL_MAGIC_utf8:
5220 vtable = &PL_vtbl_utf8;
5222 case PERL_MAGIC_substr:
5223 vtable = &PL_vtbl_substr;
5225 case PERL_MAGIC_defelem:
5226 vtable = &PL_vtbl_defelem;
5228 case PERL_MAGIC_arylen:
5229 vtable = &PL_vtbl_arylen;
5231 case PERL_MAGIC_pos:
5232 vtable = &PL_vtbl_pos;
5234 case PERL_MAGIC_backref:
5235 vtable = &PL_vtbl_backref;
5237 case PERL_MAGIC_hintselem:
5238 vtable = &PL_vtbl_hintselem;
5240 case PERL_MAGIC_hints:
5241 vtable = &PL_vtbl_hints;
5243 case PERL_MAGIC_ext:
5244 /* Reserved for use by extensions not perl internals. */
5245 /* Useful for attaching extension internal data to perl vars. */
5246 /* Note that multiple extensions may clash if magical scalars */
5247 /* etc holding private data from one are passed to another. */
5251 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5254 /* Rest of work is done else where */
5255 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5258 case PERL_MAGIC_taint:
5261 case PERL_MAGIC_ext:
5262 case PERL_MAGIC_dbfile:
5269 =for apidoc sv_unmagic
5271 Removes all magic of type C<type> from an SV.
5277 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5282 PERL_ARGS_ASSERT_SV_UNMAGIC;
5284 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5286 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5287 for (mg = *mgp; mg; mg = *mgp) {
5288 if (mg->mg_type == type) {
5289 const MGVTBL* const vtbl = mg->mg_virtual;
5290 *mgp = mg->mg_moremagic;
5291 if (vtbl && vtbl->svt_free)
5292 vtbl->svt_free(aTHX_ sv, mg);
5293 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5295 Safefree(mg->mg_ptr);
5296 else if (mg->mg_len == HEf_SVKEY)
5297 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5298 else if (mg->mg_type == PERL_MAGIC_utf8)
5299 Safefree(mg->mg_ptr);
5301 if (mg->mg_flags & MGf_REFCOUNTED)
5302 SvREFCNT_dec(mg->mg_obj);
5306 mgp = &mg->mg_moremagic;
5309 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5310 mg_magical(sv); /* else fix the flags now */
5314 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5320 =for apidoc sv_rvweaken
5322 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5323 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5324 push a back-reference to this RV onto the array of backreferences
5325 associated with that magic. If the RV is magical, set magic will be
5326 called after the RV is cleared.
5332 Perl_sv_rvweaken(pTHX_ SV *const sv)
5336 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5338 if (!SvOK(sv)) /* let undefs pass */
5341 Perl_croak(aTHX_ "Can't weaken a nonreference");
5342 else if (SvWEAKREF(sv)) {
5343 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5347 Perl_sv_add_backref(aTHX_ tsv, sv);
5353 /* Give tsv backref magic if it hasn't already got it, then push a
5354 * back-reference to sv onto the array associated with the backref magic.
5356 * As an optimisation, if there's only one backref and it's not an AV,
5357 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5358 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5361 * If an HV's backref is stored in magic, it is moved back to HvAUX.
5364 /* A discussion about the backreferences array and its refcount:
5366 * The AV holding the backreferences is pointed to either as the mg_obj of
5367 * PERL_MAGIC_backref, or in the specific case of a HV that has the hv_aux
5368 * structure, from the xhv_backreferences field. (A HV without hv_aux will
5369 * have the standard magic instead.) The array is created with a refcount
5370 * of 2. This means that if during global destruction the array gets
5371 * picked on before its parent to have its refcount decremented by the
5372 * random zapper, it won't actually be freed, meaning it's still there for
5373 * when its parent gets freed.
5375 * When the parent SV is freed, the extra ref is killed by
5376 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5377 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5379 * When a single backref SV is stored directly, it is not reference
5384 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5391 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5393 /* find slot to store array or singleton backref */
5395 if (SvTYPE(tsv) == SVt_PVHV) {
5396 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5399 if ((mg = mg_find(tsv, PERL_MAGIC_backref))) {
5400 /* Aha. They've got it stowed in magic instead.
5401 * Move it back to xhv_backreferences */
5403 /* Stop mg_free decreasing the reference count. */
5405 /* Stop mg_free even calling the destructor, given that
5406 there's no AV to free up. */
5408 sv_unmagic(tsv, PERL_MAGIC_backref);
5414 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5416 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5417 mg = mg_find(tsv, PERL_MAGIC_backref);
5419 svp = &(mg->mg_obj);
5422 /* create or retrieve the array */
5424 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5425 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5430 SvREFCNT_inc_simple_void(av);
5431 /* av now has a refcnt of 2; see discussion above */
5433 /* move single existing backref to the array */
5435 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5439 mg->mg_flags |= MGf_REFCOUNTED;
5442 av = MUTABLE_AV(*svp);
5445 /* optimisation: store single backref directly in HvAUX or mg_obj */
5449 /* push new backref */
5450 assert(SvTYPE(av) == SVt_PVAV);
5451 if (AvFILLp(av) >= AvMAX(av)) {
5452 av_extend(av, AvFILLp(av)+1);
5454 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5457 /* delete a back-reference to ourselves from the backref magic associated
5458 * with the SV we point to.
5462 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5468 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5470 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
5471 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5473 if (!svp || !*svp) {
5475 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5476 svp = mg ? &(mg->mg_obj) : NULL;
5480 Perl_croak(aTHX_ "panic: del_backref");
5482 if (SvTYPE(*svp) == SVt_PVAV) {
5484 AV * const av = (AV*)*svp;
5485 assert(!SvIS_FREED(av));
5487 for (i = AvFILLp(av); i >= 0; i--) {
5489 const SSize_t fill = AvFILLp(av);
5491 /* We weren't the last entry.
5492 An unordered list has this property that you can take the
5493 last element off the end to fill the hole, and it's still
5494 an unordered list :-)
5499 AvFILLp(av) = fill - 1;
5502 break; /* should only be one */
5509 /* optimisation: only a single backref, stored directly */
5511 Perl_croak(aTHX_ "panic: del_backref");
5518 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5524 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5529 is_array = (SvTYPE(av) == SVt_PVAV);
5531 assert(!SvIS_FREED(av));
5534 last = svp + AvFILLp(av);
5537 /* optimisation: only a single backref, stored directly */
5543 while (svp <= last) {
5545 SV *const referrer = *svp;
5546 if (SvWEAKREF(referrer)) {
5547 /* XXX Should we check that it hasn't changed? */
5548 assert(SvROK(referrer));
5549 SvRV_set(referrer, 0);
5551 SvWEAKREF_off(referrer);
5552 SvSETMAGIC(referrer);
5553 } else if (SvTYPE(referrer) == SVt_PVGV ||
5554 SvTYPE(referrer) == SVt_PVLV) {
5555 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5556 /* You lookin' at me? */
5557 assert(GvSTASH(referrer));
5558 assert(GvSTASH(referrer) == (const HV *)sv);
5559 GvSTASH(referrer) = 0;
5560 } else if (SvTYPE(referrer) == SVt_PVCV ||
5561 SvTYPE(referrer) == SVt_PVFM) {
5562 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5563 /* You lookin' at me? */
5564 assert(CvSTASH(referrer));
5565 assert(CvSTASH(referrer) == (const HV *)sv);
5566 CvSTASH(referrer) = 0;
5569 assert(SvTYPE(sv) == SVt_PVGV);
5570 /* You lookin' at me? */
5571 assert(CvGV(referrer));
5572 assert(CvGV(referrer) == (const GV *)sv);
5573 anonymise_cv_maybe(MUTABLE_GV(sv),
5574 MUTABLE_CV(referrer));
5579 "panic: magic_killbackrefs (flags=%"UVxf")",
5580 (UV)SvFLAGS(referrer));
5591 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5597 =for apidoc sv_insert
5599 Inserts a string at the specified offset/length within the SV. Similar to
5600 the Perl substr() function. Handles get magic.
5602 =for apidoc sv_insert_flags
5604 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5610 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5615 register char *midend;
5616 register char *bigend;
5620 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5623 Perl_croak(aTHX_ "Can't modify non-existent substring");
5624 SvPV_force_flags(bigstr, curlen, flags);
5625 (void)SvPOK_only_UTF8(bigstr);
5626 if (offset + len > curlen) {
5627 SvGROW(bigstr, offset+len+1);
5628 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5629 SvCUR_set(bigstr, offset+len);
5633 i = littlelen - len;
5634 if (i > 0) { /* string might grow */
5635 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5636 mid = big + offset + len;
5637 midend = bigend = big + SvCUR(bigstr);
5640 while (midend > mid) /* shove everything down */
5641 *--bigend = *--midend;
5642 Move(little,big+offset,littlelen,char);
5643 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5648 Move(little,SvPVX(bigstr)+offset,len,char);
5653 big = SvPVX(bigstr);
5656 bigend = big + SvCUR(bigstr);
5658 if (midend > bigend)
5659 Perl_croak(aTHX_ "panic: sv_insert");
5661 if (mid - big > bigend - midend) { /* faster to shorten from end */
5663 Move(little, mid, littlelen,char);
5666 i = bigend - midend;
5668 Move(midend, mid, i,char);
5672 SvCUR_set(bigstr, mid - big);
5674 else if ((i = mid - big)) { /* faster from front */
5675 midend -= littlelen;
5677 Move(big, midend - i, i, char);
5678 sv_chop(bigstr,midend-i);
5680 Move(little, mid, littlelen,char);
5682 else if (littlelen) {
5683 midend -= littlelen;
5684 sv_chop(bigstr,midend);
5685 Move(little,midend,littlelen,char);
5688 sv_chop(bigstr,midend);
5694 =for apidoc sv_replace
5696 Make the first argument a copy of the second, then delete the original.
5697 The target SV physically takes over ownership of the body of the source SV
5698 and inherits its flags; however, the target keeps any magic it owns,
5699 and any magic in the source is discarded.
5700 Note that this is a rather specialist SV copying operation; most of the
5701 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5707 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5710 const U32 refcnt = SvREFCNT(sv);
5712 PERL_ARGS_ASSERT_SV_REPLACE;
5714 SV_CHECK_THINKFIRST_COW_DROP(sv);
5715 if (SvREFCNT(nsv) != 1) {
5716 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5717 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5719 if (SvMAGICAL(sv)) {
5723 sv_upgrade(nsv, SVt_PVMG);
5724 SvMAGIC_set(nsv, SvMAGIC(sv));
5725 SvFLAGS(nsv) |= SvMAGICAL(sv);
5727 SvMAGIC_set(sv, NULL);
5731 assert(!SvREFCNT(sv));
5732 #ifdef DEBUG_LEAKING_SCALARS
5733 sv->sv_flags = nsv->sv_flags;
5734 sv->sv_any = nsv->sv_any;
5735 sv->sv_refcnt = nsv->sv_refcnt;
5736 sv->sv_u = nsv->sv_u;
5738 StructCopy(nsv,sv,SV);
5740 if(SvTYPE(sv) == SVt_IV) {
5742 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5746 #ifdef PERL_OLD_COPY_ON_WRITE
5747 if (SvIsCOW_normal(nsv)) {
5748 /* We need to follow the pointers around the loop to make the
5749 previous SV point to sv, rather than nsv. */
5752 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5755 assert(SvPVX_const(current) == SvPVX_const(nsv));
5757 /* Make the SV before us point to the SV after us. */
5759 PerlIO_printf(Perl_debug_log, "previous is\n");
5761 PerlIO_printf(Perl_debug_log,
5762 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5763 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5765 SV_COW_NEXT_SV_SET(current, sv);
5768 SvREFCNT(sv) = refcnt;
5769 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5774 /* We're about to free a GV which has a CV that refers back to us.
5775 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5779 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5785 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5788 assert(SvREFCNT(gv) == 0);
5789 assert(isGV(gv) && isGV_with_GP(gv));
5791 assert(!CvANON(cv));
5792 assert(CvGV(cv) == gv);
5794 /* will the CV shortly be freed by gp_free() ? */
5795 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5796 SvANY(cv)->xcv_gv = NULL;
5800 /* if not, anonymise: */
5801 stash = GvSTASH(gv) ? HvNAME(GvSTASH(gv)) : NULL;
5802 gvname = Perl_newSVpvf(aTHX_ "%s::__ANON__",
5803 stash ? stash : "__ANON__");
5804 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5805 SvREFCNT_dec(gvname);
5809 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5814 =for apidoc sv_clear
5816 Clear an SV: call any destructors, free up any memory used by the body,
5817 and free the body itself. The SV's head is I<not> freed, although
5818 its type is set to all 1's so that it won't inadvertently be assumed
5819 to be live during global destruction etc.
5820 This function should only be called when REFCNT is zero. Most of the time
5821 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5828 Perl_sv_clear(pTHX_ SV *const orig_sv)
5833 const struct body_details *sv_type_details;
5836 register SV *sv = orig_sv;
5838 PERL_ARGS_ASSERT_SV_CLEAR;
5840 /* within this loop, sv is the SV currently being freed, and
5841 * iter_sv is the most recent AV or whatever that's being iterated
5842 * over to provide more SVs */
5848 assert(SvREFCNT(sv) == 0);
5849 assert(SvTYPE(sv) != SVTYPEMASK);
5851 if (type <= SVt_IV) {
5852 /* See the comment in sv.h about the collusion between this
5853 * early return and the overloading of the NULL slots in the
5857 SvFLAGS(sv) &= SVf_BREAK;
5858 SvFLAGS(sv) |= SVTYPEMASK;
5863 if (PL_defstash && /* Still have a symbol table? */
5870 stash = SvSTASH(sv);
5871 destructor = StashHANDLER(stash,DESTROY);
5873 /* A constant subroutine can have no side effects, so
5874 don't bother calling it. */
5875 && !CvCONST(destructor)
5876 /* Don't bother calling an empty destructor */
5877 && (CvISXSUB(destructor)
5878 || (CvSTART(destructor)
5879 && (CvSTART(destructor)->op_next->op_type
5882 SV* const tmpref = newRV(sv);
5883 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
5885 PUSHSTACKi(PERLSI_DESTROY);
5890 call_sv(MUTABLE_SV(destructor),
5891 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
5895 if(SvREFCNT(tmpref) < 2) {
5896 /* tmpref is not kept alive! */
5898 SvRV_set(tmpref, NULL);
5901 SvREFCNT_dec(tmpref);
5903 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
5907 if (PL_in_clean_objs)
5909 "DESTROY created new reference to dead object '%s'",
5911 /* DESTROY gave object new lease on life */
5917 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
5918 SvOBJECT_off(sv); /* Curse the object. */
5919 if (type != SVt_PVIO)
5920 --PL_sv_objcount;/* XXX Might want something more general */
5923 if (type >= SVt_PVMG) {
5924 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5925 SvREFCNT_dec(SvOURSTASH(sv));
5926 } else if (SvMAGIC(sv))
5928 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5929 SvREFCNT_dec(SvSTASH(sv));
5932 /* case SVt_BIND: */
5935 IoIFP(sv) != PerlIO_stdin() &&
5936 IoIFP(sv) != PerlIO_stdout() &&
5937 IoIFP(sv) != PerlIO_stderr() &&
5938 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5940 io_close(MUTABLE_IO(sv), FALSE);
5942 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5943 PerlDir_close(IoDIRP(sv));
5944 IoDIRP(sv) = (DIR*)NULL;
5945 Safefree(IoTOP_NAME(sv));
5946 Safefree(IoFMT_NAME(sv));
5947 Safefree(IoBOTTOM_NAME(sv));
5950 /* FIXME for plugins */
5951 pregfree2((REGEXP*) sv);
5955 cv_undef(MUTABLE_CV(sv));
5956 /* If we're in a stash, we don't own a reference to it.
5957 * However it does have a back reference to us, which needs to
5959 if ((stash = CvSTASH(sv)))
5960 sv_del_backref(MUTABLE_SV(stash), sv);
5963 if (PL_last_swash_hv == (const HV *)sv) {
5964 PL_last_swash_hv = NULL;
5966 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5967 hv_undef(MUTABLE_HV(sv));
5971 AV* av = MUTABLE_AV(sv);
5972 if (PL_comppad == av) {
5976 if (AvREAL(av) && AvFILLp(av) > -1) {
5977 next_sv = AvARRAY(av)[AvFILLp(av)--];
5978 /* save old iter_sv in top-most slot of AV,
5979 * and pray that it doesn't get wiped in the meantime */
5980 AvARRAY(av)[AvMAX(av)] = iter_sv;
5982 goto get_next_sv; /* process this new sv */
5984 Safefree(AvALLOC(av));
5989 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5990 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5991 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5992 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5994 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5995 SvREFCNT_dec(LvTARG(sv));
5997 if (isGV_with_GP(sv)) {
5998 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
5999 && HvNAME_get(stash))
6000 mro_method_changed_in(stash);
6001 gp_free(MUTABLE_GV(sv));
6003 unshare_hek(GvNAME_HEK(sv));
6004 /* If we're in a stash, we don't own a reference to it.
6005 * However it does have a back reference to us, which
6006 * needs to be cleared. */
6007 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6008 sv_del_backref(MUTABLE_SV(stash), sv);
6010 /* FIXME. There are probably more unreferenced pointers to SVs
6011 * in the interpreter struct that we should check and tidy in
6012 * a similar fashion to this: */
6013 if ((const GV *)sv == PL_last_in_gv)
6014 PL_last_in_gv = NULL;
6020 /* Don't bother with SvOOK_off(sv); as we're only going to
6024 SvOOK_offset(sv, offset);
6025 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6026 /* Don't even bother with turning off the OOK flag. */
6031 SV * const target = SvRV(sv);
6033 sv_del_backref(target, sv);
6038 #ifdef PERL_OLD_COPY_ON_WRITE
6039 else if (SvPVX_const(sv)
6040 && !(SvTYPE(sv) == SVt_PVIO
6041 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6045 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6049 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6051 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6055 } else if (SvLEN(sv)) {
6056 Safefree(SvPVX_const(sv));
6060 else if (SvPVX_const(sv) && SvLEN(sv)
6061 && !(SvTYPE(sv) == SVt_PVIO
6062 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6063 Safefree(SvPVX_mutable(sv));
6064 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
6065 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6076 SvFLAGS(sv) &= SVf_BREAK;
6077 SvFLAGS(sv) |= SVTYPEMASK;
6079 sv_type_details = bodies_by_type + type;
6080 if (sv_type_details->arena) {
6081 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6082 &PL_body_roots[type]);
6084 else if (sv_type_details->body_size) {
6085 safefree(SvANY(sv));
6089 /* caller is responsible for freeing the head of the original sv */
6090 if (sv != orig_sv && !SvREFCNT(sv))
6093 /* grab and free next sv, if any */
6101 else if (!iter_sv) {
6103 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6104 AV *const av = (AV*)iter_sv;
6105 if (AvFILLp(av) > -1) {
6106 sv = AvARRAY(av)[AvFILLp(av)--];
6108 else { /* no more elements of current AV to free */
6111 /* restore previous value, squirrelled away */
6112 iter_sv = AvARRAY(av)[AvMAX(av)];
6113 Safefree(AvALLOC(av));
6118 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6122 if (!SvREFCNT(sv)) {
6126 if (--(SvREFCNT(sv)))
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;
6148 =for apidoc sv_newref
6150 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6157 Perl_sv_newref(pTHX_ SV *const sv)
6159 PERL_UNUSED_CONTEXT;
6168 Decrement an SV's reference count, and if it drops to zero, call
6169 C<sv_clear> to invoke destructors and free up any memory used by
6170 the body; finally, deallocate the SV's head itself.
6171 Normally called via a wrapper macro C<SvREFCNT_dec>.
6177 Perl_sv_free(pTHX_ SV *const sv)
6182 if (SvREFCNT(sv) == 0) {
6183 if (SvFLAGS(sv) & SVf_BREAK)
6184 /* this SV's refcnt has been artificially decremented to
6185 * trigger cleanup */
6187 if (PL_in_clean_all) /* All is fair */
6189 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6190 /* make sure SvREFCNT(sv)==0 happens very seldom */
6191 SvREFCNT(sv) = (~(U32)0)/2;
6194 if (ckWARN_d(WARN_INTERNAL)) {
6195 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6196 Perl_dump_sv_child(aTHX_ sv);
6198 #ifdef DEBUG_LEAKING_SCALARS
6201 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6202 if (PL_warnhook == PERL_WARNHOOK_FATAL
6203 || ckDEAD(packWARN(WARN_INTERNAL))) {
6204 /* Don't let Perl_warner cause us to escape our fate: */
6208 /* This may not return: */
6209 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6210 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6211 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6214 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6219 if (--(SvREFCNT(sv)) > 0)
6221 Perl_sv_free2(aTHX_ sv);
6225 Perl_sv_free2(pTHX_ SV *const sv)
6229 PERL_ARGS_ASSERT_SV_FREE2;
6233 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6234 "Attempt to free temp prematurely: SV 0x%"UVxf
6235 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6239 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6240 /* make sure SvREFCNT(sv)==0 happens very seldom */
6241 SvREFCNT(sv) = (~(U32)0)/2;
6252 Returns the length of the string in the SV. Handles magic and type
6253 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6259 Perl_sv_len(pTHX_ register SV *const sv)
6267 len = mg_length(sv);
6269 (void)SvPV_const(sv, len);
6274 =for apidoc sv_len_utf8
6276 Returns the number of characters in the string in an SV, counting wide
6277 UTF-8 bytes as a single character. Handles magic and type coercion.
6283 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6284 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6285 * (Note that the mg_len is not the length of the mg_ptr field.
6286 * This allows the cache to store the character length of the string without
6287 * needing to malloc() extra storage to attach to the mg_ptr.)
6292 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6298 return mg_length(sv);
6302 const U8 *s = (U8*)SvPV_const(sv, len);
6306 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6308 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6309 if (mg->mg_len != -1)
6312 /* We can use the offset cache for a headstart.
6313 The longer value is stored in the first pair. */
6314 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6316 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6320 if (PL_utf8cache < 0) {
6321 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6322 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6326 ulen = Perl_utf8_length(aTHX_ s, s + len);
6327 utf8_mg_len_cache_update(sv, &mg, ulen);
6331 return Perl_utf8_length(aTHX_ s, s + len);
6335 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6338 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6339 STRLEN *const uoffset_p, bool *const at_end)
6341 const U8 *s = start;
6342 STRLEN uoffset = *uoffset_p;
6344 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6346 while (s < send && uoffset) {
6353 else if (s > send) {
6355 /* This is the existing behaviour. Possibly it should be a croak, as
6356 it's actually a bounds error */
6359 *uoffset_p -= uoffset;
6363 /* Given the length of the string in both bytes and UTF-8 characters, decide
6364 whether to walk forwards or backwards to find the byte corresponding to
6365 the passed in UTF-8 offset. */
6367 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6368 STRLEN uoffset, const STRLEN uend)
6370 STRLEN backw = uend - uoffset;
6372 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6374 if (uoffset < 2 * backw) {
6375 /* The assumption is that going forwards is twice the speed of going
6376 forward (that's where the 2 * backw comes from).
6377 (The real figure of course depends on the UTF-8 data.) */
6378 const U8 *s = start;
6380 while (s < send && uoffset--)
6390 while (UTF8_IS_CONTINUATION(*send))
6393 return send - start;
6396 /* For the string representation of the given scalar, find the byte
6397 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6398 give another position in the string, *before* the sought offset, which
6399 (which is always true, as 0, 0 is a valid pair of positions), which should
6400 help reduce the amount of linear searching.
6401 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6402 will be used to reduce the amount of linear searching. The cache will be
6403 created if necessary, and the found value offered to it for update. */
6405 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6406 const U8 *const send, STRLEN uoffset,
6407 STRLEN uoffset0, STRLEN boffset0)
6409 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6411 bool at_end = FALSE;
6413 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6415 assert (uoffset >= uoffset0);
6422 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6423 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6424 if ((*mgp)->mg_ptr) {
6425 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6426 if (cache[0] == uoffset) {
6427 /* An exact match. */
6430 if (cache[2] == uoffset) {
6431 /* An exact match. */
6435 if (cache[0] < uoffset) {
6436 /* The cache already knows part of the way. */
6437 if (cache[0] > uoffset0) {
6438 /* The cache knows more than the passed in pair */
6439 uoffset0 = cache[0];
6440 boffset0 = cache[1];
6442 if ((*mgp)->mg_len != -1) {
6443 /* And we know the end too. */
6445 + sv_pos_u2b_midway(start + boffset0, send,
6447 (*mgp)->mg_len - uoffset0);
6449 uoffset -= uoffset0;
6451 + sv_pos_u2b_forwards(start + boffset0,
6452 send, &uoffset, &at_end);
6453 uoffset += uoffset0;
6456 else if (cache[2] < uoffset) {
6457 /* We're between the two cache entries. */
6458 if (cache[2] > uoffset0) {
6459 /* and the cache knows more than the passed in pair */
6460 uoffset0 = cache[2];
6461 boffset0 = cache[3];
6465 + sv_pos_u2b_midway(start + boffset0,
6468 cache[0] - uoffset0);
6471 + sv_pos_u2b_midway(start + boffset0,
6474 cache[2] - uoffset0);
6478 else if ((*mgp)->mg_len != -1) {
6479 /* If we can take advantage of a passed in offset, do so. */
6480 /* In fact, offset0 is either 0, or less than offset, so don't
6481 need to worry about the other possibility. */
6483 + sv_pos_u2b_midway(start + boffset0, send,
6485 (*mgp)->mg_len - uoffset0);
6490 if (!found || PL_utf8cache < 0) {
6491 STRLEN real_boffset;
6492 uoffset -= uoffset0;
6493 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6494 send, &uoffset, &at_end);
6495 uoffset += uoffset0;
6497 if (found && PL_utf8cache < 0)
6498 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6500 boffset = real_boffset;
6505 utf8_mg_len_cache_update(sv, mgp, uoffset);
6507 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6514 =for apidoc sv_pos_u2b_flags
6516 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6517 the start of the string, to a count of the equivalent number of bytes; if
6518 lenp is non-zero, it does the same to lenp, but this time starting from
6519 the offset, rather than from the start of the string. Handles type coercion.
6520 I<flags> is passed to C<SvPV_flags>, and usually should be
6521 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6527 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6528 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6529 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6534 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6541 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6543 start = (U8*)SvPV_flags(sv, len, flags);
6545 const U8 * const send = start + len;
6547 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6550 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6551 is 0, and *lenp is already set to that. */) {
6552 /* Convert the relative offset to absolute. */
6553 const STRLEN uoffset2 = uoffset + *lenp;
6554 const STRLEN boffset2
6555 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6556 uoffset, boffset) - boffset;
6570 =for apidoc sv_pos_u2b
6572 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6573 the start of the string, to a count of the equivalent number of bytes; if
6574 lenp is non-zero, it does the same to lenp, but this time starting from
6575 the offset, rather than from the start of the string. Handles magic and
6578 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6585 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6586 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6587 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6591 /* This function is subject to size and sign problems */
6594 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6596 PERL_ARGS_ASSERT_SV_POS_U2B;
6599 STRLEN ulen = (STRLEN)*lenp;
6600 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6601 SV_GMAGIC|SV_CONST_RETURN);
6604 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6605 SV_GMAGIC|SV_CONST_RETURN);
6610 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6613 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6617 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6618 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6619 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6623 (*mgp)->mg_len = ulen;
6624 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6625 if (ulen != (STRLEN) (*mgp)->mg_len)
6626 (*mgp)->mg_len = -1;
6629 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6630 byte length pairing. The (byte) length of the total SV is passed in too,
6631 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6632 may not have updated SvCUR, so we can't rely on reading it directly.
6634 The proffered utf8/byte length pairing isn't used if the cache already has
6635 two pairs, and swapping either for the proffered pair would increase the
6636 RMS of the intervals between known byte offsets.
6638 The cache itself consists of 4 STRLEN values
6639 0: larger UTF-8 offset
6640 1: corresponding byte offset
6641 2: smaller UTF-8 offset
6642 3: corresponding byte offset
6644 Unused cache pairs have the value 0, 0.
6645 Keeping the cache "backwards" means that the invariant of
6646 cache[0] >= cache[2] is maintained even with empty slots, which means that
6647 the code that uses it doesn't need to worry if only 1 entry has actually
6648 been set to non-zero. It also makes the "position beyond the end of the
6649 cache" logic much simpler, as the first slot is always the one to start
6653 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6654 const STRLEN utf8, const STRLEN blen)
6658 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6663 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6664 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6665 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6667 (*mgp)->mg_len = -1;
6671 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6672 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6673 (*mgp)->mg_ptr = (char *) cache;
6677 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6678 /* SvPOKp() because it's possible that sv has string overloading, and
6679 therefore is a reference, hence SvPVX() is actually a pointer.
6680 This cures the (very real) symptoms of RT 69422, but I'm not actually
6681 sure whether we should even be caching the results of UTF-8
6682 operations on overloading, given that nothing stops overloading
6683 returning a different value every time it's called. */
6684 const U8 *start = (const U8 *) SvPVX_const(sv);
6685 const STRLEN realutf8 = utf8_length(start, start + byte);
6687 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6691 /* Cache is held with the later position first, to simplify the code
6692 that deals with unbounded ends. */
6694 ASSERT_UTF8_CACHE(cache);
6695 if (cache[1] == 0) {
6696 /* Cache is totally empty */
6699 } else if (cache[3] == 0) {
6700 if (byte > cache[1]) {
6701 /* New one is larger, so goes first. */
6702 cache[2] = cache[0];
6703 cache[3] = cache[1];
6711 #define THREEWAY_SQUARE(a,b,c,d) \
6712 ((float)((d) - (c))) * ((float)((d) - (c))) \
6713 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6714 + ((float)((b) - (a))) * ((float)((b) - (a)))
6716 /* Cache has 2 slots in use, and we know three potential pairs.
6717 Keep the two that give the lowest RMS distance. Do the
6718 calcualation in bytes simply because we always know the byte
6719 length. squareroot has the same ordering as the positive value,
6720 so don't bother with the actual square root. */
6721 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6722 if (byte > cache[1]) {
6723 /* New position is after the existing pair of pairs. */
6724 const float keep_earlier
6725 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6726 const float keep_later
6727 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6729 if (keep_later < keep_earlier) {
6730 if (keep_later < existing) {
6731 cache[2] = cache[0];
6732 cache[3] = cache[1];
6738 if (keep_earlier < existing) {
6744 else if (byte > cache[3]) {
6745 /* New position is between the existing pair of pairs. */
6746 const float keep_earlier
6747 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6748 const float keep_later
6749 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6751 if (keep_later < keep_earlier) {
6752 if (keep_later < existing) {
6758 if (keep_earlier < existing) {
6765 /* New position is before the existing pair of pairs. */
6766 const float keep_earlier
6767 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6768 const float keep_later
6769 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6771 if (keep_later < keep_earlier) {
6772 if (keep_later < existing) {
6778 if (keep_earlier < existing) {
6779 cache[0] = cache[2];
6780 cache[1] = cache[3];
6787 ASSERT_UTF8_CACHE(cache);
6790 /* We already know all of the way, now we may be able to walk back. The same
6791 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6792 backward is half the speed of walking forward. */
6794 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6795 const U8 *end, STRLEN endu)
6797 const STRLEN forw = target - s;
6798 STRLEN backw = end - target;
6800 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6802 if (forw < 2 * backw) {
6803 return utf8_length(s, target);
6806 while (end > target) {
6808 while (UTF8_IS_CONTINUATION(*end)) {
6817 =for apidoc sv_pos_b2u
6819 Converts the value pointed to by offsetp from a count of bytes from the
6820 start of the string, to a count of the equivalent number of UTF-8 chars.
6821 Handles magic and type coercion.
6827 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6828 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6833 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6836 const STRLEN byte = *offsetp;
6837 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6843 PERL_ARGS_ASSERT_SV_POS_B2U;
6848 s = (const U8*)SvPV_const(sv, blen);
6851 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
6857 && SvTYPE(sv) >= SVt_PVMG
6858 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
6861 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
6862 if (cache[1] == byte) {
6863 /* An exact match. */
6864 *offsetp = cache[0];
6867 if (cache[3] == byte) {
6868 /* An exact match. */
6869 *offsetp = cache[2];
6873 if (cache[1] < byte) {
6874 /* We already know part of the way. */
6875 if (mg->mg_len != -1) {
6876 /* Actually, we know the end too. */
6878 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
6879 s + blen, mg->mg_len - cache[0]);
6881 len = cache[0] + utf8_length(s + cache[1], send);
6884 else if (cache[3] < byte) {
6885 /* We're between the two cached pairs, so we do the calculation
6886 offset by the byte/utf-8 positions for the earlier pair,
6887 then add the utf-8 characters from the string start to
6889 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
6890 s + cache[1], cache[0] - cache[2])
6894 else { /* cache[3] > byte */
6895 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
6899 ASSERT_UTF8_CACHE(cache);
6901 } else if (mg->mg_len != -1) {
6902 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
6906 if (!found || PL_utf8cache < 0) {
6907 const STRLEN real_len = utf8_length(s, send);
6909 if (found && PL_utf8cache < 0)
6910 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
6917 utf8_mg_len_cache_update(sv, &mg, len);
6919 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
6924 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
6925 STRLEN real, SV *const sv)
6927 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
6929 /* As this is debugging only code, save space by keeping this test here,
6930 rather than inlining it in all the callers. */
6931 if (from_cache == real)
6934 /* Need to turn the assertions off otherwise we may recurse infinitely
6935 while printing error messages. */
6936 SAVEI8(PL_utf8cache);
6938 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
6939 func, (UV) from_cache, (UV) real, SVfARG(sv));
6945 Returns a boolean indicating whether the strings in the two SVs are
6946 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6947 coerce its args to strings if necessary.
6949 =for apidoc sv_eq_flags
6951 Returns a boolean indicating whether the strings in the two SVs are
6952 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
6953 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
6959 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const I32 flags)
6968 SV* svrecode = NULL;
6975 /* if pv1 and pv2 are the same, second SvPV_const call may
6976 * invalidate pv1 (if we are handling magic), so we may need to
6978 if (sv1 == sv2 && flags & SV_GMAGIC
6979 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
6980 pv1 = SvPV_const(sv1, cur1);
6981 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
6983 pv1 = SvPV_flags_const(sv1, cur1, flags);
6991 pv2 = SvPV_flags_const(sv2, cur2, flags);
6993 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6994 /* Differing utf8ness.
6995 * Do not UTF8size the comparands as a side-effect. */
6998 svrecode = newSVpvn(pv2, cur2);
6999 sv_recode_to_utf8(svrecode, PL_encoding);
7000 pv2 = SvPV_const(svrecode, cur2);
7003 svrecode = newSVpvn(pv1, cur1);
7004 sv_recode_to_utf8(svrecode, PL_encoding);
7005 pv1 = SvPV_const(svrecode, cur1);
7007 /* Now both are in UTF-8. */
7009 SvREFCNT_dec(svrecode);
7014 bool is_utf8 = TRUE;
7017 /* sv1 is the UTF-8 one,
7018 * if is equal it must be downgrade-able */
7019 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
7025 /* sv2 is the UTF-8 one,
7026 * if is equal it must be downgrade-able */
7027 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
7033 /* Downgrade not possible - cannot be eq */
7041 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7043 SvREFCNT_dec(svrecode);
7053 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7054 string in C<sv1> is less than, equal to, or greater than the string in
7055 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7056 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7058 =for apidoc sv_cmp_flags
7060 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7061 string in C<sv1> is less than, equal to, or greater than the string in
7062 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7063 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7064 also C<sv_cmp_locale_flags>.
7070 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7072 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7076 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2, const I32 flags)
7080 const char *pv1, *pv2;
7083 SV *svrecode = NULL;
7090 pv1 = SvPV_flags_const(sv1, cur1, flags);
7097 pv2 = SvPV_flags_const(sv2, cur2, flags);
7099 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7100 /* Differing utf8ness.
7101 * Do not UTF8size the comparands as a side-effect. */
7104 svrecode = newSVpvn(pv2, cur2);
7105 sv_recode_to_utf8(svrecode, PL_encoding);
7106 pv2 = SvPV_const(svrecode, cur2);
7109 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
7114 svrecode = newSVpvn(pv1, cur1);
7115 sv_recode_to_utf8(svrecode, PL_encoding);
7116 pv1 = SvPV_const(svrecode, cur1);
7119 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
7125 cmp = cur2 ? -1 : 0;
7129 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7132 cmp = retval < 0 ? -1 : 1;
7133 } else if (cur1 == cur2) {
7136 cmp = cur1 < cur2 ? -1 : 1;
7140 SvREFCNT_dec(svrecode);
7148 =for apidoc sv_cmp_locale
7150 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7151 'use bytes' aware, handles get magic, and will coerce its args to strings
7152 if necessary. See also C<sv_cmp>.
7154 =for apidoc sv_cmp_locale_flags
7156 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7157 'use bytes' aware and will coerce its args to strings if necessary. If the
7158 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7164 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7166 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7170 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2, const I32 flags)
7173 #ifdef USE_LOCALE_COLLATE
7179 if (PL_collation_standard)
7183 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7185 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7187 if (!pv1 || !len1) {
7198 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7201 return retval < 0 ? -1 : 1;
7204 * When the result of collation is equality, that doesn't mean
7205 * that there are no differences -- some locales exclude some
7206 * characters from consideration. So to avoid false equalities,
7207 * we use the raw string as a tiebreaker.
7213 #endif /* USE_LOCALE_COLLATE */
7215 return sv_cmp(sv1, sv2);
7219 #ifdef USE_LOCALE_COLLATE
7222 =for apidoc sv_collxfrm
7224 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7225 C<sv_collxfrm_flags>.
7227 =for apidoc sv_collxfrm_flags
7229 Add Collate Transform magic to an SV if it doesn't already have it. If the
7230 flags contain SV_GMAGIC, it handles get-magic.
7232 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7233 scalar data of the variable, but transformed to such a format that a normal
7234 memory comparison can be used to compare the data according to the locale
7241 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7246 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7248 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7249 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7255 Safefree(mg->mg_ptr);
7256 s = SvPV_flags_const(sv, len, flags);
7257 if ((xf = mem_collxfrm(s, len, &xlen))) {
7259 #ifdef PERL_OLD_COPY_ON_WRITE
7261 sv_force_normal_flags(sv, 0);
7263 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7277 if (mg && mg->mg_ptr) {
7279 return mg->mg_ptr + sizeof(PL_collation_ix);
7287 #endif /* USE_LOCALE_COLLATE */
7292 Get a line from the filehandle and store it into the SV, optionally
7293 appending to the currently-stored string.
7299 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7304 register STDCHAR rslast;
7305 register STDCHAR *bp;
7310 PERL_ARGS_ASSERT_SV_GETS;
7312 if (SvTHINKFIRST(sv))
7313 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7314 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7316 However, perlbench says it's slower, because the existing swipe code
7317 is faster than copy on write.
7318 Swings and roundabouts. */
7319 SvUPGRADE(sv, SVt_PV);
7324 if (PerlIO_isutf8(fp)) {
7326 sv_utf8_upgrade_nomg(sv);
7327 sv_pos_u2b(sv,&append,0);
7329 } else if (SvUTF8(sv)) {
7330 SV * const tsv = newSV(0);
7331 sv_gets(tsv, fp, 0);
7332 sv_utf8_upgrade_nomg(tsv);
7333 SvCUR_set(sv,append);
7336 goto return_string_or_null;
7344 if (PerlIO_isutf8(fp))
7347 if (IN_PERL_COMPILETIME) {
7348 /* we always read code in line mode */
7352 else if (RsSNARF(PL_rs)) {
7353 /* If it is a regular disk file use size from stat() as estimate
7354 of amount we are going to read -- may result in mallocing
7355 more memory than we really need if the layers below reduce
7356 the size we read (e.g. CRLF or a gzip layer).
7359 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7360 const Off_t offset = PerlIO_tell(fp);
7361 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7362 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7368 else if (RsRECORD(PL_rs)) {
7376 /* Grab the size of the record we're getting */
7377 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7378 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7381 /* VMS wants read instead of fread, because fread doesn't respect */
7382 /* RMS record boundaries. This is not necessarily a good thing to be */
7383 /* doing, but we've got no other real choice - except avoid stdio
7384 as implementation - perhaps write a :vms layer ?
7386 fd = PerlIO_fileno(fp);
7387 if (fd == -1) { /* in-memory file from PerlIO::Scalar */
7388 bytesread = PerlIO_read(fp, buffer, recsize);
7391 bytesread = PerlLIO_read(fd, buffer, recsize);
7394 bytesread = PerlIO_read(fp, buffer, recsize);
7398 SvCUR_set(sv, bytesread + append);
7399 buffer[bytesread] = '\0';
7400 goto return_string_or_null;
7402 else if (RsPARA(PL_rs)) {
7408 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7409 if (PerlIO_isutf8(fp)) {
7410 rsptr = SvPVutf8(PL_rs, rslen);
7413 if (SvUTF8(PL_rs)) {
7414 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7415 Perl_croak(aTHX_ "Wide character in $/");
7418 rsptr = SvPV_const(PL_rs, rslen);
7422 rslast = rslen ? rsptr[rslen - 1] : '\0';
7424 if (rspara) { /* have to do this both before and after */
7425 do { /* to make sure file boundaries work right */
7428 i = PerlIO_getc(fp);
7432 PerlIO_ungetc(fp,i);
7438 /* See if we know enough about I/O mechanism to cheat it ! */
7440 /* This used to be #ifdef test - it is made run-time test for ease
7441 of abstracting out stdio interface. One call should be cheap
7442 enough here - and may even be a macro allowing compile
7446 if (PerlIO_fast_gets(fp)) {
7449 * We're going to steal some values from the stdio struct
7450 * and put EVERYTHING in the innermost loop into registers.
7452 register STDCHAR *ptr;
7456 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7457 /* An ungetc()d char is handled separately from the regular
7458 * buffer, so we getc() it back out and stuff it in the buffer.
7460 i = PerlIO_getc(fp);
7461 if (i == EOF) return 0;
7462 *(--((*fp)->_ptr)) = (unsigned char) i;
7466 /* Here is some breathtakingly efficient cheating */
7468 cnt = PerlIO_get_cnt(fp); /* get count into register */
7469 /* make sure we have the room */
7470 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7471 /* Not room for all of it
7472 if we are looking for a separator and room for some
7474 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7475 /* just process what we have room for */
7476 shortbuffered = cnt - SvLEN(sv) + append + 1;
7477 cnt -= shortbuffered;
7481 /* remember that cnt can be negative */
7482 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7487 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7488 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7489 DEBUG_P(PerlIO_printf(Perl_debug_log,
7490 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7491 DEBUG_P(PerlIO_printf(Perl_debug_log,
7492 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7493 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7494 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7499 while (cnt > 0) { /* this | eat */
7501 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7502 goto thats_all_folks; /* screams | sed :-) */
7506 Copy(ptr, bp, cnt, char); /* this | eat */
7507 bp += cnt; /* screams | dust */
7508 ptr += cnt; /* louder | sed :-) */
7513 if (shortbuffered) { /* oh well, must extend */
7514 cnt = shortbuffered;
7516 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7518 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7519 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7523 DEBUG_P(PerlIO_printf(Perl_debug_log,
7524 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7525 PTR2UV(ptr),(long)cnt));
7526 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7528 DEBUG_P(PerlIO_printf(Perl_debug_log,
7529 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7530 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7531 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7533 /* This used to call 'filbuf' in stdio form, but as that behaves like
7534 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7535 another abstraction. */
7536 i = PerlIO_getc(fp); /* get more characters */
7538 DEBUG_P(PerlIO_printf(Perl_debug_log,
7539 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7540 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7541 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7543 cnt = PerlIO_get_cnt(fp);
7544 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7545 DEBUG_P(PerlIO_printf(Perl_debug_log,
7546 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7548 if (i == EOF) /* all done for ever? */
7549 goto thats_really_all_folks;
7551 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7553 SvGROW(sv, bpx + cnt + 2);
7554 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7556 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7558 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7559 goto thats_all_folks;
7563 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7564 memNE((char*)bp - rslen, rsptr, rslen))
7565 goto screamer; /* go back to the fray */
7566 thats_really_all_folks:
7568 cnt += shortbuffered;
7569 DEBUG_P(PerlIO_printf(Perl_debug_log,
7570 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7571 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7572 DEBUG_P(PerlIO_printf(Perl_debug_log,
7573 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7574 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7575 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7577 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7578 DEBUG_P(PerlIO_printf(Perl_debug_log,
7579 "Screamer: done, len=%ld, string=|%.*s|\n",
7580 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7584 /*The big, slow, and stupid way. */
7585 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7586 STDCHAR *buf = NULL;
7587 Newx(buf, 8192, STDCHAR);
7595 register const STDCHAR * const bpe = buf + sizeof(buf);
7597 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7598 ; /* keep reading */
7602 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7603 /* Accomodate broken VAXC compiler, which applies U8 cast to
7604 * both args of ?: operator, causing EOF to change into 255
7607 i = (U8)buf[cnt - 1];
7613 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7615 sv_catpvn(sv, (char *) buf, cnt);
7617 sv_setpvn(sv, (char *) buf, cnt);
7619 if (i != EOF && /* joy */
7621 SvCUR(sv) < rslen ||
7622 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7626 * If we're reading from a TTY and we get a short read,
7627 * indicating that the user hit his EOF character, we need
7628 * to notice it now, because if we try to read from the TTY
7629 * again, the EOF condition will disappear.
7631 * The comparison of cnt to sizeof(buf) is an optimization
7632 * that prevents unnecessary calls to feof().
7636 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7640 #ifdef USE_HEAP_INSTEAD_OF_STACK
7645 if (rspara) { /* have to do this both before and after */
7646 while (i != EOF) { /* to make sure file boundaries work right */
7647 i = PerlIO_getc(fp);
7649 PerlIO_ungetc(fp,i);
7655 return_string_or_null:
7656 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7662 Auto-increment of the value in the SV, doing string to numeric conversion
7663 if necessary. Handles 'get' magic and operator overloading.
7669 Perl_sv_inc(pTHX_ register SV *const sv)
7678 =for apidoc sv_inc_nomg
7680 Auto-increment of the value in the SV, doing string to numeric conversion
7681 if necessary. Handles operator overloading. Skips handling 'get' magic.
7687 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7695 if (SvTHINKFIRST(sv)) {
7697 sv_force_normal_flags(sv, 0);
7698 if (SvREADONLY(sv)) {
7699 if (IN_PERL_RUNTIME)
7700 Perl_croak_no_modify(aTHX);
7704 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
7706 i = PTR2IV(SvRV(sv));
7711 flags = SvFLAGS(sv);
7712 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7713 /* It's (privately or publicly) a float, but not tested as an
7714 integer, so test it to see. */
7716 flags = SvFLAGS(sv);
7718 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7719 /* It's publicly an integer, or privately an integer-not-float */
7720 #ifdef PERL_PRESERVE_IVUV
7724 if (SvUVX(sv) == UV_MAX)
7725 sv_setnv(sv, UV_MAX_P1);
7727 (void)SvIOK_only_UV(sv);
7728 SvUV_set(sv, SvUVX(sv) + 1);
7730 if (SvIVX(sv) == IV_MAX)
7731 sv_setuv(sv, (UV)IV_MAX + 1);
7733 (void)SvIOK_only(sv);
7734 SvIV_set(sv, SvIVX(sv) + 1);
7739 if (flags & SVp_NOK) {
7740 const NV was = SvNVX(sv);
7741 if (NV_OVERFLOWS_INTEGERS_AT &&
7742 was >= NV_OVERFLOWS_INTEGERS_AT) {
7743 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7744 "Lost precision when incrementing %" NVff " by 1",
7747 (void)SvNOK_only(sv);
7748 SvNV_set(sv, was + 1.0);
7752 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7753 if ((flags & SVTYPEMASK) < SVt_PVIV)
7754 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7755 (void)SvIOK_only(sv);
7760 while (isALPHA(*d)) d++;
7761 while (isDIGIT(*d)) d++;
7762 if (d < SvEND(sv)) {
7763 #ifdef PERL_PRESERVE_IVUV
7764 /* Got to punt this as an integer if needs be, but we don't issue
7765 warnings. Probably ought to make the sv_iv_please() that does
7766 the conversion if possible, and silently. */
7767 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7768 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7769 /* Need to try really hard to see if it's an integer.
7770 9.22337203685478e+18 is an integer.
7771 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7772 so $a="9.22337203685478e+18"; $a+0; $a++
7773 needs to be the same as $a="9.22337203685478e+18"; $a++
7780 /* sv_2iv *should* have made this an NV */
7781 if (flags & SVp_NOK) {
7782 (void)SvNOK_only(sv);
7783 SvNV_set(sv, SvNVX(sv) + 1.0);
7786 /* I don't think we can get here. Maybe I should assert this
7787 And if we do get here I suspect that sv_setnv will croak. NWC
7789 #if defined(USE_LONG_DOUBLE)
7790 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",
7791 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7793 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7794 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7797 #endif /* PERL_PRESERVE_IVUV */
7798 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7802 while (d >= SvPVX_const(sv)) {
7810 /* MKS: The original code here died if letters weren't consecutive.
7811 * at least it didn't have to worry about non-C locales. The
7812 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7813 * arranged in order (although not consecutively) and that only
7814 * [A-Za-z] are accepted by isALPHA in the C locale.
7816 if (*d != 'z' && *d != 'Z') {
7817 do { ++*d; } while (!isALPHA(*d));
7820 *(d--) -= 'z' - 'a';
7825 *(d--) -= 'z' - 'a' + 1;
7829 /* oh,oh, the number grew */
7830 SvGROW(sv, SvCUR(sv) + 2);
7831 SvCUR_set(sv, SvCUR(sv) + 1);
7832 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7843 Auto-decrement of the value in the SV, doing string to numeric conversion
7844 if necessary. Handles 'get' magic and operator overloading.
7850 Perl_sv_dec(pTHX_ register SV *const sv)
7860 =for apidoc sv_dec_nomg
7862 Auto-decrement of the value in the SV, doing string to numeric conversion
7863 if necessary. Handles operator overloading. Skips handling 'get' magic.
7869 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
7876 if (SvTHINKFIRST(sv)) {
7878 sv_force_normal_flags(sv, 0);
7879 if (SvREADONLY(sv)) {
7880 if (IN_PERL_RUNTIME)
7881 Perl_croak_no_modify(aTHX);
7885 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
7887 i = PTR2IV(SvRV(sv));
7892 /* Unlike sv_inc we don't have to worry about string-never-numbers
7893 and keeping them magic. But we mustn't warn on punting */
7894 flags = SvFLAGS(sv);
7895 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7896 /* It's publicly an integer, or privately an integer-not-float */
7897 #ifdef PERL_PRESERVE_IVUV
7901 if (SvUVX(sv) == 0) {
7902 (void)SvIOK_only(sv);
7906 (void)SvIOK_only_UV(sv);
7907 SvUV_set(sv, SvUVX(sv) - 1);
7910 if (SvIVX(sv) == IV_MIN) {
7911 sv_setnv(sv, (NV)IV_MIN);
7915 (void)SvIOK_only(sv);
7916 SvIV_set(sv, SvIVX(sv) - 1);
7921 if (flags & SVp_NOK) {
7924 const NV was = SvNVX(sv);
7925 if (NV_OVERFLOWS_INTEGERS_AT &&
7926 was <= -NV_OVERFLOWS_INTEGERS_AT) {
7927 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7928 "Lost precision when decrementing %" NVff " by 1",
7931 (void)SvNOK_only(sv);
7932 SvNV_set(sv, was - 1.0);
7936 if (!(flags & SVp_POK)) {
7937 if ((flags & SVTYPEMASK) < SVt_PVIV)
7938 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
7940 (void)SvIOK_only(sv);
7943 #ifdef PERL_PRESERVE_IVUV
7945 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7946 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7947 /* Need to try really hard to see if it's an integer.
7948 9.22337203685478e+18 is an integer.
7949 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7950 so $a="9.22337203685478e+18"; $a+0; $a--
7951 needs to be the same as $a="9.22337203685478e+18"; $a--
7958 /* sv_2iv *should* have made this an NV */
7959 if (flags & SVp_NOK) {
7960 (void)SvNOK_only(sv);
7961 SvNV_set(sv, SvNVX(sv) - 1.0);
7964 /* I don't think we can get here. Maybe I should assert this
7965 And if we do get here I suspect that sv_setnv will croak. NWC
7967 #if defined(USE_LONG_DOUBLE)
7968 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",
7969 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7971 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7972 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7976 #endif /* PERL_PRESERVE_IVUV */
7977 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
7980 /* this define is used to eliminate a chunk of duplicated but shared logic
7981 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
7982 * used anywhere but here - yves
7984 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
7987 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
7991 =for apidoc sv_mortalcopy
7993 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
7994 The new SV is marked as mortal. It will be destroyed "soon", either by an
7995 explicit call to FREETMPS, or by an implicit call at places such as
7996 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8001 /* Make a string that will exist for the duration of the expression
8002 * evaluation. Actually, it may have to last longer than that, but
8003 * hopefully we won't free it until it has been assigned to a
8004 * permanent location. */
8007 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8013 sv_setsv(sv,oldstr);
8014 PUSH_EXTEND_MORTAL__SV_C(sv);
8020 =for apidoc sv_newmortal
8022 Creates a new null SV which is mortal. The reference count of the SV is
8023 set to 1. It will be destroyed "soon", either by an explicit call to
8024 FREETMPS, or by an implicit call at places such as statement boundaries.
8025 See also C<sv_mortalcopy> and C<sv_2mortal>.
8031 Perl_sv_newmortal(pTHX)
8037 SvFLAGS(sv) = SVs_TEMP;
8038 PUSH_EXTEND_MORTAL__SV_C(sv);
8044 =for apidoc newSVpvn_flags
8046 Creates a new SV and copies a string into it. The reference count for the
8047 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8048 string. You are responsible for ensuring that the source string is at least
8049 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8050 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8051 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8052 returning. If C<SVf_UTF8> is set, C<s> is considered to be in UTF-8 and the
8053 C<SVf_UTF8> flag will be set on the new SV.
8054 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8056 #define newSVpvn_utf8(s, len, u) \
8057 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8063 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8068 /* All the flags we don't support must be zero.
8069 And we're new code so I'm going to assert this from the start. */
8070 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8072 sv_setpvn(sv,s,len);
8074 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8075 * and do what it does outselves here.
8076 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8077 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8078 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8079 * eleminate quite a few steps than it looks - Yves (explaining patch by gfx)
8082 SvFLAGS(sv) |= flags;
8084 if(flags & SVs_TEMP){
8085 PUSH_EXTEND_MORTAL__SV_C(sv);
8092 =for apidoc sv_2mortal
8094 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8095 by an explicit call to FREETMPS, or by an implicit call at places such as
8096 statement boundaries. SvTEMP() is turned on which means that the SV's
8097 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8098 and C<sv_mortalcopy>.
8104 Perl_sv_2mortal(pTHX_ register SV *const sv)
8109 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8111 PUSH_EXTEND_MORTAL__SV_C(sv);
8119 Creates a new SV and copies a string into it. The reference count for the
8120 SV is set to 1. If C<len> is zero, Perl will compute the length using
8121 strlen(). For efficiency, consider using C<newSVpvn> instead.
8127 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8133 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8138 =for apidoc newSVpvn
8140 Creates a new SV and copies a string into it. The reference count for the
8141 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8142 string. You are responsible for ensuring that the source string is at least
8143 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8149 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
8155 sv_setpvn(sv,s,len);
8160 =for apidoc newSVhek
8162 Creates a new SV from the hash key structure. It will generate scalars that
8163 point to the shared string table where possible. Returns a new (undefined)
8164 SV if the hek is NULL.
8170 Perl_newSVhek(pTHX_ const HEK *const hek)
8180 if (HEK_LEN(hek) == HEf_SVKEY) {
8181 return newSVsv(*(SV**)HEK_KEY(hek));
8183 const int flags = HEK_FLAGS(hek);
8184 if (flags & HVhek_WASUTF8) {
8186 Andreas would like keys he put in as utf8 to come back as utf8
8188 STRLEN utf8_len = HEK_LEN(hek);
8189 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8190 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
8193 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
8195 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8196 /* We don't have a pointer to the hv, so we have to replicate the
8197 flag into every HEK. This hv is using custom a hasing
8198 algorithm. Hence we can't return a shared string scalar, as
8199 that would contain the (wrong) hash value, and might get passed
8200 into an hv routine with a regular hash.
8201 Similarly, a hash that isn't using shared hash keys has to have
8202 the flag in every key so that we know not to try to call
8203 share_hek_kek on it. */
8205 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8210 /* This will be overwhelminly the most common case. */
8212 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8213 more efficient than sharepvn(). */
8217 sv_upgrade(sv, SVt_PV);
8218 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8219 SvCUR_set(sv, HEK_LEN(hek));
8232 =for apidoc newSVpvn_share
8234 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8235 table. If the string does not already exist in the table, it is created
8236 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
8237 value is used; otherwise the hash is computed. The string's hash can be later
8238 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
8239 that as the string table is used for shared hash keys these strings will have
8240 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8246 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8250 bool is_utf8 = FALSE;
8251 const char *const orig_src = src;
8254 STRLEN tmplen = -len;
8256 /* See the note in hv.c:hv_fetch() --jhi */
8257 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8261 PERL_HASH(hash, src, len);
8263 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8264 changes here, update it there too. */
8265 sv_upgrade(sv, SVt_PV);
8266 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8274 if (src != orig_src)
8280 =for apidoc newSVpv_share
8282 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8289 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8291 return newSVpvn_share(src, strlen(src), hash);
8294 #if defined(PERL_IMPLICIT_CONTEXT)
8296 /* pTHX_ magic can't cope with varargs, so this is a no-context
8297 * version of the main function, (which may itself be aliased to us).
8298 * Don't access this version directly.
8302 Perl_newSVpvf_nocontext(const char *const pat, ...)
8308 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8310 va_start(args, pat);
8311 sv = vnewSVpvf(pat, &args);
8318 =for apidoc newSVpvf
8320 Creates a new SV and initializes it with the string formatted like
8327 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8332 PERL_ARGS_ASSERT_NEWSVPVF;
8334 va_start(args, pat);
8335 sv = vnewSVpvf(pat, &args);
8340 /* backend for newSVpvf() and newSVpvf_nocontext() */
8343 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8348 PERL_ARGS_ASSERT_VNEWSVPVF;
8351 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8358 Creates a new SV and copies a floating point value into it.
8359 The reference count for the SV is set to 1.
8365 Perl_newSVnv(pTHX_ const NV n)
8378 Creates a new SV and copies an integer into it. The reference count for the
8385 Perl_newSViv(pTHX_ const IV i)
8398 Creates a new SV and copies an unsigned integer into it.
8399 The reference count for the SV is set to 1.
8405 Perl_newSVuv(pTHX_ const UV u)
8416 =for apidoc newSV_type
8418 Creates a new SV, of the type specified. The reference count for the new SV
8425 Perl_newSV_type(pTHX_ const svtype type)
8430 sv_upgrade(sv, type);
8435 =for apidoc newRV_noinc
8437 Creates an RV wrapper for an SV. The reference count for the original
8438 SV is B<not> incremented.
8444 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8447 register SV *sv = newSV_type(SVt_IV);
8449 PERL_ARGS_ASSERT_NEWRV_NOINC;
8452 SvRV_set(sv, tmpRef);
8457 /* newRV_inc is the official function name to use now.
8458 * newRV_inc is in fact #defined to newRV in sv.h
8462 Perl_newRV(pTHX_ SV *const sv)
8466 PERL_ARGS_ASSERT_NEWRV;
8468 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8474 Creates a new SV which is an exact duplicate of the original SV.
8481 Perl_newSVsv(pTHX_ register SV *const old)
8488 if (SvTYPE(old) == SVTYPEMASK) {
8489 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8493 /* SV_GMAGIC is the default for sv_setv()
8494 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8495 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8496 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8501 =for apidoc sv_reset
8503 Underlying implementation for the C<reset> Perl function.
8504 Note that the perl-level function is vaguely deprecated.
8510 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8513 char todo[PERL_UCHAR_MAX+1];
8515 PERL_ARGS_ASSERT_SV_RESET;
8520 if (!*s) { /* reset ?? searches */
8521 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8523 const U32 count = mg->mg_len / sizeof(PMOP**);
8524 PMOP **pmp = (PMOP**) mg->mg_ptr;
8525 PMOP *const *const end = pmp + count;
8529 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8531 (*pmp)->op_pmflags &= ~PMf_USED;
8539 /* reset variables */
8541 if (!HvARRAY(stash))
8544 Zero(todo, 256, char);
8547 I32 i = (unsigned char)*s;
8551 max = (unsigned char)*s++;
8552 for ( ; i <= max; i++) {
8555 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8557 for (entry = HvARRAY(stash)[i];
8559 entry = HeNEXT(entry))
8564 if (!todo[(U8)*HeKEY(entry)])
8566 gv = MUTABLE_GV(HeVAL(entry));
8569 if (SvTHINKFIRST(sv)) {
8570 if (!SvREADONLY(sv) && SvROK(sv))
8572 /* XXX Is this continue a bug? Why should THINKFIRST
8573 exempt us from resetting arrays and hashes? */
8577 if (SvTYPE(sv) >= SVt_PV) {
8579 if (SvPVX_const(sv) != NULL)
8587 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8589 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8592 # if defined(USE_ENVIRON_ARRAY)
8595 # endif /* USE_ENVIRON_ARRAY */
8606 Using various gambits, try to get an IO from an SV: the IO slot if its a
8607 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8608 named after the PV if we're a string.
8614 Perl_sv_2io(pTHX_ SV *const sv)
8619 PERL_ARGS_ASSERT_SV_2IO;
8621 switch (SvTYPE(sv)) {
8623 io = MUTABLE_IO(sv);
8627 if (isGV_with_GP(sv)) {
8628 gv = MUTABLE_GV(sv);
8631 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8637 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8639 return sv_2io(SvRV(sv));
8640 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8646 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8655 Using various gambits, try to get a CV from an SV; in addition, try if
8656 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8657 The flags in C<lref> are passed to gv_fetchsv.
8663 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8669 PERL_ARGS_ASSERT_SV_2CV;
8676 switch (SvTYPE(sv)) {
8680 return MUTABLE_CV(sv);
8687 if (isGV_with_GP(sv)) {
8688 gv = MUTABLE_GV(sv);
8697 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
8699 tryAMAGICunDEREF(to_cv);
8702 if (SvTYPE(sv) == SVt_PVCV) {
8703 cv = MUTABLE_CV(sv);
8708 else if(isGV_with_GP(sv))
8709 gv = MUTABLE_GV(sv);
8711 Perl_croak(aTHX_ "Not a subroutine reference");
8713 else if (isGV_with_GP(sv)) {
8715 gv = MUTABLE_GV(sv);
8718 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8724 /* Some flags to gv_fetchsv mean don't really create the GV */
8725 if (!isGV_with_GP(gv)) {
8731 if (lref && !GvCVu(gv)) {
8735 gv_efullname3(tmpsv, gv, NULL);
8736 /* XXX this is probably not what they think they're getting.
8737 * It has the same effect as "sub name;", i.e. just a forward
8739 newSUB(start_subparse(FALSE, 0),
8740 newSVOP(OP_CONST, 0, tmpsv),
8744 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8745 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8754 Returns true if the SV has a true value by Perl's rules.
8755 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8756 instead use an in-line version.
8762 Perl_sv_true(pTHX_ register SV *const sv)
8767 register const XPV* const tXpv = (XPV*)SvANY(sv);
8769 (tXpv->xpv_cur > 1 ||
8770 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8777 return SvIVX(sv) != 0;
8780 return SvNVX(sv) != 0.0;
8782 return sv_2bool(sv);
8788 =for apidoc sv_pvn_force
8790 Get a sensible string out of the SV somehow.
8791 A private implementation of the C<SvPV_force> macro for compilers which
8792 can't cope with complex macro expressions. Always use the macro instead.
8794 =for apidoc sv_pvn_force_flags
8796 Get a sensible string out of the SV somehow.
8797 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8798 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8799 implemented in terms of this function.
8800 You normally want to use the various wrapper macros instead: see
8801 C<SvPV_force> and C<SvPV_force_nomg>
8807 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8811 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8813 if (SvTHINKFIRST(sv) && !SvROK(sv))
8814 sv_force_normal_flags(sv, 0);
8824 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8825 const char * const ref = sv_reftype(sv,0);
8827 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8828 ref, OP_DESC(PL_op));
8830 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8832 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8833 || isGV_with_GP(sv))
8834 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8836 s = sv_2pv_flags(sv, &len, flags);
8840 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8843 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8844 SvGROW(sv, len + 1);
8845 Move(s,SvPVX(sv),len,char);
8847 SvPVX(sv)[len] = '\0';
8850 SvPOK_on(sv); /* validate pointer */
8852 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
8853 PTR2UV(sv),SvPVX_const(sv)));
8856 return SvPVX_mutable(sv);
8860 =for apidoc sv_pvbyten_force
8862 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
8868 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
8870 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
8872 sv_pvn_force(sv,lp);
8873 sv_utf8_downgrade(sv,0);
8879 =for apidoc sv_pvutf8n_force
8881 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
8887 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
8889 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
8891 sv_pvn_force(sv,lp);
8892 sv_utf8_upgrade(sv);
8898 =for apidoc sv_reftype
8900 Returns a string describing what the SV is a reference to.
8906 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
8908 PERL_ARGS_ASSERT_SV_REFTYPE;
8910 /* The fact that I don't need to downcast to char * everywhere, only in ?:
8911 inside return suggests a const propagation bug in g++. */
8912 if (ob && SvOBJECT(sv)) {
8913 char * const name = HvNAME_get(SvSTASH(sv));
8914 return name ? name : (char *) "__ANON__";
8917 switch (SvTYPE(sv)) {
8932 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
8933 /* tied lvalues should appear to be
8934 * scalars for backwards compatitbility */
8935 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
8936 ? "SCALAR" : "LVALUE");
8937 case SVt_PVAV: return "ARRAY";
8938 case SVt_PVHV: return "HASH";
8939 case SVt_PVCV: return "CODE";
8940 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
8941 ? "GLOB" : "SCALAR");
8942 case SVt_PVFM: return "FORMAT";
8943 case SVt_PVIO: return "IO";
8944 case SVt_BIND: return "BIND";
8945 case SVt_REGEXP: return "REGEXP";
8946 default: return "UNKNOWN";
8952 =for apidoc sv_isobject
8954 Returns a boolean indicating whether the SV is an RV pointing to a blessed
8955 object. If the SV is not an RV, or if the object is not blessed, then this
8962 Perl_sv_isobject(pTHX_ SV *sv)
8978 Returns a boolean indicating whether the SV is blessed into the specified
8979 class. This does not check for subtypes; use C<sv_derived_from> to verify
8980 an inheritance relationship.
8986 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
8990 PERL_ARGS_ASSERT_SV_ISA;
9000 hvname = HvNAME_get(SvSTASH(sv));
9004 return strEQ(hvname, name);
9010 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9011 it will be upgraded to one. If C<classname> is non-null then the new SV will
9012 be blessed in the specified package. The new SV is returned and its
9013 reference count is 1.
9019 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9024 PERL_ARGS_ASSERT_NEWSVRV;
9028 SV_CHECK_THINKFIRST_COW_DROP(rv);
9029 (void)SvAMAGIC_off(rv);
9031 if (SvTYPE(rv) >= SVt_PVMG) {
9032 const U32 refcnt = SvREFCNT(rv);
9036 SvREFCNT(rv) = refcnt;
9038 sv_upgrade(rv, SVt_IV);
9039 } else if (SvROK(rv)) {
9040 SvREFCNT_dec(SvRV(rv));
9042 prepare_SV_for_RV(rv);
9050 HV* const stash = gv_stashpv(classname, GV_ADD);
9051 (void)sv_bless(rv, stash);
9057 =for apidoc sv_setref_pv
9059 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9060 argument will be upgraded to an RV. That RV will be modified to point to
9061 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9062 into the SV. The C<classname> argument indicates the package for the
9063 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9064 will have a reference count of 1, and the RV will be returned.
9066 Do not use with other Perl types such as HV, AV, SV, CV, because those
9067 objects will become corrupted by the pointer copy process.
9069 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9075 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9079 PERL_ARGS_ASSERT_SV_SETREF_PV;
9082 sv_setsv(rv, &PL_sv_undef);
9086 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9091 =for apidoc sv_setref_iv
9093 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9094 argument will be upgraded to an RV. That RV will be modified to point to
9095 the new SV. The C<classname> argument indicates the package for the
9096 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9097 will have a reference count of 1, and the RV will be returned.
9103 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9105 PERL_ARGS_ASSERT_SV_SETREF_IV;
9107 sv_setiv(newSVrv(rv,classname), iv);
9112 =for apidoc sv_setref_uv
9114 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9115 argument will be upgraded to an RV. That RV will be modified to point to
9116 the new SV. The C<classname> argument indicates the package for the
9117 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9118 will have a reference count of 1, and the RV will be returned.
9124 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9126 PERL_ARGS_ASSERT_SV_SETREF_UV;
9128 sv_setuv(newSVrv(rv,classname), uv);
9133 =for apidoc sv_setref_nv
9135 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9136 argument will be upgraded to an RV. That RV will be modified to point to
9137 the new SV. The C<classname> argument indicates the package for the
9138 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9139 will have a reference count of 1, and the RV will be returned.
9145 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9147 PERL_ARGS_ASSERT_SV_SETREF_NV;
9149 sv_setnv(newSVrv(rv,classname), nv);
9154 =for apidoc sv_setref_pvn
9156 Copies a string into a new SV, optionally blessing the SV. The length of the
9157 string must be specified with C<n>. The C<rv> argument will be upgraded to
9158 an RV. That RV will be modified to point to the new SV. The C<classname>
9159 argument indicates the package for the blessing. Set C<classname> to
9160 C<NULL> to avoid the blessing. The new SV will have a reference count
9161 of 1, and the RV will be returned.
9163 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9169 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9170 const char *const pv, const STRLEN n)
9172 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9174 sv_setpvn(newSVrv(rv,classname), pv, n);
9179 =for apidoc sv_bless
9181 Blesses an SV into a specified package. The SV must be an RV. The package
9182 must be designated by its stash (see C<gv_stashpv()>). The reference count
9183 of the SV is unaffected.
9189 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9194 PERL_ARGS_ASSERT_SV_BLESS;
9197 Perl_croak(aTHX_ "Can't bless non-reference value");
9199 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9200 if (SvIsCOW(tmpRef))
9201 sv_force_normal_flags(tmpRef, 0);
9202 if (SvREADONLY(tmpRef))
9203 Perl_croak_no_modify(aTHX);
9204 if (SvOBJECT(tmpRef)) {
9205 if (SvTYPE(tmpRef) != SVt_PVIO)
9207 SvREFCNT_dec(SvSTASH(tmpRef));
9210 SvOBJECT_on(tmpRef);
9211 if (SvTYPE(tmpRef) != SVt_PVIO)
9213 SvUPGRADE(tmpRef, SVt_PVMG);
9214 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9219 (void)SvAMAGIC_off(sv);
9221 if(SvSMAGICAL(tmpRef))
9222 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9230 /* Downgrades a PVGV to a PVMG. If it’s actually a PVLV, we leave the type
9231 * as it is after unglobbing it.
9235 S_sv_unglob(pTHX_ SV *const sv)
9240 SV * const temp = sv_newmortal();
9242 PERL_ARGS_ASSERT_SV_UNGLOB;
9244 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9246 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9249 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9250 && HvNAME_get(stash))
9251 mro_method_changed_in(stash);
9252 gp_free(MUTABLE_GV(sv));
9255 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9259 if (GvNAME_HEK(sv)) {
9260 unshare_hek(GvNAME_HEK(sv));
9262 isGV_with_GP_off(sv);
9264 if(SvTYPE(sv) == SVt_PVGV) {
9265 /* need to keep SvANY(sv) in the right arena */
9266 xpvmg = new_XPVMG();
9267 StructCopy(SvANY(sv), xpvmg, XPVMG);
9268 del_XPVGV(SvANY(sv));
9271 SvFLAGS(sv) &= ~SVTYPEMASK;
9272 SvFLAGS(sv) |= SVt_PVMG;
9275 /* Intentionally not calling any local SET magic, as this isn't so much a
9276 set operation as merely an internal storage change. */
9277 sv_setsv_flags(sv, temp, 0);
9281 =for apidoc sv_unref_flags
9283 Unsets the RV status of the SV, and decrements the reference count of
9284 whatever was being referenced by the RV. This can almost be thought of
9285 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9286 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9287 (otherwise the decrementing is conditional on the reference count being
9288 different from one or the reference being a readonly SV).
9295 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9297 SV* const target = SvRV(ref);
9299 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9301 if (SvWEAKREF(ref)) {
9302 sv_del_backref(target, ref);
9304 SvRV_set(ref, NULL);
9307 SvRV_set(ref, NULL);
9309 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9310 assigned to as BEGIN {$a = \"Foo"} will fail. */
9311 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9312 SvREFCNT_dec(target);
9313 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9314 sv_2mortal(target); /* Schedule for freeing later */
9318 =for apidoc sv_untaint
9320 Untaint an SV. Use C<SvTAINTED_off> instead.
9325 Perl_sv_untaint(pTHX_ SV *const sv)
9327 PERL_ARGS_ASSERT_SV_UNTAINT;
9329 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9330 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9337 =for apidoc sv_tainted
9339 Test an SV for taintedness. Use C<SvTAINTED> instead.
9344 Perl_sv_tainted(pTHX_ SV *const sv)
9346 PERL_ARGS_ASSERT_SV_TAINTED;
9348 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9349 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9350 if (mg && (mg->mg_len & 1) )
9357 =for apidoc sv_setpviv
9359 Copies an integer into the given SV, also updating its string value.
9360 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9366 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9368 char buf[TYPE_CHARS(UV)];
9370 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9372 PERL_ARGS_ASSERT_SV_SETPVIV;
9374 sv_setpvn(sv, ptr, ebuf - ptr);
9378 =for apidoc sv_setpviv_mg
9380 Like C<sv_setpviv>, but also handles 'set' magic.
9386 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9388 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9394 #if defined(PERL_IMPLICIT_CONTEXT)
9396 /* pTHX_ magic can't cope with varargs, so this is a no-context
9397 * version of the main function, (which may itself be aliased to us).
9398 * Don't access this version directly.
9402 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9407 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9409 va_start(args, pat);
9410 sv_vsetpvf(sv, pat, &args);
9414 /* pTHX_ magic can't cope with varargs, so this is a no-context
9415 * version of the main function, (which may itself be aliased to us).
9416 * Don't access this version directly.
9420 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9425 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9427 va_start(args, pat);
9428 sv_vsetpvf_mg(sv, pat, &args);
9434 =for apidoc sv_setpvf
9436 Works like C<sv_catpvf> but copies the text into the SV instead of
9437 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9443 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9447 PERL_ARGS_ASSERT_SV_SETPVF;
9449 va_start(args, pat);
9450 sv_vsetpvf(sv, pat, &args);
9455 =for apidoc sv_vsetpvf
9457 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9458 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9460 Usually used via its frontend C<sv_setpvf>.
9466 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9468 PERL_ARGS_ASSERT_SV_VSETPVF;
9470 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9474 =for apidoc sv_setpvf_mg
9476 Like C<sv_setpvf>, but also handles 'set' magic.
9482 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9486 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9488 va_start(args, pat);
9489 sv_vsetpvf_mg(sv, pat, &args);
9494 =for apidoc sv_vsetpvf_mg
9496 Like C<sv_vsetpvf>, but also handles 'set' magic.
9498 Usually used via its frontend C<sv_setpvf_mg>.
9504 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9506 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9508 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9512 #if defined(PERL_IMPLICIT_CONTEXT)
9514 /* pTHX_ magic can't cope with varargs, so this is a no-context
9515 * version of the main function, (which may itself be aliased to us).
9516 * Don't access this version directly.
9520 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9525 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9527 va_start(args, pat);
9528 sv_vcatpvf(sv, pat, &args);
9532 /* pTHX_ magic can't cope with varargs, so this is a no-context
9533 * version of the main function, (which may itself be aliased to us).
9534 * Don't access this version directly.
9538 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9543 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9545 va_start(args, pat);
9546 sv_vcatpvf_mg(sv, pat, &args);
9552 =for apidoc sv_catpvf
9554 Processes its arguments like C<sprintf> and appends the formatted
9555 output to an SV. If the appended data contains "wide" characters
9556 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9557 and characters >255 formatted with %c), the original SV might get
9558 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9559 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9560 valid UTF-8; if the original SV was bytes, the pattern should be too.
9565 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9569 PERL_ARGS_ASSERT_SV_CATPVF;
9571 va_start(args, pat);
9572 sv_vcatpvf(sv, pat, &args);
9577 =for apidoc sv_vcatpvf
9579 Processes its arguments like C<vsprintf> and appends the formatted output
9580 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9582 Usually used via its frontend C<sv_catpvf>.
9588 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9590 PERL_ARGS_ASSERT_SV_VCATPVF;
9592 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9596 =for apidoc sv_catpvf_mg
9598 Like C<sv_catpvf>, but also handles 'set' magic.
9604 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9608 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9610 va_start(args, pat);
9611 sv_vcatpvf_mg(sv, pat, &args);
9616 =for apidoc sv_vcatpvf_mg
9618 Like C<sv_vcatpvf>, but also handles 'set' magic.
9620 Usually used via its frontend C<sv_catpvf_mg>.
9626 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9628 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9630 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9635 =for apidoc sv_vsetpvfn
9637 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9640 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9646 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9647 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9649 PERL_ARGS_ASSERT_SV_VSETPVFN;
9652 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9657 * Warn of missing argument to sprintf, and then return a defined value
9658 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9660 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9662 S_vcatpvfn_missing_argument(pTHX) {
9663 if (ckWARN(WARN_MISSING)) {
9664 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9665 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9672 S_expect_number(pTHX_ char **const pattern)
9677 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9679 switch (**pattern) {
9680 case '1': case '2': case '3':
9681 case '4': case '5': case '6':
9682 case '7': case '8': case '9':
9683 var = *(*pattern)++ - '0';
9684 while (isDIGIT(**pattern)) {
9685 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9687 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9695 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9697 const int neg = nv < 0;
9700 PERL_ARGS_ASSERT_F0CONVERT;
9708 if (uv & 1 && uv == nv)
9709 uv--; /* Round to even */
9711 const unsigned dig = uv % 10;
9724 =for apidoc sv_vcatpvfn
9726 Processes its arguments like C<vsprintf> and appends the formatted output
9727 to an SV. Uses an array of SVs if the C style variable argument list is
9728 missing (NULL). When running with taint checks enabled, indicates via
9729 C<maybe_tainted> if results are untrustworthy (often due to the use of
9732 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9738 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9739 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9740 vec_utf8 = DO_UTF8(vecsv);
9742 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9745 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9746 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9754 static const char nullstr[] = "(null)";
9756 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9757 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9759 /* Times 4: a decimal digit takes more than 3 binary digits.
9760 * NV_DIG: mantissa takes than many decimal digits.
9761 * Plus 32: Playing safe. */
9762 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9763 /* large enough for "%#.#f" --chip */
9764 /* what about long double NVs? --jhi */
9766 PERL_ARGS_ASSERT_SV_VCATPVFN;
9767 PERL_UNUSED_ARG(maybe_tainted);
9769 /* no matter what, this is a string now */
9770 (void)SvPV_force(sv, origlen);
9772 /* special-case "", "%s", and "%-p" (SVf - see below) */
9775 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9777 const char * const s = va_arg(*args, char*);
9778 sv_catpv(sv, s ? s : nullstr);
9780 else if (svix < svmax) {
9781 sv_catsv(sv, *svargs);
9784 S_vcatpvfn_missing_argument(aTHX);
9787 if (args && patlen == 3 && pat[0] == '%' &&
9788 pat[1] == '-' && pat[2] == 'p') {
9789 argsv = MUTABLE_SV(va_arg(*args, void*));
9790 sv_catsv(sv, argsv);
9794 #ifndef USE_LONG_DOUBLE
9795 /* special-case "%.<number>[gf]" */
9796 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9797 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9798 unsigned digits = 0;
9802 while (*pp >= '0' && *pp <= '9')
9803 digits = 10 * digits + (*pp++ - '0');
9804 if (pp - pat == (int)patlen - 1 && svix < svmax) {
9805 const NV nv = SvNV(*svargs);
9807 /* Add check for digits != 0 because it seems that some
9808 gconverts are buggy in this case, and we don't yet have
9809 a Configure test for this. */
9810 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9811 /* 0, point, slack */
9812 Gconvert(nv, (int)digits, 0, ebuf);
9814 if (*ebuf) /* May return an empty string for digits==0 */
9817 } else if (!digits) {
9820 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9821 sv_catpvn(sv, p, l);
9827 #endif /* !USE_LONG_DOUBLE */
9829 if (!args && svix < svmax && DO_UTF8(*svargs))
9832 patend = (char*)pat + patlen;
9833 for (p = (char*)pat; p < patend; p = q) {
9836 bool vectorize = FALSE;
9837 bool vectorarg = FALSE;
9838 bool vec_utf8 = FALSE;
9844 bool has_precis = FALSE;
9846 const I32 osvix = svix;
9847 bool is_utf8 = FALSE; /* is this item utf8? */
9848 #ifdef HAS_LDBL_SPRINTF_BUG
9849 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
9850 with sfio - Allen <allens@cpan.org> */
9851 bool fix_ldbl_sprintf_bug = FALSE;
9855 U8 utf8buf[UTF8_MAXBYTES+1];
9856 STRLEN esignlen = 0;
9858 const char *eptr = NULL;
9859 const char *fmtstart;
9862 const U8 *vecstr = NULL;
9869 /* we need a long double target in case HAS_LONG_DOUBLE but
9872 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9880 const char *dotstr = ".";
9881 STRLEN dotstrlen = 1;
9882 I32 efix = 0; /* explicit format parameter index */
9883 I32 ewix = 0; /* explicit width index */
9884 I32 epix = 0; /* explicit precision index */
9885 I32 evix = 0; /* explicit vector index */
9886 bool asterisk = FALSE;
9888 /* echo everything up to the next format specification */
9889 for (q = p; q < patend && *q != '%'; ++q) ;
9891 if (has_utf8 && !pat_utf8)
9892 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
9894 sv_catpvn(sv, p, q - p);
9903 We allow format specification elements in this order:
9904 \d+\$ explicit format parameter index
9906 v|\*(\d+\$)?v vector with optional (optionally specified) arg
9907 0 flag (as above): repeated to allow "v02"
9908 \d+|\*(\d+\$)? width using optional (optionally specified) arg
9909 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
9911 [%bcdefginopsuxDFOUX] format (mandatory)
9916 As of perl5.9.3, printf format checking is on by default.
9917 Internally, perl uses %p formats to provide an escape to
9918 some extended formatting. This block deals with those
9919 extensions: if it does not match, (char*)q is reset and
9920 the normal format processing code is used.
9922 Currently defined extensions are:
9923 %p include pointer address (standard)
9924 %-p (SVf) include an SV (previously %_)
9925 %-<num>p include an SV with precision <num>
9926 %<num>p reserved for future extensions
9928 Robin Barker 2005-07-14
9930 %1p (VDf) removed. RMB 2007-10-19
9937 n = expect_number(&q);
9944 argsv = MUTABLE_SV(va_arg(*args, void*));
9945 eptr = SvPV_const(argsv, elen);
9951 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
9952 "internal %%<num>p might conflict with future printf extensions");
9958 if ( (width = expect_number(&q)) ) {
9973 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10002 if ( (ewix = expect_number(&q)) )
10011 if ((vectorarg = asterisk)) {
10024 width = expect_number(&q);
10030 vecsv = va_arg(*args, SV*);
10032 vecsv = (evix > 0 && evix <= svmax)
10033 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10035 vecsv = svix < svmax
10036 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10038 dotstr = SvPV_const(vecsv, dotstrlen);
10039 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10040 bad with tied or overloaded values that return UTF8. */
10041 if (DO_UTF8(vecsv))
10043 else if (has_utf8) {
10044 vecsv = sv_mortalcopy(vecsv);
10045 sv_utf8_upgrade(vecsv);
10046 dotstr = SvPV_const(vecsv, dotstrlen);
10053 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10054 vecsv = svargs[efix ? efix-1 : svix++];
10055 vecstr = (U8*)SvPV_const(vecsv,veclen);
10056 vec_utf8 = DO_UTF8(vecsv);
10058 /* if this is a version object, we need to convert
10059 * back into v-string notation and then let the
10060 * vectorize happen normally
10062 if (sv_derived_from(vecsv, "version")) {
10063 char *version = savesvpv(vecsv);
10064 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10065 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
10066 "vector argument not supported with alpha versions");
10069 vecsv = sv_newmortal();
10070 scan_vstring(version, version + veclen, vecsv);
10071 vecstr = (U8*)SvPV_const(vecsv, veclen);
10072 vec_utf8 = DO_UTF8(vecsv);
10084 i = va_arg(*args, int);
10086 i = (ewix ? ewix <= svmax : svix < svmax) ?
10087 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10089 width = (i < 0) ? -i : i;
10099 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10101 /* XXX: todo, support specified precision parameter */
10105 i = va_arg(*args, int);
10107 i = (ewix ? ewix <= svmax : svix < svmax)
10108 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10110 has_precis = !(i < 0);
10114 while (isDIGIT(*q))
10115 precis = precis * 10 + (*q++ - '0');
10124 case 'I': /* Ix, I32x, and I64x */
10126 if (q[1] == '6' && q[2] == '4') {
10132 if (q[1] == '3' && q[2] == '2') {
10142 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10153 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10154 if (*(q + 1) == 'l') { /* lld, llf */
10180 if (!vectorize && !args) {
10182 const I32 i = efix-1;
10183 argsv = (i >= 0 && i < svmax)
10184 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10186 argsv = (svix >= 0 && svix < svmax)
10187 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10191 switch (c = *q++) {
10198 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10200 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10202 eptr = (char*)utf8buf;
10203 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10217 eptr = va_arg(*args, char*);
10219 elen = strlen(eptr);
10221 eptr = (char *)nullstr;
10222 elen = sizeof nullstr - 1;
10226 eptr = SvPV_const(argsv, elen);
10227 if (DO_UTF8(argsv)) {
10228 STRLEN old_precis = precis;
10229 if (has_precis && precis < elen) {
10230 STRLEN ulen = sv_len_utf8(argsv);
10231 I32 p = precis > ulen ? ulen : precis;
10232 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
10235 if (width) { /* fudge width (can't fudge elen) */
10236 if (has_precis && precis < elen)
10237 width += precis - old_precis;
10239 width += elen - sv_len_utf8(argsv);
10246 if (has_precis && precis < elen)
10253 if (alt || vectorize)
10255 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10276 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10285 esignbuf[esignlen++] = plus;
10289 case 'h': iv = (short)va_arg(*args, int); break;
10290 case 'l': iv = va_arg(*args, long); break;
10291 case 'V': iv = va_arg(*args, IV); break;
10292 default: iv = va_arg(*args, int); break;
10295 iv = va_arg(*args, Quad_t); break;
10302 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10304 case 'h': iv = (short)tiv; break;
10305 case 'l': iv = (long)tiv; break;
10307 default: iv = tiv; break;
10310 iv = (Quad_t)tiv; break;
10316 if ( !vectorize ) /* we already set uv above */
10321 esignbuf[esignlen++] = plus;
10325 esignbuf[esignlen++] = '-';
10369 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10380 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10381 case 'l': uv = va_arg(*args, unsigned long); break;
10382 case 'V': uv = va_arg(*args, UV); break;
10383 default: uv = va_arg(*args, unsigned); break;
10386 uv = va_arg(*args, Uquad_t); break;
10393 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10395 case 'h': uv = (unsigned short)tuv; break;
10396 case 'l': uv = (unsigned long)tuv; break;
10398 default: uv = tuv; break;
10401 uv = (Uquad_t)tuv; break;
10410 char *ptr = ebuf + sizeof ebuf;
10411 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10417 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10421 } while (uv >>= 4);
10423 esignbuf[esignlen++] = '0';
10424 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10430 *--ptr = '0' + dig;
10431 } while (uv >>= 3);
10432 if (alt && *ptr != '0')
10438 *--ptr = '0' + dig;
10439 } while (uv >>= 1);
10441 esignbuf[esignlen++] = '0';
10442 esignbuf[esignlen++] = c;
10445 default: /* it had better be ten or less */
10448 *--ptr = '0' + dig;
10449 } while (uv /= base);
10452 elen = (ebuf + sizeof ebuf) - ptr;
10456 zeros = precis - elen;
10457 else if (precis == 0 && elen == 1 && *eptr == '0'
10458 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10461 /* a precision nullifies the 0 flag. */
10468 /* FLOATING POINT */
10471 c = 'f'; /* maybe %F isn't supported here */
10473 case 'e': case 'E':
10475 case 'g': case 'G':
10479 /* This is evil, but floating point is even more evil */
10481 /* for SV-style calling, we can only get NV
10482 for C-style calling, we assume %f is double;
10483 for simplicity we allow any of %Lf, %llf, %qf for long double
10487 #if defined(USE_LONG_DOUBLE)
10491 /* [perl #20339] - we should accept and ignore %lf rather than die */
10495 #if defined(USE_LONG_DOUBLE)
10496 intsize = args ? 0 : 'q';
10500 #if defined(HAS_LONG_DOUBLE)
10509 /* now we need (long double) if intsize == 'q', else (double) */
10511 #if LONG_DOUBLESIZE > DOUBLESIZE
10513 va_arg(*args, long double) :
10514 va_arg(*args, double)
10516 va_arg(*args, double)
10521 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10522 else. frexp() has some unspecified behaviour for those three */
10523 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10525 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10526 will cast our (long double) to (double) */
10527 (void)Perl_frexp(nv, &i);
10528 if (i == PERL_INT_MIN)
10529 Perl_die(aTHX_ "panic: frexp");
10531 need = BIT_DIGITS(i);
10533 need += has_precis ? precis : 6; /* known default */
10538 #ifdef HAS_LDBL_SPRINTF_BUG
10539 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10540 with sfio - Allen <allens@cpan.org> */
10543 # define MY_DBL_MAX DBL_MAX
10544 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10545 # if DOUBLESIZE >= 8
10546 # define MY_DBL_MAX 1.7976931348623157E+308L
10548 # define MY_DBL_MAX 3.40282347E+38L
10552 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10553 # define MY_DBL_MAX_BUG 1L
10555 # define MY_DBL_MAX_BUG MY_DBL_MAX
10559 # define MY_DBL_MIN DBL_MIN
10560 # else /* XXX guessing! -Allen */
10561 # if DOUBLESIZE >= 8
10562 # define MY_DBL_MIN 2.2250738585072014E-308L
10564 # define MY_DBL_MIN 1.17549435E-38L
10568 if ((intsize == 'q') && (c == 'f') &&
10569 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10570 (need < DBL_DIG)) {
10571 /* it's going to be short enough that
10572 * long double precision is not needed */
10574 if ((nv <= 0L) && (nv >= -0L))
10575 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10577 /* would use Perl_fp_class as a double-check but not
10578 * functional on IRIX - see perl.h comments */
10580 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10581 /* It's within the range that a double can represent */
10582 #if defined(DBL_MAX) && !defined(DBL_MIN)
10583 if ((nv >= ((long double)1/DBL_MAX)) ||
10584 (nv <= (-(long double)1/DBL_MAX)))
10586 fix_ldbl_sprintf_bug = TRUE;
10589 if (fix_ldbl_sprintf_bug == TRUE) {
10599 # undef MY_DBL_MAX_BUG
10602 #endif /* HAS_LDBL_SPRINTF_BUG */
10604 need += 20; /* fudge factor */
10605 if (PL_efloatsize < need) {
10606 Safefree(PL_efloatbuf);
10607 PL_efloatsize = need + 20; /* more fudge */
10608 Newx(PL_efloatbuf, PL_efloatsize, char);
10609 PL_efloatbuf[0] = '\0';
10612 if ( !(width || left || plus || alt) && fill != '0'
10613 && has_precis && intsize != 'q' ) { /* Shortcuts */
10614 /* See earlier comment about buggy Gconvert when digits,
10616 if ( c == 'g' && precis) {
10617 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10618 /* May return an empty string for digits==0 */
10619 if (*PL_efloatbuf) {
10620 elen = strlen(PL_efloatbuf);
10621 goto float_converted;
10623 } else if ( c == 'f' && !precis) {
10624 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10629 char *ptr = ebuf + sizeof ebuf;
10632 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10633 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10634 if (intsize == 'q') {
10635 /* Copy the one or more characters in a long double
10636 * format before the 'base' ([efgEFG]) character to
10637 * the format string. */
10638 static char const prifldbl[] = PERL_PRIfldbl;
10639 char const *p = prifldbl + sizeof(prifldbl) - 3;
10640 while (p >= prifldbl) { *--ptr = *p--; }
10645 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10650 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10662 /* No taint. Otherwise we are in the strange situation
10663 * where printf() taints but print($float) doesn't.
10665 #if defined(HAS_LONG_DOUBLE)
10666 elen = ((intsize == 'q')
10667 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10668 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10670 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10674 eptr = PL_efloatbuf;
10682 i = SvCUR(sv) - origlen;
10685 case 'h': *(va_arg(*args, short*)) = i; break;
10686 default: *(va_arg(*args, int*)) = i; break;
10687 case 'l': *(va_arg(*args, long*)) = i; break;
10688 case 'V': *(va_arg(*args, IV*)) = i; break;
10691 *(va_arg(*args, Quad_t*)) = i; break;
10698 sv_setuv_mg(argsv, (UV)i);
10699 continue; /* not "break" */
10706 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10707 && ckWARN(WARN_PRINTF))
10709 SV * const msg = sv_newmortal();
10710 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10711 (PL_op->op_type == OP_PRTF) ? "" : "s");
10712 if (fmtstart < patend) {
10713 const char * const fmtend = q < patend ? q : patend;
10715 sv_catpvs(msg, "\"%");
10716 for (f = fmtstart; f < fmtend; f++) {
10718 sv_catpvn(msg, f, 1);
10720 Perl_sv_catpvf(aTHX_ msg,
10721 "\\%03"UVof, (UV)*f & 0xFF);
10724 sv_catpvs(msg, "\"");
10726 sv_catpvs(msg, "end of string");
10728 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10731 /* output mangled stuff ... */
10737 /* ... right here, because formatting flags should not apply */
10738 SvGROW(sv, SvCUR(sv) + elen + 1);
10740 Copy(eptr, p, elen, char);
10743 SvCUR_set(sv, p - SvPVX_const(sv));
10745 continue; /* not "break" */
10748 if (is_utf8 != has_utf8) {
10751 sv_utf8_upgrade(sv);
10754 const STRLEN old_elen = elen;
10755 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10756 sv_utf8_upgrade(nsv);
10757 eptr = SvPVX_const(nsv);
10760 if (width) { /* fudge width (can't fudge elen) */
10761 width += elen - old_elen;
10767 have = esignlen + zeros + elen;
10769 Perl_croak_nocontext("%s", PL_memory_wrap);
10771 need = (have > width ? have : width);
10774 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10775 Perl_croak_nocontext("%s", PL_memory_wrap);
10776 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10778 if (esignlen && fill == '0') {
10780 for (i = 0; i < (int)esignlen; i++)
10781 *p++ = esignbuf[i];
10783 if (gap && !left) {
10784 memset(p, fill, gap);
10787 if (esignlen && fill != '0') {
10789 for (i = 0; i < (int)esignlen; i++)
10790 *p++ = esignbuf[i];
10794 for (i = zeros; i; i--)
10798 Copy(eptr, p, elen, char);
10802 memset(p, ' ', gap);
10807 Copy(dotstr, p, dotstrlen, char);
10811 vectorize = FALSE; /* done iterating over vecstr */
10818 SvCUR_set(sv, p - SvPVX_const(sv));
10827 /* =========================================================================
10829 =head1 Cloning an interpreter
10831 All the macros and functions in this section are for the private use of
10832 the main function, perl_clone().
10834 The foo_dup() functions make an exact copy of an existing foo thingy.
10835 During the course of a cloning, a hash table is used to map old addresses
10836 to new addresses. The table is created and manipulated with the
10837 ptr_table_* functions.
10841 * =========================================================================*/
10844 #if defined(USE_ITHREADS)
10846 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
10847 #ifndef GpREFCNT_inc
10848 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
10852 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
10853 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
10854 If this changes, please unmerge ss_dup.
10855 Likewise, sv_dup_inc_multiple() relies on this fact. */
10856 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
10857 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
10858 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
10859 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
10860 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
10861 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
10862 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
10863 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
10864 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
10865 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
10866 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
10867 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
10868 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
10870 /* clone a parser */
10873 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
10877 PERL_ARGS_ASSERT_PARSER_DUP;
10882 /* look for it in the table first */
10883 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
10887 /* create anew and remember what it is */
10888 Newxz(parser, 1, yy_parser);
10889 ptr_table_store(PL_ptr_table, proto, parser);
10891 /* XXX these not yet duped */
10892 parser->old_parser = NULL;
10893 parser->stack = NULL;
10895 parser->stack_size = 0;
10896 /* XXX parser->stack->state = 0; */
10898 /* XXX eventually, just Copy() most of the parser struct ? */
10900 parser->lex_brackets = proto->lex_brackets;
10901 parser->lex_casemods = proto->lex_casemods;
10902 parser->lex_brackstack = savepvn(proto->lex_brackstack,
10903 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
10904 parser->lex_casestack = savepvn(proto->lex_casestack,
10905 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
10906 parser->lex_defer = proto->lex_defer;
10907 parser->lex_dojoin = proto->lex_dojoin;
10908 parser->lex_expect = proto->lex_expect;
10909 parser->lex_formbrack = proto->lex_formbrack;
10910 parser->lex_inpat = proto->lex_inpat;
10911 parser->lex_inwhat = proto->lex_inwhat;
10912 parser->lex_op = proto->lex_op;
10913 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
10914 parser->lex_starts = proto->lex_starts;
10915 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
10916 parser->multi_close = proto->multi_close;
10917 parser->multi_open = proto->multi_open;
10918 parser->multi_start = proto->multi_start;
10919 parser->multi_end = proto->multi_end;
10920 parser->pending_ident = proto->pending_ident;
10921 parser->preambled = proto->preambled;
10922 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
10923 parser->linestr = sv_dup_inc(proto->linestr, param);
10924 parser->expect = proto->expect;
10925 parser->copline = proto->copline;
10926 parser->last_lop_op = proto->last_lop_op;
10927 parser->lex_state = proto->lex_state;
10928 parser->rsfp = fp_dup(proto->rsfp, '<', param);
10929 /* rsfp_filters entries have fake IoDIRP() */
10930 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
10931 parser->in_my = proto->in_my;
10932 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
10933 parser->error_count = proto->error_count;
10936 parser->linestr = sv_dup_inc(proto->linestr, param);
10939 char * const ols = SvPVX(proto->linestr);
10940 char * const ls = SvPVX(parser->linestr);
10942 parser->bufptr = ls + (proto->bufptr >= ols ?
10943 proto->bufptr - ols : 0);
10944 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
10945 proto->oldbufptr - ols : 0);
10946 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
10947 proto->oldoldbufptr - ols : 0);
10948 parser->linestart = ls + (proto->linestart >= ols ?
10949 proto->linestart - ols : 0);
10950 parser->last_uni = ls + (proto->last_uni >= ols ?
10951 proto->last_uni - ols : 0);
10952 parser->last_lop = ls + (proto->last_lop >= ols ?
10953 proto->last_lop - ols : 0);
10955 parser->bufend = ls + SvCUR(parser->linestr);
10958 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
10962 parser->endwhite = proto->endwhite;
10963 parser->faketokens = proto->faketokens;
10964 parser->lasttoke = proto->lasttoke;
10965 parser->nextwhite = proto->nextwhite;
10966 parser->realtokenstart = proto->realtokenstart;
10967 parser->skipwhite = proto->skipwhite;
10968 parser->thisclose = proto->thisclose;
10969 parser->thismad = proto->thismad;
10970 parser->thisopen = proto->thisopen;
10971 parser->thisstuff = proto->thisstuff;
10972 parser->thistoken = proto->thistoken;
10973 parser->thiswhite = proto->thiswhite;
10975 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
10976 parser->curforce = proto->curforce;
10978 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
10979 Copy(proto->nexttype, parser->nexttype, 5, I32);
10980 parser->nexttoke = proto->nexttoke;
10983 /* XXX should clone saved_curcop here, but we aren't passed
10984 * proto_perl; so do it in perl_clone_using instead */
10990 /* duplicate a file handle */
10993 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
10997 PERL_ARGS_ASSERT_FP_DUP;
10998 PERL_UNUSED_ARG(type);
11001 return (PerlIO*)NULL;
11003 /* look for it in the table first */
11004 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11008 /* create anew and remember what it is */
11009 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11010 ptr_table_store(PL_ptr_table, fp, ret);
11014 /* duplicate a directory handle */
11017 Perl_dirp_dup(pTHX_ DIR *const dp)
11022 register const Direntry_t *dirent;
11023 char smallbuf[256];
11029 PERL_UNUSED_CONTEXT;
11034 /* look for it in the table first */
11035 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11041 /* open the current directory (so we can switch back) */
11042 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11044 /* chdir to our dir handle and open the present working directory */
11045 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11046 PerlDir_close(pwd);
11047 return (DIR *)NULL;
11049 /* Now we should have two dir handles pointing to the same dir. */
11051 /* Be nice to the calling code and chdir back to where we were. */
11052 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11054 /* We have no need of the pwd handle any more. */
11055 PerlDir_close(pwd);
11058 # define d_namlen(d) (d)->d_namlen
11060 # define d_namlen(d) strlen((d)->d_name)
11062 /* Iterate once through dp, to get the file name at the current posi-
11063 tion. Then step back. */
11064 pos = PerlDir_tell(dp);
11065 if ((dirent = PerlDir_read(dp))) {
11066 len = d_namlen(dirent);
11067 if (len <= sizeof smallbuf) name = smallbuf;
11068 else Newx(name, len, char);
11069 Move(dirent->d_name, name, len, char);
11071 PerlDir_seek(dp, pos);
11073 /* Iterate through the new dir handle, till we find a file with the
11075 if (!dirent) /* just before the end */
11077 pos = PerlDir_tell(ret);
11078 if (PerlDir_read(ret)) continue; /* not there yet */
11079 PerlDir_seek(ret, pos); /* step back */
11083 const long pos0 = PerlDir_tell(ret);
11085 pos = PerlDir_tell(ret);
11086 if ((dirent = PerlDir_read(ret))) {
11087 if (len == d_namlen(dirent)
11088 && memEQ(name, dirent->d_name, len)) {
11090 PerlDir_seek(ret, pos); /* step back */
11093 /* else we are not there yet; keep iterating */
11095 else { /* This is not meant to happen. The best we can do is
11096 reset the iterator to the beginning. */
11097 PerlDir_seek(ret, pos0);
11104 if (name && name != smallbuf)
11107 /* pop it in the pointer table */
11108 ptr_table_store(PL_ptr_table, dp, ret);
11116 /* duplicate a typeglob */
11119 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11123 PERL_ARGS_ASSERT_GP_DUP;
11127 /* look for it in the table first */
11128 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11132 /* create anew and remember what it is */
11134 ptr_table_store(PL_ptr_table, gp, ret);
11137 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11138 on Newxz() to do this for us. */
11139 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11140 ret->gp_io = io_dup_inc(gp->gp_io, param);
11141 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11142 ret->gp_av = av_dup_inc(gp->gp_av, param);
11143 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11144 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11145 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11146 ret->gp_cvgen = gp->gp_cvgen;
11147 ret->gp_line = gp->gp_line;
11148 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11152 /* duplicate a chain of magic */
11155 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11157 MAGIC *mgret = NULL;
11158 MAGIC **mgprev_p = &mgret;
11160 PERL_ARGS_ASSERT_MG_DUP;
11162 for (; mg; mg = mg->mg_moremagic) {
11165 if ((param->flags & CLONEf_JOIN_IN)
11166 && mg->mg_type == PERL_MAGIC_backref)
11167 /* when joining, we let the individual SVs add themselves to
11168 * backref as needed. */
11171 Newx(nmg, 1, MAGIC);
11173 mgprev_p = &(nmg->mg_moremagic);
11175 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11176 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11177 from the original commit adding Perl_mg_dup() - revision 4538.
11178 Similarly there is the annotation "XXX random ptr?" next to the
11179 assignment to nmg->mg_ptr. */
11182 /* FIXME for plugins
11183 if (nmg->mg_type == PERL_MAGIC_qr) {
11184 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11188 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11189 ? nmg->mg_type == PERL_MAGIC_backref
11190 /* The backref AV has its reference
11191 * count deliberately bumped by 1 */
11192 ? SvREFCNT_inc(av_dup_inc((const AV *)
11193 nmg->mg_obj, param))
11194 : sv_dup_inc(nmg->mg_obj, param)
11195 : sv_dup(nmg->mg_obj, param);
11197 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11198 if (nmg->mg_len > 0) {
11199 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11200 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11201 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11203 AMT * const namtp = (AMT*)nmg->mg_ptr;
11204 sv_dup_inc_multiple((SV**)(namtp->table),
11205 (SV**)(namtp->table), NofAMmeth, param);
11208 else if (nmg->mg_len == HEf_SVKEY)
11209 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11211 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11212 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11218 #endif /* USE_ITHREADS */
11220 struct ptr_tbl_arena {
11221 struct ptr_tbl_arena *next;
11222 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11225 /* create a new pointer-mapping table */
11228 Perl_ptr_table_new(pTHX)
11231 PERL_UNUSED_CONTEXT;
11233 Newx(tbl, 1, PTR_TBL_t);
11234 tbl->tbl_max = 511;
11235 tbl->tbl_items = 0;
11236 tbl->tbl_arena = NULL;
11237 tbl->tbl_arena_next = NULL;
11238 tbl->tbl_arena_end = NULL;
11239 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11243 #define PTR_TABLE_HASH(ptr) \
11244 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11246 /* map an existing pointer using a table */
11248 STATIC PTR_TBL_ENT_t *
11249 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11251 PTR_TBL_ENT_t *tblent;
11252 const UV hash = PTR_TABLE_HASH(sv);
11254 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11256 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11257 for (; tblent; tblent = tblent->next) {
11258 if (tblent->oldval == sv)
11265 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11267 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11269 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11270 PERL_UNUSED_CONTEXT;
11272 return tblent ? tblent->newval : NULL;
11275 /* add a new entry to a pointer-mapping table */
11278 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11280 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11282 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11283 PERL_UNUSED_CONTEXT;
11286 tblent->newval = newsv;
11288 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11290 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11291 struct ptr_tbl_arena *new_arena;
11293 Newx(new_arena, 1, struct ptr_tbl_arena);
11294 new_arena->next = tbl->tbl_arena;
11295 tbl->tbl_arena = new_arena;
11296 tbl->tbl_arena_next = new_arena->array;
11297 tbl->tbl_arena_end = new_arena->array
11298 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11301 tblent = tbl->tbl_arena_next++;
11303 tblent->oldval = oldsv;
11304 tblent->newval = newsv;
11305 tblent->next = tbl->tbl_ary[entry];
11306 tbl->tbl_ary[entry] = tblent;
11308 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11309 ptr_table_split(tbl);
11313 /* double the hash bucket size of an existing ptr table */
11316 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11318 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11319 const UV oldsize = tbl->tbl_max + 1;
11320 UV newsize = oldsize * 2;
11323 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11324 PERL_UNUSED_CONTEXT;
11326 Renew(ary, newsize, PTR_TBL_ENT_t*);
11327 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11328 tbl->tbl_max = --newsize;
11329 tbl->tbl_ary = ary;
11330 for (i=0; i < oldsize; i++, ary++) {
11331 PTR_TBL_ENT_t **entp = ary;
11332 PTR_TBL_ENT_t *ent = *ary;
11333 PTR_TBL_ENT_t **curentp;
11336 curentp = ary + oldsize;
11338 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11340 ent->next = *curentp;
11350 /* remove all the entries from a ptr table */
11351 /* Deprecated - will be removed post 5.14 */
11354 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11356 if (tbl && tbl->tbl_items) {
11357 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11359 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11362 struct ptr_tbl_arena *next = arena->next;
11368 tbl->tbl_items = 0;
11369 tbl->tbl_arena = NULL;
11370 tbl->tbl_arena_next = NULL;
11371 tbl->tbl_arena_end = NULL;
11375 /* clear and free a ptr table */
11378 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11380 struct ptr_tbl_arena *arena;
11386 arena = tbl->tbl_arena;
11389 struct ptr_tbl_arena *next = arena->next;
11395 Safefree(tbl->tbl_ary);
11399 #if defined(USE_ITHREADS)
11402 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11404 PERL_ARGS_ASSERT_RVPV_DUP;
11407 if (SvWEAKREF(sstr)) {
11408 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11409 if (param->flags & CLONEf_JOIN_IN) {
11410 /* if joining, we add any back references individually rather
11411 * than copying the whole backref array */
11412 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11416 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11418 else if (SvPVX_const(sstr)) {
11419 /* Has something there */
11421 /* Normal PV - clone whole allocated space */
11422 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11423 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11424 /* Not that normal - actually sstr is copy on write.
11425 But we are a true, independant SV, so: */
11426 SvREADONLY_off(dstr);
11431 /* Special case - not normally malloced for some reason */
11432 if (isGV_with_GP(sstr)) {
11433 /* Don't need to do anything here. */
11435 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11436 /* A "shared" PV - clone it as "shared" PV */
11438 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11442 /* Some other special case - random pointer */
11443 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11448 /* Copy the NULL */
11449 SvPV_set(dstr, NULL);
11453 /* duplicate a list of SVs. source and dest may point to the same memory. */
11455 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11456 SSize_t items, CLONE_PARAMS *const param)
11458 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11460 while (items-- > 0) {
11461 *dest++ = sv_dup_inc(*source++, param);
11467 /* duplicate an SV of any type (including AV, HV etc) */
11470 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11475 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11477 if (SvTYPE(sstr) == SVTYPEMASK) {
11478 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11483 /* look for it in the table first */
11484 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11488 if(param->flags & CLONEf_JOIN_IN) {
11489 /** We are joining here so we don't want do clone
11490 something that is bad **/
11491 if (SvTYPE(sstr) == SVt_PVHV) {
11492 const HEK * const hvname = HvNAME_HEK(sstr);
11494 /** don't clone stashes if they already exist **/
11495 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
11496 ptr_table_store(PL_ptr_table, sstr, dstr);
11502 /* create anew and remember what it is */
11505 #ifdef DEBUG_LEAKING_SCALARS
11506 dstr->sv_debug_optype = sstr->sv_debug_optype;
11507 dstr->sv_debug_line = sstr->sv_debug_line;
11508 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11509 dstr->sv_debug_parent = (SV*)sstr;
11510 FREE_SV_DEBUG_FILE(dstr);
11511 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11514 ptr_table_store(PL_ptr_table, sstr, dstr);
11517 SvFLAGS(dstr) = SvFLAGS(sstr);
11518 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11519 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11522 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11523 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11524 (void*)PL_watch_pvx, SvPVX_const(sstr));
11527 /* don't clone objects whose class has asked us not to */
11528 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11533 switch (SvTYPE(sstr)) {
11535 SvANY(dstr) = NULL;
11538 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11540 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11542 SvIV_set(dstr, SvIVX(sstr));
11546 SvANY(dstr) = new_XNV();
11547 SvNV_set(dstr, SvNVX(sstr));
11549 /* case SVt_BIND: */
11552 /* These are all the types that need complex bodies allocating. */
11554 const svtype sv_type = SvTYPE(sstr);
11555 const struct body_details *const sv_type_details
11556 = bodies_by_type + sv_type;
11560 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11575 assert(sv_type_details->body_size);
11576 if (sv_type_details->arena) {
11577 new_body_inline(new_body, sv_type);
11579 = (void*)((char*)new_body - sv_type_details->offset);
11581 new_body = new_NOARENA(sv_type_details);
11585 SvANY(dstr) = new_body;
11588 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11589 ((char*)SvANY(dstr)) + sv_type_details->offset,
11590 sv_type_details->copy, char);
11592 Copy(((char*)SvANY(sstr)),
11593 ((char*)SvANY(dstr)),
11594 sv_type_details->body_size + sv_type_details->offset, char);
11597 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11598 && !isGV_with_GP(dstr)
11599 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11600 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11602 /* The Copy above means that all the source (unduplicated) pointers
11603 are now in the destination. We can check the flags and the
11604 pointers in either, but it's possible that there's less cache
11605 missing by always going for the destination.
11606 FIXME - instrument and check that assumption */
11607 if (sv_type >= SVt_PVMG) {
11608 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11609 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11610 } else if (SvMAGIC(dstr))
11611 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11613 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11616 /* The cast silences a GCC warning about unhandled types. */
11617 switch ((int)sv_type) {
11627 /* FIXME for plugins */
11628 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11631 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11632 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11633 LvTARG(dstr) = dstr;
11634 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11635 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11637 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11639 /* non-GP case already handled above */
11640 if(isGV_with_GP(sstr)) {
11641 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11642 /* Don't call sv_add_backref here as it's going to be
11643 created as part of the magic cloning of the symbol
11644 table--unless this is during a join and the stash
11645 is not actually being cloned. */
11646 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11647 at the point of this comment. */
11648 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11649 if (param->flags & CLONEf_JOIN_IN)
11650 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
11651 GvGP(dstr) = gp_dup(GvGP(sstr), param);
11652 (void)GpREFCNT_inc(GvGP(dstr));
11656 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11657 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11658 /* I have no idea why fake dirp (rsfps)
11659 should be treated differently but otherwise
11660 we end up with leaks -- sky*/
11661 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11662 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11663 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11665 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11666 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11667 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11668 if (IoDIRP(dstr)) {
11669 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
11672 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11674 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
11676 if (IoOFP(dstr) == IoIFP(sstr))
11677 IoOFP(dstr) = IoIFP(dstr);
11679 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11680 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11681 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11682 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11685 /* avoid cloning an empty array */
11686 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11687 SV **dst_ary, **src_ary;
11688 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11690 src_ary = AvARRAY((const AV *)sstr);
11691 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11692 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11693 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11694 AvALLOC((const AV *)dstr) = dst_ary;
11695 if (AvREAL((const AV *)sstr)) {
11696 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11700 while (items-- > 0)
11701 *dst_ary++ = sv_dup(*src_ary++, param);
11703 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11704 while (items-- > 0) {
11705 *dst_ary++ = &PL_sv_undef;
11709 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11710 AvALLOC((const AV *)dstr) = (SV**)NULL;
11711 AvMAX( (const AV *)dstr) = -1;
11712 AvFILLp((const AV *)dstr) = -1;
11716 if (HvARRAY((const HV *)sstr)) {
11718 const bool sharekeys = !!HvSHAREKEYS(sstr);
11719 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11720 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11722 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11723 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11725 HvARRAY(dstr) = (HE**)darray;
11726 while (i <= sxhv->xhv_max) {
11727 const HE * const source = HvARRAY(sstr)[i];
11728 HvARRAY(dstr)[i] = source
11729 ? he_dup(source, sharekeys, param) : 0;
11734 const struct xpvhv_aux * const saux = HvAUX(sstr);
11735 struct xpvhv_aux * const daux = HvAUX(dstr);
11736 /* This flag isn't copied. */
11737 /* SvOOK_on(hv) attacks the IV flags. */
11738 SvFLAGS(dstr) |= SVf_OOK;
11740 hvname = saux->xhv_name;
11741 daux->xhv_name = hek_dup(hvname, param);
11743 daux->xhv_riter = saux->xhv_riter;
11744 daux->xhv_eiter = saux->xhv_eiter
11745 ? he_dup(saux->xhv_eiter,
11746 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
11747 /* backref array needs refcnt=2; see sv_add_backref */
11748 daux->xhv_backreferences =
11749 (param->flags & CLONEf_JOIN_IN)
11750 /* when joining, we let the individual GVs and
11751 * CVs add themselves to backref as
11752 * needed. This avoids pulling in stuff
11753 * that isn't required, and simplifies the
11754 * case where stashes aren't cloned back
11755 * if they already exist in the parent
11758 : saux->xhv_backreferences
11759 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
11760 ? MUTABLE_AV(SvREFCNT_inc(
11761 sv_dup_inc((const SV *)
11762 saux->xhv_backreferences, param)))
11763 : MUTABLE_AV(sv_dup((const SV *)
11764 saux->xhv_backreferences, param))
11767 daux->xhv_mro_meta = saux->xhv_mro_meta
11768 ? mro_meta_dup(saux->xhv_mro_meta, param)
11771 /* Record stashes for possible cloning in Perl_clone(). */
11773 av_push(param->stashes, dstr);
11777 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11780 if (!(param->flags & CLONEf_COPY_STACKS)) {
11785 /* NOTE: not refcounted */
11786 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
11787 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
11788 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
11790 if (!CvISXSUB(dstr))
11791 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
11793 if (CvCONST(dstr) && CvISXSUB(dstr)) {
11794 CvXSUBANY(dstr).any_ptr =
11795 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
11797 /* don't dup if copying back - CvGV isn't refcounted, so the
11798 * duped GV may never be freed. A bit of a hack! DAPM */
11799 SvANY(MUTABLE_CV(dstr))->xcv_gv =
11801 ? gv_dup_inc(CvGV(sstr), param)
11802 : (param->flags & CLONEf_JOIN_IN)
11804 : gv_dup(CvGV(sstr), param);
11806 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
11808 CvWEAKOUTSIDE(sstr)
11809 ? cv_dup( CvOUTSIDE(dstr), param)
11810 : cv_dup_inc(CvOUTSIDE(dstr), param);
11811 if (!CvISXSUB(dstr))
11812 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
11818 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
11825 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11827 PERL_ARGS_ASSERT_SV_DUP_INC;
11828 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
11832 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11834 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
11835 PERL_ARGS_ASSERT_SV_DUP;
11837 /* Track every SV that (at least initially) had a reference count of 0.
11838 We need to do this by holding an actual reference to it in this array.
11839 If we attempt to cheat, turn AvREAL_off(), and store only pointers
11840 (akin to the stashes hash, and the perl stack), we come unstuck if
11841 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
11842 thread) is manipulated in a CLONE method, because CLONE runs before the
11843 unreferenced array is walked to find SVs still with SvREFCNT() == 0
11844 (and fix things up by giving each a reference via the temps stack).
11845 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
11846 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
11847 before the walk of unreferenced happens and a reference to that is SV
11848 added to the temps stack. At which point we have the same SV considered
11849 to be in use, and free to be re-used. Not good.
11851 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
11852 assert(param->unreferenced);
11853 av_push(param->unreferenced, SvREFCNT_inc(dstr));
11859 /* duplicate a context */
11862 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
11864 PERL_CONTEXT *ncxs;
11866 PERL_ARGS_ASSERT_CX_DUP;
11869 return (PERL_CONTEXT*)NULL;
11871 /* look for it in the table first */
11872 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
11876 /* create anew and remember what it is */
11877 Newx(ncxs, max + 1, PERL_CONTEXT);
11878 ptr_table_store(PL_ptr_table, cxs, ncxs);
11879 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
11882 PERL_CONTEXT * const ncx = &ncxs[ix];
11883 if (CxTYPE(ncx) == CXt_SUBST) {
11884 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
11887 switch (CxTYPE(ncx)) {
11889 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
11890 ? cv_dup_inc(ncx->blk_sub.cv, param)
11891 : cv_dup(ncx->blk_sub.cv,param));
11892 ncx->blk_sub.argarray = (CxHASARGS(ncx)
11893 ? av_dup_inc(ncx->blk_sub.argarray,
11896 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
11898 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
11899 ncx->blk_sub.oldcomppad);
11902 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
11904 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
11906 case CXt_LOOP_LAZYSV:
11907 ncx->blk_loop.state_u.lazysv.end
11908 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
11909 /* We are taking advantage of av_dup_inc and sv_dup_inc
11910 actually being the same function, and order equivalance of
11912 We can assert the later [but only at run time :-(] */
11913 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
11914 (void *) &ncx->blk_loop.state_u.lazysv.cur);
11916 ncx->blk_loop.state_u.ary.ary
11917 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
11918 case CXt_LOOP_LAZYIV:
11919 case CXt_LOOP_PLAIN:
11920 if (CxPADLOOP(ncx)) {
11921 ncx->blk_loop.itervar_u.oldcomppad
11922 = (PAD*)ptr_table_fetch(PL_ptr_table,
11923 ncx->blk_loop.itervar_u.oldcomppad);
11925 ncx->blk_loop.itervar_u.gv
11926 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
11931 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
11932 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
11933 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
11946 /* duplicate a stack info structure */
11949 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
11953 PERL_ARGS_ASSERT_SI_DUP;
11956 return (PERL_SI*)NULL;
11958 /* look for it in the table first */
11959 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
11963 /* create anew and remember what it is */
11964 Newxz(nsi, 1, PERL_SI);
11965 ptr_table_store(PL_ptr_table, si, nsi);
11967 nsi->si_stack = av_dup_inc(si->si_stack, param);
11968 nsi->si_cxix = si->si_cxix;
11969 nsi->si_cxmax = si->si_cxmax;
11970 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
11971 nsi->si_type = si->si_type;
11972 nsi->si_prev = si_dup(si->si_prev, param);
11973 nsi->si_next = si_dup(si->si_next, param);
11974 nsi->si_markoff = si->si_markoff;
11979 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
11980 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
11981 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
11982 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
11983 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
11984 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
11985 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
11986 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
11987 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
11988 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
11989 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
11990 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
11991 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
11992 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
11993 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
11994 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
11997 #define pv_dup_inc(p) SAVEPV(p)
11998 #define pv_dup(p) SAVEPV(p)
11999 #define svp_dup_inc(p,pp) any_dup(p,pp)
12001 /* map any object to the new equivent - either something in the
12002 * ptr table, or something in the interpreter structure
12006 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12010 PERL_ARGS_ASSERT_ANY_DUP;
12013 return (void*)NULL;
12015 /* look for it in the table first */
12016 ret = ptr_table_fetch(PL_ptr_table, v);
12020 /* see if it is part of the interpreter structure */
12021 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12022 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12030 /* duplicate the save stack */
12033 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12036 ANY * const ss = proto_perl->Isavestack;
12037 const I32 max = proto_perl->Isavestack_max;
12038 I32 ix = proto_perl->Isavestack_ix;
12051 void (*dptr) (void*);
12052 void (*dxptr) (pTHX_ void*);
12054 PERL_ARGS_ASSERT_SS_DUP;
12056 Newxz(nss, max, ANY);
12059 const UV uv = POPUV(ss,ix);
12060 const U8 type = (U8)uv & SAVE_MASK;
12062 TOPUV(nss,ix) = uv;
12064 case SAVEt_CLEARSV:
12066 case SAVEt_HELEM: /* hash element */
12067 sv = (const SV *)POPPTR(ss,ix);
12068 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12070 case SAVEt_ITEM: /* normal string */
12071 case SAVEt_GVSV: /* scalar slot in GV */
12072 case SAVEt_SV: /* scalar reference */
12073 sv = (const SV *)POPPTR(ss,ix);
12074 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12077 case SAVEt_MORTALIZESV:
12078 sv = (const SV *)POPPTR(ss,ix);
12079 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12081 case SAVEt_SHARED_PVREF: /* char* in shared space */
12082 c = (char*)POPPTR(ss,ix);
12083 TOPPTR(nss,ix) = savesharedpv(c);
12084 ptr = POPPTR(ss,ix);
12085 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12087 case SAVEt_GENERIC_SVREF: /* generic sv */
12088 case SAVEt_SVREF: /* scalar reference */
12089 sv = (const SV *)POPPTR(ss,ix);
12090 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12091 ptr = POPPTR(ss,ix);
12092 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12094 case SAVEt_HV: /* hash reference */
12095 case SAVEt_AV: /* array reference */
12096 sv = (const SV *) POPPTR(ss,ix);
12097 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12099 case SAVEt_COMPPAD:
12101 sv = (const SV *) POPPTR(ss,ix);
12102 TOPPTR(nss,ix) = sv_dup(sv, param);
12104 case SAVEt_INT: /* int reference */
12105 ptr = POPPTR(ss,ix);
12106 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12107 intval = (int)POPINT(ss,ix);
12108 TOPINT(nss,ix) = intval;
12110 case SAVEt_LONG: /* long reference */
12111 ptr = POPPTR(ss,ix);
12112 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12113 longval = (long)POPLONG(ss,ix);
12114 TOPLONG(nss,ix) = longval;
12116 case SAVEt_I32: /* I32 reference */
12117 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
12118 ptr = POPPTR(ss,ix);
12119 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12121 TOPINT(nss,ix) = i;
12123 case SAVEt_IV: /* IV reference */
12124 ptr = POPPTR(ss,ix);
12125 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12127 TOPIV(nss,ix) = iv;
12129 case SAVEt_HPTR: /* HV* reference */
12130 case SAVEt_APTR: /* AV* reference */
12131 case SAVEt_SPTR: /* SV* reference */
12132 ptr = POPPTR(ss,ix);
12133 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12134 sv = (const SV *)POPPTR(ss,ix);
12135 TOPPTR(nss,ix) = sv_dup(sv, param);
12137 case SAVEt_VPTR: /* random* reference */
12138 ptr = POPPTR(ss,ix);
12139 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12141 case SAVEt_INT_SMALL:
12142 case SAVEt_I32_SMALL:
12143 case SAVEt_I16: /* I16 reference */
12144 case SAVEt_I8: /* I8 reference */
12146 ptr = POPPTR(ss,ix);
12147 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12149 case SAVEt_GENERIC_PVREF: /* generic char* */
12150 case SAVEt_PPTR: /* char* reference */
12151 ptr = POPPTR(ss,ix);
12152 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12153 c = (char*)POPPTR(ss,ix);
12154 TOPPTR(nss,ix) = pv_dup(c);
12156 case SAVEt_GP: /* scalar reference */
12157 gv = (const GV *)POPPTR(ss,ix);
12158 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12159 gp = (GP*)POPPTR(ss,ix);
12160 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12161 (void)GpREFCNT_inc(gp);
12163 TOPINT(nss,ix) = i;
12166 ptr = POPPTR(ss,ix);
12167 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12168 /* these are assumed to be refcounted properly */
12170 switch (((OP*)ptr)->op_type) {
12172 case OP_LEAVESUBLV:
12176 case OP_LEAVEWRITE:
12177 TOPPTR(nss,ix) = ptr;
12180 (void) OpREFCNT_inc(o);
12184 TOPPTR(nss,ix) = NULL;
12189 TOPPTR(nss,ix) = NULL;
12192 hv = (const HV *)POPPTR(ss,ix);
12193 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12195 TOPINT(nss,ix) = i;
12198 c = (char*)POPPTR(ss,ix);
12199 TOPPTR(nss,ix) = pv_dup_inc(c);
12201 case SAVEt_STACK_POS: /* Position on Perl stack */
12203 TOPINT(nss,ix) = i;
12205 case SAVEt_DESTRUCTOR:
12206 ptr = POPPTR(ss,ix);
12207 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12208 dptr = POPDPTR(ss,ix);
12209 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12210 any_dup(FPTR2DPTR(void *, dptr),
12213 case SAVEt_DESTRUCTOR_X:
12214 ptr = POPPTR(ss,ix);
12215 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12216 dxptr = POPDXPTR(ss,ix);
12217 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12218 any_dup(FPTR2DPTR(void *, dxptr),
12221 case SAVEt_REGCONTEXT:
12223 ix -= uv >> SAVE_TIGHT_SHIFT;
12225 case SAVEt_AELEM: /* array element */
12226 sv = (const SV *)POPPTR(ss,ix);
12227 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12229 TOPINT(nss,ix) = i;
12230 av = (const AV *)POPPTR(ss,ix);
12231 TOPPTR(nss,ix) = av_dup_inc(av, param);
12234 ptr = POPPTR(ss,ix);
12235 TOPPTR(nss,ix) = ptr;
12238 ptr = POPPTR(ss,ix);
12241 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
12242 HINTS_REFCNT_UNLOCK;
12244 TOPPTR(nss,ix) = ptr;
12246 TOPINT(nss,ix) = i;
12247 if (i & HINT_LOCALIZE_HH) {
12248 hv = (const HV *)POPPTR(ss,ix);
12249 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12252 case SAVEt_PADSV_AND_MORTALIZE:
12253 longval = (long)POPLONG(ss,ix);
12254 TOPLONG(nss,ix) = longval;
12255 ptr = POPPTR(ss,ix);
12256 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12257 sv = (const SV *)POPPTR(ss,ix);
12258 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12260 case SAVEt_SET_SVFLAGS:
12262 TOPINT(nss,ix) = i;
12264 TOPINT(nss,ix) = i;
12265 sv = (const SV *)POPPTR(ss,ix);
12266 TOPPTR(nss,ix) = sv_dup(sv, param);
12268 case SAVEt_RE_STATE:
12270 const struct re_save_state *const old_state
12271 = (struct re_save_state *)
12272 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12273 struct re_save_state *const new_state
12274 = (struct re_save_state *)
12275 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12277 Copy(old_state, new_state, 1, struct re_save_state);
12278 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12280 new_state->re_state_bostr
12281 = pv_dup(old_state->re_state_bostr);
12282 new_state->re_state_reginput
12283 = pv_dup(old_state->re_state_reginput);
12284 new_state->re_state_regeol
12285 = pv_dup(old_state->re_state_regeol);
12286 new_state->re_state_regoffs
12287 = (regexp_paren_pair*)
12288 any_dup(old_state->re_state_regoffs, proto_perl);
12289 new_state->re_state_reglastparen
12290 = (U32*) any_dup(old_state->re_state_reglastparen,
12292 new_state->re_state_reglastcloseparen
12293 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
12295 /* XXX This just has to be broken. The old save_re_context
12296 code did SAVEGENERICPV(PL_reg_start_tmp);
12297 PL_reg_start_tmp is char **.
12298 Look above to what the dup code does for
12299 SAVEt_GENERIC_PVREF
12300 It can never have worked.
12301 So this is merely a faithful copy of the exiting bug: */
12302 new_state->re_state_reg_start_tmp
12303 = (char **) pv_dup((char *)
12304 old_state->re_state_reg_start_tmp);
12305 /* I assume that it only ever "worked" because no-one called
12306 (pseudo)fork while the regexp engine had re-entered itself.
12308 #ifdef PERL_OLD_COPY_ON_WRITE
12309 new_state->re_state_nrs
12310 = sv_dup(old_state->re_state_nrs, param);
12312 new_state->re_state_reg_magic
12313 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12315 new_state->re_state_reg_oldcurpm
12316 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12318 new_state->re_state_reg_curpm
12319 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12321 new_state->re_state_reg_oldsaved
12322 = pv_dup(old_state->re_state_reg_oldsaved);
12323 new_state->re_state_reg_poscache
12324 = pv_dup(old_state->re_state_reg_poscache);
12325 new_state->re_state_reg_starttry
12326 = pv_dup(old_state->re_state_reg_starttry);
12329 case SAVEt_COMPILE_WARNINGS:
12330 ptr = POPPTR(ss,ix);
12331 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12334 ptr = POPPTR(ss,ix);
12335 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12339 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12347 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12348 * flag to the result. This is done for each stash before cloning starts,
12349 * so we know which stashes want their objects cloned */
12352 do_mark_cloneable_stash(pTHX_ SV *const sv)
12354 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12356 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12357 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12358 if (cloner && GvCV(cloner)) {
12365 mXPUSHs(newSVhek(hvname));
12367 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12374 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12382 =for apidoc perl_clone
12384 Create and return a new interpreter by cloning the current one.
12386 perl_clone takes these flags as parameters:
12388 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12389 without it we only clone the data and zero the stacks,
12390 with it we copy the stacks and the new perl interpreter is
12391 ready to run at the exact same point as the previous one.
12392 The pseudo-fork code uses COPY_STACKS while the
12393 threads->create doesn't.
12395 CLONEf_KEEP_PTR_TABLE
12396 perl_clone keeps a ptr_table with the pointer of the old
12397 variable as a key and the new variable as a value,
12398 this allows it to check if something has been cloned and not
12399 clone it again but rather just use the value and increase the
12400 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12401 the ptr_table using the function
12402 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12403 reason to keep it around is if you want to dup some of your own
12404 variable who are outside the graph perl scans, example of this
12405 code is in threads.xs create
12408 This is a win32 thing, it is ignored on unix, it tells perls
12409 win32host code (which is c++) to clone itself, this is needed on
12410 win32 if you want to run two threads at the same time,
12411 if you just want to do some stuff in a separate perl interpreter
12412 and then throw it away and return to the original one,
12413 you don't need to do anything.
12418 /* XXX the above needs expanding by someone who actually understands it ! */
12419 EXTERN_C PerlInterpreter *
12420 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12423 perl_clone(PerlInterpreter *proto_perl, UV flags)
12426 #ifdef PERL_IMPLICIT_SYS
12428 PERL_ARGS_ASSERT_PERL_CLONE;
12430 /* perlhost.h so we need to call into it
12431 to clone the host, CPerlHost should have a c interface, sky */
12433 if (flags & CLONEf_CLONE_HOST) {
12434 return perl_clone_host(proto_perl,flags);
12436 return perl_clone_using(proto_perl, flags,
12438 proto_perl->IMemShared,
12439 proto_perl->IMemParse,
12441 proto_perl->IStdIO,
12445 proto_perl->IProc);
12449 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12450 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12451 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12452 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12453 struct IPerlDir* ipD, struct IPerlSock* ipS,
12454 struct IPerlProc* ipP)
12456 /* XXX many of the string copies here can be optimized if they're
12457 * constants; they need to be allocated as common memory and just
12458 * their pointers copied. */
12461 CLONE_PARAMS clone_params;
12462 CLONE_PARAMS* const param = &clone_params;
12464 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12466 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12467 #else /* !PERL_IMPLICIT_SYS */
12469 CLONE_PARAMS clone_params;
12470 CLONE_PARAMS* param = &clone_params;
12471 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12473 PERL_ARGS_ASSERT_PERL_CLONE;
12474 #endif /* PERL_IMPLICIT_SYS */
12476 /* for each stash, determine whether its objects should be cloned */
12477 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12478 PERL_SET_THX(my_perl);
12481 PoisonNew(my_perl, 1, PerlInterpreter);
12486 PL_scopestack_name = 0;
12488 PL_savestack_ix = 0;
12489 PL_savestack_max = -1;
12490 PL_sig_pending = 0;
12492 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12493 # ifdef DEBUG_LEAKING_SCALARS
12494 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12496 #else /* !DEBUGGING */
12497 Zero(my_perl, 1, PerlInterpreter);
12498 #endif /* DEBUGGING */
12500 #ifdef PERL_IMPLICIT_SYS
12501 /* host pointers */
12503 PL_MemShared = ipMS;
12504 PL_MemParse = ipMP;
12511 #endif /* PERL_IMPLICIT_SYS */
12513 param->flags = flags;
12514 /* Nothing in the core code uses this, but we make it available to
12515 extensions (using mg_dup). */
12516 param->proto_perl = proto_perl;
12517 /* Likely nothing will use this, but it is initialised to be consistent
12518 with Perl_clone_params_new(). */
12519 param->proto_perl = my_perl;
12520 param->unreferenced = NULL;
12522 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12524 PL_body_arenas = NULL;
12525 Zero(&PL_body_roots, 1, PL_body_roots);
12528 PL_sv_objcount = 0;
12530 PL_sv_arenaroot = NULL;
12532 PL_debug = proto_perl->Idebug;
12534 PL_hash_seed = proto_perl->Ihash_seed;
12535 PL_rehash_seed = proto_perl->Irehash_seed;
12537 #ifdef USE_REENTRANT_API
12538 /* XXX: things like -Dm will segfault here in perlio, but doing
12539 * PERL_SET_CONTEXT(proto_perl);
12540 * breaks too many other things
12542 Perl_reentrant_init(aTHX);
12545 /* create SV map for pointer relocation */
12546 PL_ptr_table = ptr_table_new();
12548 /* initialize these special pointers as early as possible */
12549 SvANY(&PL_sv_undef) = NULL;
12550 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12551 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12552 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
12554 SvANY(&PL_sv_no) = new_XPVNV();
12555 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12556 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12557 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12558 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
12559 SvCUR_set(&PL_sv_no, 0);
12560 SvLEN_set(&PL_sv_no, 1);
12561 SvIV_set(&PL_sv_no, 0);
12562 SvNV_set(&PL_sv_no, 0);
12563 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
12565 SvANY(&PL_sv_yes) = new_XPVNV();
12566 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12567 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12568 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12569 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
12570 SvCUR_set(&PL_sv_yes, 1);
12571 SvLEN_set(&PL_sv_yes, 2);
12572 SvIV_set(&PL_sv_yes, 1);
12573 SvNV_set(&PL_sv_yes, 1);
12574 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
12576 /* dbargs array probably holds garbage */
12579 /* create (a non-shared!) shared string table */
12580 PL_strtab = newHV();
12581 HvSHAREKEYS_off(PL_strtab);
12582 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
12583 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
12585 PL_compiling = proto_perl->Icompiling;
12587 /* These two PVs will be free'd special way so must set them same way op.c does */
12588 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
12589 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
12591 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
12592 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
12594 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
12595 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
12596 if (PL_compiling.cop_hints_hash) {
12598 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
12599 HINTS_REFCNT_UNLOCK;
12601 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
12602 #ifdef PERL_DEBUG_READONLY_OPS
12607 /* pseudo environmental stuff */
12608 PL_origargc = proto_perl->Iorigargc;
12609 PL_origargv = proto_perl->Iorigargv;
12611 param->stashes = newAV(); /* Setup array of objects to call clone on */
12612 /* This makes no difference to the implementation, as it always pushes
12613 and shifts pointers to other SVs without changing their reference
12614 count, with the array becoming empty before it is freed. However, it
12615 makes it conceptually clear what is going on, and will avoid some
12616 work inside av.c, filling slots between AvFILL() and AvMAX() with
12617 &PL_sv_undef, and SvREFCNT_dec()ing those. */
12618 AvREAL_off(param->stashes);
12620 if (!(flags & CLONEf_COPY_STACKS)) {
12621 param->unreferenced = newAV();
12624 /* Set tainting stuff before PerlIO_debug can possibly get called */
12625 PL_tainting = proto_perl->Itainting;
12626 PL_taint_warn = proto_perl->Itaint_warn;
12628 #ifdef PERLIO_LAYERS
12629 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
12630 PerlIO_clone(aTHX_ proto_perl, param);
12633 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
12634 PL_incgv = gv_dup(proto_perl->Iincgv, param);
12635 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
12636 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
12637 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
12638 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
12641 PL_minus_c = proto_perl->Iminus_c;
12642 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
12643 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
12644 PL_localpatches = proto_perl->Ilocalpatches;
12645 PL_splitstr = proto_perl->Isplitstr;
12646 PL_minus_n = proto_perl->Iminus_n;
12647 PL_minus_p = proto_perl->Iminus_p;
12648 PL_minus_l = proto_perl->Iminus_l;
12649 PL_minus_a = proto_perl->Iminus_a;
12650 PL_minus_E = proto_perl->Iminus_E;
12651 PL_minus_F = proto_perl->Iminus_F;
12652 PL_doswitches = proto_perl->Idoswitches;
12653 PL_dowarn = proto_perl->Idowarn;
12654 PL_doextract = proto_perl->Idoextract;
12655 PL_sawampersand = proto_perl->Isawampersand;
12656 PL_unsafe = proto_perl->Iunsafe;
12657 PL_inplace = SAVEPV(proto_perl->Iinplace);
12658 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
12659 PL_perldb = proto_perl->Iperldb;
12660 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12661 PL_exit_flags = proto_perl->Iexit_flags;
12663 /* magical thingies */
12664 /* XXX time(&PL_basetime) when asked for? */
12665 PL_basetime = proto_perl->Ibasetime;
12666 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
12668 PL_maxsysfd = proto_perl->Imaxsysfd;
12669 PL_statusvalue = proto_perl->Istatusvalue;
12671 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12673 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12675 PL_encoding = sv_dup(proto_perl->Iencoding, param);
12677 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
12678 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
12679 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
12682 /* RE engine related */
12683 Zero(&PL_reg_state, 1, struct re_save_state);
12684 PL_reginterp_cnt = 0;
12685 PL_regmatch_slab = NULL;
12687 /* Clone the regex array */
12688 /* ORANGE FIXME for plugins, probably in the SV dup code.
12689 newSViv(PTR2IV(CALLREGDUPE(
12690 INT2PTR(REGEXP *, SvIVX(regex)), param))))
12692 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
12693 PL_regex_pad = AvARRAY(PL_regex_padav);
12695 /* shortcuts to various I/O objects */
12696 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
12697 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
12698 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
12699 PL_defgv = gv_dup(proto_perl->Idefgv, param);
12700 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
12701 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
12702 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
12704 /* shortcuts to regexp stuff */
12705 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
12707 /* shortcuts to misc objects */
12708 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
12710 /* shortcuts to debugging objects */
12711 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
12712 PL_DBline = gv_dup(proto_perl->IDBline, param);
12713 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
12714 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
12715 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
12716 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
12718 /* symbol tables */
12719 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
12720 PL_curstash = hv_dup(proto_perl->Icurstash, param);
12721 PL_debstash = hv_dup(proto_perl->Idebstash, param);
12722 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
12723 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
12725 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
12726 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
12727 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
12728 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
12729 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
12730 PL_endav = av_dup_inc(proto_perl->Iendav, param);
12731 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
12732 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
12734 PL_sub_generation = proto_perl->Isub_generation;
12735 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
12737 /* funky return mechanisms */
12738 PL_forkprocess = proto_perl->Iforkprocess;
12740 /* subprocess state */
12741 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
12743 /* internal state */
12744 PL_maxo = proto_perl->Imaxo;
12745 if (proto_perl->Iop_mask)
12746 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
12749 /* PL_asserting = proto_perl->Iasserting; */
12751 /* current interpreter roots */
12752 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
12754 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
12756 PL_main_start = proto_perl->Imain_start;
12757 PL_eval_root = proto_perl->Ieval_root;
12758 PL_eval_start = proto_perl->Ieval_start;
12760 /* runtime control stuff */
12761 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
12763 PL_filemode = proto_perl->Ifilemode;
12764 PL_lastfd = proto_perl->Ilastfd;
12765 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12768 PL_gensym = proto_perl->Igensym;
12769 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
12770 PL_laststatval = proto_perl->Ilaststatval;
12771 PL_laststype = proto_perl->Ilaststype;
12774 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
12776 /* interpreter atexit processing */
12777 PL_exitlistlen = proto_perl->Iexitlistlen;
12778 if (PL_exitlistlen) {
12779 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12780 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12783 PL_exitlist = (PerlExitListEntry*)NULL;
12785 PL_my_cxt_size = proto_perl->Imy_cxt_size;
12786 if (PL_my_cxt_size) {
12787 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
12788 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
12789 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12790 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
12791 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
12795 PL_my_cxt_list = (void**)NULL;
12796 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12797 PL_my_cxt_keys = (const char**)NULL;
12800 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
12801 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
12802 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
12804 PL_profiledata = NULL;
12806 PL_compcv = cv_dup(proto_perl->Icompcv, param);
12808 PAD_CLONE_VARS(proto_perl, param);
12810 #ifdef HAVE_INTERP_INTERN
12811 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
12814 /* more statics moved here */
12815 PL_generation = proto_perl->Igeneration;
12816 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
12818 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12819 PL_in_clean_all = proto_perl->Iin_clean_all;
12821 PL_uid = proto_perl->Iuid;
12822 PL_euid = proto_perl->Ieuid;
12823 PL_gid = proto_perl->Igid;
12824 PL_egid = proto_perl->Iegid;
12825 PL_nomemok = proto_perl->Inomemok;
12826 PL_an = proto_perl->Ian;
12827 PL_evalseq = proto_perl->Ievalseq;
12828 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12829 PL_origalen = proto_perl->Iorigalen;
12830 #ifdef PERL_USES_PL_PIDSTATUS
12831 PL_pidstatus = newHV(); /* XXX flag for cloning? */
12833 PL_osname = SAVEPV(proto_perl->Iosname);
12834 PL_sighandlerp = proto_perl->Isighandlerp;
12836 PL_runops = proto_perl->Irunops;
12838 PL_parser = parser_dup(proto_perl->Iparser, param);
12840 /* XXX this only works if the saved cop has already been cloned */
12841 if (proto_perl->Iparser) {
12842 PL_parser->saved_curcop = (COP*)any_dup(
12843 proto_perl->Iparser->saved_curcop,
12847 PL_subline = proto_perl->Isubline;
12848 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
12851 PL_cryptseen = proto_perl->Icryptseen;
12854 PL_hints = proto_perl->Ihints;
12856 PL_amagic_generation = proto_perl->Iamagic_generation;
12858 #ifdef USE_LOCALE_COLLATE
12859 PL_collation_ix = proto_perl->Icollation_ix;
12860 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
12861 PL_collation_standard = proto_perl->Icollation_standard;
12862 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12863 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12864 #endif /* USE_LOCALE_COLLATE */
12866 #ifdef USE_LOCALE_NUMERIC
12867 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
12868 PL_numeric_standard = proto_perl->Inumeric_standard;
12869 PL_numeric_local = proto_perl->Inumeric_local;
12870 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
12871 #endif /* !USE_LOCALE_NUMERIC */
12873 /* utf8 character classes */
12874 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
12875 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
12876 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
12877 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
12878 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
12879 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
12880 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
12881 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
12882 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
12883 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
12884 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
12885 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
12886 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
12887 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
12888 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
12889 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
12890 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
12891 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
12892 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
12893 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
12894 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
12895 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
12896 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
12897 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
12898 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
12899 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
12900 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
12901 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
12902 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
12904 /* Did the locale setup indicate UTF-8? */
12905 PL_utf8locale = proto_perl->Iutf8locale;
12906 /* Unicode features (see perlrun/-C) */
12907 PL_unicode = proto_perl->Iunicode;
12909 /* Pre-5.8 signals control */
12910 PL_signals = proto_perl->Isignals;
12912 /* times() ticks per second */
12913 PL_clocktick = proto_perl->Iclocktick;
12915 /* Recursion stopper for PerlIO_find_layer */
12916 PL_in_load_module = proto_perl->Iin_load_module;
12918 /* sort() routine */
12919 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12921 /* Not really needed/useful since the reenrant_retint is "volatile",
12922 * but do it for consistency's sake. */
12923 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12925 /* Hooks to shared SVs and locks. */
12926 PL_sharehook = proto_perl->Isharehook;
12927 PL_lockhook = proto_perl->Ilockhook;
12928 PL_unlockhook = proto_perl->Iunlockhook;
12929 PL_threadhook = proto_perl->Ithreadhook;
12930 PL_destroyhook = proto_perl->Idestroyhook;
12931 PL_signalhook = proto_perl->Isignalhook;
12933 #ifdef THREADS_HAVE_PIDS
12934 PL_ppid = proto_perl->Ippid;
12938 PL_last_swash_hv = NULL; /* reinits on demand */
12939 PL_last_swash_klen = 0;
12940 PL_last_swash_key[0]= '\0';
12941 PL_last_swash_tmps = (U8*)NULL;
12942 PL_last_swash_slen = 0;
12944 PL_glob_index = proto_perl->Iglob_index;
12945 PL_srand_called = proto_perl->Isrand_called;
12947 if (proto_perl->Ipsig_pend) {
12948 Newxz(PL_psig_pend, SIG_SIZE, int);
12951 PL_psig_pend = (int*)NULL;
12954 if (proto_perl->Ipsig_name) {
12955 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
12956 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
12958 PL_psig_ptr = PL_psig_name + SIG_SIZE;
12961 PL_psig_ptr = (SV**)NULL;
12962 PL_psig_name = (SV**)NULL;
12965 /* intrpvar.h stuff */
12967 if (flags & CLONEf_COPY_STACKS) {
12968 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12969 PL_tmps_ix = proto_perl->Itmps_ix;
12970 PL_tmps_max = proto_perl->Itmps_max;
12971 PL_tmps_floor = proto_perl->Itmps_floor;
12972 Newx(PL_tmps_stack, PL_tmps_max, SV*);
12973 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
12974 PL_tmps_ix+1, param);
12976 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
12977 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
12978 Newxz(PL_markstack, i, I32);
12979 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
12980 - proto_perl->Imarkstack);
12981 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
12982 - proto_perl->Imarkstack);
12983 Copy(proto_perl->Imarkstack, PL_markstack,
12984 PL_markstack_ptr - PL_markstack + 1, I32);
12986 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12987 * NOTE: unlike the others! */
12988 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12989 PL_scopestack_max = proto_perl->Iscopestack_max;
12990 Newxz(PL_scopestack, PL_scopestack_max, I32);
12991 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
12994 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
12995 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
12997 /* NOTE: si_dup() looks at PL_markstack */
12998 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13000 /* PL_curstack = PL_curstackinfo->si_stack; */
13001 PL_curstack = av_dup(proto_perl->Icurstack, param);
13002 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13004 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13005 PL_stack_base = AvARRAY(PL_curstack);
13006 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13007 - proto_perl->Istack_base);
13008 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13010 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
13011 * NOTE: unlike the others! */
13012 PL_savestack_ix = proto_perl->Isavestack_ix;
13013 PL_savestack_max = proto_perl->Isavestack_max;
13014 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13015 PL_savestack = ss_dup(proto_perl, param);
13019 ENTER; /* perl_destruct() wants to LEAVE; */
13022 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
13023 PL_top_env = &PL_start_env;
13025 PL_op = proto_perl->Iop;
13028 PL_Xpv = (XPV*)NULL;
13029 my_perl->Ina = proto_perl->Ina;
13031 PL_statbuf = proto_perl->Istatbuf;
13032 PL_statcache = proto_perl->Istatcache;
13033 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13034 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13036 PL_timesbuf = proto_perl->Itimesbuf;
13039 PL_tainted = proto_perl->Itainted;
13040 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
13041 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13042 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13043 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13044 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
13045 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13046 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13047 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13049 PL_restartjmpenv = proto_perl->Irestartjmpenv;
13050 PL_restartop = proto_perl->Irestartop;
13051 PL_in_eval = proto_perl->Iin_eval;
13052 PL_delaymagic = proto_perl->Idelaymagic;
13053 PL_dirty = proto_perl->Idirty;
13054 PL_localizing = proto_perl->Ilocalizing;
13056 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13057 PL_hv_fetch_ent_mh = NULL;
13058 PL_modcount = proto_perl->Imodcount;
13059 PL_lastgotoprobe = NULL;
13060 PL_dumpindent = proto_perl->Idumpindent;
13062 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13063 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13064 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13065 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13066 PL_efloatbuf = NULL; /* reinits on demand */
13067 PL_efloatsize = 0; /* reinits on demand */
13071 PL_screamfirst = NULL;
13072 PL_screamnext = NULL;
13073 PL_maxscream = -1; /* reinits on demand */
13074 PL_lastscream = NULL;
13077 PL_regdummy = proto_perl->Iregdummy;
13078 PL_colorset = 0; /* reinits PL_colors[] */
13079 /*PL_colors[6] = {0,0,0,0,0,0};*/
13083 /* Pluggable optimizer */
13084 PL_peepp = proto_perl->Ipeepp;
13085 PL_rpeepp = proto_perl->Irpeepp;
13086 /* op_free() hook */
13087 PL_opfreehook = proto_perl->Iopfreehook;
13089 PL_stashcache = newHV();
13091 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13092 proto_perl->Iwatchaddr);
13093 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13094 if (PL_debug && PL_watchaddr) {
13095 PerlIO_printf(Perl_debug_log,
13096 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13097 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13098 PTR2UV(PL_watchok));
13101 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13102 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13104 /* Call the ->CLONE method, if it exists, for each of the stashes
13105 identified by sv_dup() above.
13107 while(av_len(param->stashes) != -1) {
13108 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13109 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13110 if (cloner && GvCV(cloner)) {
13115 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13117 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13123 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13124 ptr_table_free(PL_ptr_table);
13125 PL_ptr_table = NULL;
13128 if (!(flags & CLONEf_COPY_STACKS)) {
13129 unreferenced_to_tmp_stack(param->unreferenced);
13132 SvREFCNT_dec(param->stashes);
13134 /* orphaned? eg threads->new inside BEGIN or use */
13135 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13136 SvREFCNT_inc_simple_void(PL_compcv);
13137 SAVEFREESV(PL_compcv);
13144 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13146 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13148 if (AvFILLp(unreferenced) > -1) {
13149 SV **svp = AvARRAY(unreferenced);
13150 SV **const last = svp + AvFILLp(unreferenced);
13154 if (SvREFCNT(*svp) == 1)
13156 } while (++svp <= last);
13158 EXTEND_MORTAL(count);
13159 svp = AvARRAY(unreferenced);
13162 if (SvREFCNT(*svp) == 1) {
13163 /* Our reference is the only one to this SV. This means that
13164 in this thread, the scalar effectively has a 0 reference.
13165 That doesn't work (cleanup never happens), so donate our
13166 reference to it onto the save stack. */
13167 PL_tmps_stack[++PL_tmps_ix] = *svp;
13169 /* As an optimisation, because we are already walking the
13170 entire array, instead of above doing either
13171 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13172 release our reference to the scalar, so that at the end of
13173 the array owns zero references to the scalars it happens to
13174 point to. We are effectively converting the array from
13175 AvREAL() on to AvREAL() off. This saves the av_clear()
13176 (triggered by the SvREFCNT_dec(unreferenced) below) from
13177 walking the array a second time. */
13178 SvREFCNT_dec(*svp);
13181 } while (++svp <= last);
13182 AvREAL_off(unreferenced);
13184 SvREFCNT_dec(unreferenced);
13188 Perl_clone_params_del(CLONE_PARAMS *param)
13190 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13192 PerlInterpreter *const to = param->new_perl;
13194 PerlInterpreter *const was = PERL_GET_THX;
13196 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13202 SvREFCNT_dec(param->stashes);
13203 if (param->unreferenced)
13204 unreferenced_to_tmp_stack(param->unreferenced);
13214 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13217 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13218 does a dTHX; to get the context from thread local storage.
13219 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13220 a version that passes in my_perl. */
13221 PerlInterpreter *const was = PERL_GET_THX;
13222 CLONE_PARAMS *param;
13224 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13230 /* Given that we've set the context, we can do this unshared. */
13231 Newx(param, 1, CLONE_PARAMS);
13234 param->proto_perl = from;
13235 param->new_perl = to;
13236 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13237 AvREAL_off(param->stashes);
13238 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13246 #endif /* USE_ITHREADS */
13249 =head1 Unicode Support
13251 =for apidoc sv_recode_to_utf8
13253 The encoding is assumed to be an Encode object, on entry the PV
13254 of the sv is assumed to be octets in that encoding, and the sv
13255 will be converted into Unicode (and UTF-8).
13257 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13258 is not a reference, nothing is done to the sv. If the encoding is not
13259 an C<Encode::XS> Encoding object, bad things will happen.
13260 (See F<lib/encoding.pm> and L<Encode>).
13262 The PV of the sv is returned.
13267 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13271 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13273 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13287 Passing sv_yes is wrong - it needs to be or'ed set of constants
13288 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13289 remove converted chars from source.
13291 Both will default the value - let them.
13293 XPUSHs(&PL_sv_yes);
13296 call_method("decode", G_SCALAR);
13300 s = SvPV_const(uni, len);
13301 if (s != SvPVX_const(sv)) {
13302 SvGROW(sv, len + 1);
13303 Move(s, SvPVX(sv), len + 1, char);
13304 SvCUR_set(sv, len);
13311 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13315 =for apidoc sv_cat_decode
13317 The encoding is assumed to be an Encode object, the PV of the ssv is
13318 assumed to be octets in that encoding and decoding the input starts
13319 from the position which (PV + *offset) pointed to. The dsv will be
13320 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13321 when the string tstr appears in decoding output or the input ends on
13322 the PV of the ssv. The value which the offset points will be modified
13323 to the last input position on the ssv.
13325 Returns TRUE if the terminator was found, else returns FALSE.
13330 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13331 SV *ssv, int *offset, char *tstr, int tlen)
13336 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13338 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13349 offsv = newSViv(*offset);
13351 mXPUSHp(tstr, tlen);
13353 call_method("cat_decode", G_SCALAR);
13355 ret = SvTRUE(TOPs);
13356 *offset = SvIV(offsv);
13362 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13367 /* ---------------------------------------------------------------------
13369 * support functions for report_uninit()
13372 /* the maxiumum size of array or hash where we will scan looking
13373 * for the undefined element that triggered the warning */
13375 #define FUV_MAX_SEARCH_SIZE 1000
13377 /* Look for an entry in the hash whose value has the same SV as val;
13378 * If so, return a mortal copy of the key. */
13381 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13384 register HE **array;
13387 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13389 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13390 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13393 array = HvARRAY(hv);
13395 for (i=HvMAX(hv); i>0; i--) {
13396 register HE *entry;
13397 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13398 if (HeVAL(entry) != val)
13400 if ( HeVAL(entry) == &PL_sv_undef ||
13401 HeVAL(entry) == &PL_sv_placeholder)
13405 if (HeKLEN(entry) == HEf_SVKEY)
13406 return sv_mortalcopy(HeKEY_sv(entry));
13407 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13413 /* Look for an entry in the array whose value has the same SV as val;
13414 * If so, return the index, otherwise return -1. */
13417 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13421 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13423 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13424 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13427 if (val != &PL_sv_undef) {
13428 SV ** const svp = AvARRAY(av);
13431 for (i=AvFILLp(av); i>=0; i--)
13438 /* S_varname(): return the name of a variable, optionally with a subscript.
13439 * If gv is non-zero, use the name of that global, along with gvtype (one
13440 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13441 * targ. Depending on the value of the subscript_type flag, return:
13444 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13445 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13446 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13447 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13450 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13451 const SV *const keyname, I32 aindex, int subscript_type)
13454 SV * const name = sv_newmortal();
13457 buffer[0] = gvtype;
13460 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13462 gv_fullname4(name, gv, buffer, 0);
13464 if ((unsigned int)SvPVX(name)[1] <= 26) {
13466 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13468 /* Swap the 1 unprintable control character for the 2 byte pretty
13469 version - ie substr($name, 1, 1) = $buffer; */
13470 sv_insert(name, 1, 1, buffer, 2);
13474 CV * const cv = find_runcv(NULL);
13478 if (!cv || !CvPADLIST(cv))
13480 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13481 sv = *av_fetch(av, targ, FALSE);
13482 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
13485 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13486 SV * const sv = newSV(0);
13487 *SvPVX(name) = '$';
13488 Perl_sv_catpvf(aTHX_ name, "{%s}",
13489 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
13492 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13493 *SvPVX(name) = '$';
13494 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13496 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13497 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13498 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13506 =for apidoc find_uninit_var
13508 Find the name of the undefined variable (if any) that caused the operator o
13509 to issue a "Use of uninitialized value" warning.
13510 If match is true, only return a name if it's value matches uninit_sv.
13511 So roughly speaking, if a unary operator (such as OP_COS) generates a
13512 warning, then following the direct child of the op may yield an
13513 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13514 other hand, with OP_ADD there are two branches to follow, so we only print
13515 the variable name if we get an exact match.
13517 The name is returned as a mortal SV.
13519 Assumes that PL_op is the op that originally triggered the error, and that
13520 PL_comppad/PL_curpad points to the currently executing pad.
13526 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13532 const OP *o, *o2, *kid;
13534 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13535 uninit_sv == &PL_sv_placeholder)))
13538 switch (obase->op_type) {
13545 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13546 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13549 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13551 if (pad) { /* @lex, %lex */
13552 sv = PAD_SVl(obase->op_targ);
13556 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13557 /* @global, %global */
13558 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13561 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13563 else /* @{expr}, %{expr} */
13564 return find_uninit_var(cUNOPx(obase)->op_first,
13568 /* attempt to find a match within the aggregate */
13570 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13572 subscript_type = FUV_SUBSCRIPT_HASH;
13575 index = find_array_subscript((const AV *)sv, uninit_sv);
13577 subscript_type = FUV_SUBSCRIPT_ARRAY;
13580 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
13583 return varname(gv, hash ? '%' : '@', obase->op_targ,
13584 keysv, index, subscript_type);
13588 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
13590 return varname(NULL, '$', obase->op_targ,
13591 NULL, 0, FUV_SUBSCRIPT_NONE);
13594 gv = cGVOPx_gv(obase);
13595 if (!gv || (match && GvSV(gv) != uninit_sv))
13597 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13600 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
13603 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
13604 if (!av || SvRMAGICAL(av))
13606 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13607 if (!svp || *svp != uninit_sv)
13610 return varname(NULL, '$', obase->op_targ,
13611 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13614 gv = cGVOPx_gv(obase);
13619 AV *const av = GvAV(gv);
13620 if (!av || SvRMAGICAL(av))
13622 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13623 if (!svp || *svp != uninit_sv)
13626 return varname(gv, '$', 0,
13627 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13632 o = cUNOPx(obase)->op_first;
13633 if (!o || o->op_type != OP_NULL ||
13634 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
13636 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
13640 if (PL_op == obase)
13641 /* $a[uninit_expr] or $h{uninit_expr} */
13642 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
13645 o = cBINOPx(obase)->op_first;
13646 kid = cBINOPx(obase)->op_last;
13648 /* get the av or hv, and optionally the gv */
13650 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
13651 sv = PAD_SV(o->op_targ);
13653 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
13654 && cUNOPo->op_first->op_type == OP_GV)
13656 gv = cGVOPx_gv(cUNOPo->op_first);
13660 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
13665 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
13666 /* index is constant */
13670 if (obase->op_type == OP_HELEM) {
13671 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
13672 if (!he || HeVAL(he) != uninit_sv)
13676 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
13677 if (!svp || *svp != uninit_sv)
13681 if (obase->op_type == OP_HELEM)
13682 return varname(gv, '%', o->op_targ,
13683 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
13685 return varname(gv, '@', o->op_targ, NULL,
13686 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
13689 /* index is an expression;
13690 * attempt to find a match within the aggregate */
13691 if (obase->op_type == OP_HELEM) {
13692 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13694 return varname(gv, '%', o->op_targ,
13695 keysv, 0, FUV_SUBSCRIPT_HASH);
13699 = find_array_subscript((const AV *)sv, uninit_sv);
13701 return varname(gv, '@', o->op_targ,
13702 NULL, index, FUV_SUBSCRIPT_ARRAY);
13707 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
13709 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
13714 /* only examine RHS */
13715 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
13718 o = cUNOPx(obase)->op_first;
13719 if (o->op_type == OP_PUSHMARK)
13722 if (!o->op_sibling) {
13723 /* one-arg version of open is highly magical */
13725 if (o->op_type == OP_GV) { /* open FOO; */
13727 if (match && GvSV(gv) != uninit_sv)
13729 return varname(gv, '$', 0,
13730 NULL, 0, FUV_SUBSCRIPT_NONE);
13732 /* other possibilities not handled are:
13733 * open $x; or open my $x; should return '${*$x}'
13734 * open expr; should return '$'.expr ideally
13740 /* ops where $_ may be an implicit arg */
13744 if ( !(obase->op_flags & OPf_STACKED)) {
13745 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
13746 ? PAD_SVl(obase->op_targ)
13749 sv = sv_newmortal();
13750 sv_setpvs(sv, "$_");
13759 match = 1; /* print etc can return undef on defined args */
13760 /* skip filehandle as it can't produce 'undef' warning */
13761 o = cUNOPx(obase)->op_first;
13762 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
13763 o = o->op_sibling->op_sibling;
13767 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
13769 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
13771 /* the following ops are capable of returning PL_sv_undef even for
13772 * defined arg(s) */
13791 case OP_GETPEERNAME:
13839 case OP_SMARTMATCH:
13848 /* XXX tmp hack: these two may call an XS sub, and currently
13849 XS subs don't have a SUB entry on the context stack, so CV and
13850 pad determination goes wrong, and BAD things happen. So, just
13851 don't try to determine the value under those circumstances.
13852 Need a better fix at dome point. DAPM 11/2007 */
13858 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
13859 if (gv && GvSV(gv) == uninit_sv)
13860 return newSVpvs_flags("$.", SVs_TEMP);
13865 /* def-ness of rval pos() is independent of the def-ness of its arg */
13866 if ( !(obase->op_flags & OPf_MOD))
13871 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
13872 return newSVpvs_flags("${$/}", SVs_TEMP);
13877 if (!(obase->op_flags & OPf_KIDS))
13879 o = cUNOPx(obase)->op_first;
13885 /* if all except one arg are constant, or have no side-effects,
13886 * or are optimized away, then it's unambiguous */
13888 for (kid=o; kid; kid = kid->op_sibling) {
13890 const OPCODE type = kid->op_type;
13891 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
13892 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
13893 || (type == OP_PUSHMARK)
13897 if (o2) { /* more than one found */
13904 return find_uninit_var(o2, uninit_sv, match);
13906 /* scan all args */
13908 sv = find_uninit_var(o, uninit_sv, 1);
13920 =for apidoc report_uninit
13922 Print appropriate "Use of uninitialized variable" warning
13928 Perl_report_uninit(pTHX_ const SV *uninit_sv)
13932 SV* varname = NULL;
13934 varname = find_uninit_var(PL_op, uninit_sv,0);
13936 sv_insert(varname, 0, 0, " ", 1);
13938 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13939 varname ? SvPV_nolen_const(varname) : "",
13940 " in ", OP_DESC(PL_op));
13943 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13949 * c-indentation-style: bsd
13950 * c-basic-offset: 4
13951 * indent-tabs-mode: t
13954 * ex: set ts=8 sts=4 sw=4 noet: