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 Attempt to free all objects pointed to by RVs,
128 and, unless DISABLE_DESTRUCTOR_KLUDGE is defined,
129 try to do the same for all objects indirectly
130 referenced by typeglobs too. Called once from
131 perl_destruct(), prior to calling sv_clean_all()
134 sv_clean_all() / do_clean_all()
135 SvREFCNT_dec(sv) each remaining SV, possibly
136 triggering an sv_free(). It also sets the
137 SVf_BREAK flag on the SV to indicate that the
138 refcnt has been artificially lowered, and thus
139 stopping sv_free() from giving spurious warnings
140 about SVs which unexpectedly have a refcnt
141 of zero. called repeatedly from perl_destruct()
142 until there are no SVs left.
144 =head2 Arena allocator API Summary
146 Private API to rest of sv.c
150 new_XPVNV(), del_XPVGV(),
155 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
159 * ========================================================================= */
162 * "A time to plant, and a time to uproot what was planted..."
166 Perl_offer_nice_chunk(pTHX_ void *const chunk, const U32 chunk_size)
172 PERL_ARGS_ASSERT_OFFER_NICE_CHUNK;
174 new_chunk = (void *)(chunk);
175 new_chunk_size = (chunk_size);
176 if (new_chunk_size > PL_nice_chunk_size) {
177 Safefree(PL_nice_chunk);
178 PL_nice_chunk = (char *) new_chunk;
179 PL_nice_chunk_size = new_chunk_size;
186 # define MEM_LOG_NEW_SV(sv, file, line, func) \
187 Perl_mem_log_new_sv(sv, file, line, func)
188 # define MEM_LOG_DEL_SV(sv, file, line, func) \
189 Perl_mem_log_del_sv(sv, file, line, func)
191 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
192 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
195 #ifdef DEBUG_LEAKING_SCALARS
196 # define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
197 # define DEBUG_SV_SERIAL(sv) \
198 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
199 PTR2UV(sv), (long)(sv)->sv_debug_serial))
201 # define FREE_SV_DEBUG_FILE(sv)
202 # define DEBUG_SV_SERIAL(sv) NOOP
206 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
207 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
208 /* Whilst I'd love to do this, it seems that things like to check on
210 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
212 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
213 PoisonNew(&SvREFCNT(sv), 1, U32)
215 # define SvARENA_CHAIN(sv) SvANY(sv)
216 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
217 # define POSION_SV_HEAD(sv)
220 /* Mark an SV head as unused, and add to free list.
222 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
223 * its refcount artificially decremented during global destruction, so
224 * there may be dangling pointers to it. The last thing we want in that
225 * case is for it to be reused. */
227 #define plant_SV(p) \
229 const U32 old_flags = SvFLAGS(p); \
230 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
231 DEBUG_SV_SERIAL(p); \
232 FREE_SV_DEBUG_FILE(p); \
234 SvFLAGS(p) = SVTYPEMASK; \
235 if (!(old_flags & SVf_BREAK)) { \
236 SvARENA_CHAIN_SET(p, PL_sv_root); \
242 #define uproot_SV(p) \
245 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
250 /* make some more SVs by adding another arena */
259 sv_add_arena(PL_nice_chunk, PL_nice_chunk_size, 0);
260 PL_nice_chunk = NULL;
261 PL_nice_chunk_size = 0;
264 char *chunk; /* must use New here to match call to */
265 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
266 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
272 /* new_SV(): return a new, empty SV head */
274 #ifdef DEBUG_LEAKING_SCALARS
275 /* provide a real function for a debugger to play with */
277 S_new_SV(pTHX_ const char *file, int line, const char *func)
284 sv = S_more_sv(aTHX);
288 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
289 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
295 sv->sv_debug_inpad = 0;
296 sv->sv_debug_parent = NULL;
297 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
299 sv->sv_debug_serial = PL_sv_serial++;
301 MEM_LOG_NEW_SV(sv, file, line, func);
302 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
303 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
307 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
315 (p) = S_more_sv(aTHX); \
319 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
324 /* del_SV(): return an empty SV head to the free list */
337 S_del_sv(pTHX_ SV *p)
341 PERL_ARGS_ASSERT_DEL_SV;
346 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
347 const SV * const sv = sva + 1;
348 const SV * const svend = &sva[SvREFCNT(sva)];
349 if (p >= sv && p < svend) {
355 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
356 "Attempt to free non-arena SV: 0x%"UVxf
357 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
364 #else /* ! DEBUGGING */
366 #define del_SV(p) plant_SV(p)
368 #endif /* DEBUGGING */
372 =head1 SV Manipulation Functions
374 =for apidoc sv_add_arena
376 Given a chunk of memory, link it to the head of the list of arenas,
377 and split it into a list of free SVs.
383 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
386 SV *const sva = MUTABLE_SV(ptr);
390 PERL_ARGS_ASSERT_SV_ADD_ARENA;
392 /* The first SV in an arena isn't an SV. */
393 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
394 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
395 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
397 PL_sv_arenaroot = sva;
398 PL_sv_root = sva + 1;
400 svend = &sva[SvREFCNT(sva) - 1];
403 SvARENA_CHAIN_SET(sv, (sv + 1));
407 /* Must always set typemask because it's always checked in on cleanup
408 when the arenas are walked looking for objects. */
409 SvFLAGS(sv) = SVTYPEMASK;
412 SvARENA_CHAIN_SET(sv, 0);
416 SvFLAGS(sv) = SVTYPEMASK;
419 /* visit(): call the named function for each non-free SV in the arenas
420 * whose flags field matches the flags/mask args. */
423 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
429 PERL_ARGS_ASSERT_VISIT;
431 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
432 register const SV * const svend = &sva[SvREFCNT(sva)];
434 for (sv = sva + 1; sv < svend; ++sv) {
435 if (SvTYPE(sv) != SVTYPEMASK
436 && (sv->sv_flags & mask) == flags
449 /* called by sv_report_used() for each live SV */
452 do_report_used(pTHX_ SV *const sv)
454 if (SvTYPE(sv) != SVTYPEMASK) {
455 PerlIO_printf(Perl_debug_log, "****\n");
462 =for apidoc sv_report_used
464 Dump the contents of all SVs not yet freed. (Debugging aid).
470 Perl_sv_report_used(pTHX)
473 visit(do_report_used, 0, 0);
479 /* called by sv_clean_objs() for each live SV */
482 do_clean_objs(pTHX_ SV *const ref)
487 SV * const target = SvRV(ref);
488 if (SvOBJECT(target)) {
489 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
490 if (SvWEAKREF(ref)) {
491 sv_del_backref(target, ref);
497 SvREFCNT_dec(target);
502 /* XXX Might want to check arrays, etc. */
505 /* called by sv_clean_objs() for each live SV */
507 #ifndef DISABLE_DESTRUCTOR_KLUDGE
509 do_clean_named_objs(pTHX_ SV *const sv)
512 assert(SvTYPE(sv) == SVt_PVGV);
513 assert(isGV_with_GP(sv));
516 #ifdef PERL_DONT_CREATE_GVSV
519 SvOBJECT(GvSV(sv))) ||
520 (GvAV(sv) && SvOBJECT(GvAV(sv))) ||
521 (GvHV(sv) && SvOBJECT(GvHV(sv))) ||
522 /* In certain rare cases GvIOp(sv) can be NULL, which would make SvOBJECT(GvIO(sv)) dereference NULL. */
523 (GvIO(sv) ? (SvFLAGS(GvIOp(sv)) & SVs_OBJECT) : 0) ||
524 (GvCV(sv) && SvOBJECT(GvCV(sv))) )
526 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning named glob object:\n "), sv_dump(sv)));
527 SvFLAGS(sv) |= SVf_BREAK;
535 =for apidoc sv_clean_objs
537 Attempt to destroy all objects not yet freed
543 Perl_sv_clean_objs(pTHX)
546 PL_in_clean_objs = TRUE;
547 visit(do_clean_objs, SVf_ROK, SVf_ROK);
548 #ifndef DISABLE_DESTRUCTOR_KLUDGE
549 /* some barnacles may yet remain, clinging to typeglobs */
550 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
552 PL_in_clean_objs = FALSE;
555 /* called by sv_clean_all() for each live SV */
558 do_clean_all(pTHX_ SV *const sv)
561 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
562 /* don't clean pid table and strtab */
565 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
566 SvFLAGS(sv) |= SVf_BREAK;
571 =for apidoc sv_clean_all
573 Decrement the refcnt of each remaining SV, possibly triggering a
574 cleanup. This function may have to be called multiple times to free
575 SVs which are in complex self-referential hierarchies.
581 Perl_sv_clean_all(pTHX)
585 PL_in_clean_all = TRUE;
586 cleaned = visit(do_clean_all, 0,0);
587 PL_in_clean_all = FALSE;
592 ARENASETS: a meta-arena implementation which separates arena-info
593 into struct arena_set, which contains an array of struct
594 arena_descs, each holding info for a single arena. By separating
595 the meta-info from the arena, we recover the 1st slot, formerly
596 borrowed for list management. The arena_set is about the size of an
597 arena, avoiding the needless malloc overhead of a naive linked-list.
599 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
600 memory in the last arena-set (1/2 on average). In trade, we get
601 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
602 smaller types). The recovery of the wasted space allows use of
603 small arenas for large, rare body types, by changing array* fields
604 in body_details_by_type[] below.
607 char *arena; /* the raw storage, allocated aligned */
608 size_t size; /* its size ~4k typ */
609 svtype utype; /* bodytype stored in arena */
614 /* Get the maximum number of elements in set[] such that struct arena_set
615 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
616 therefore likely to be 1 aligned memory page. */
618 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
619 - 2 * sizeof(int)) / sizeof (struct arena_desc))
622 struct arena_set* next;
623 unsigned int set_size; /* ie ARENAS_PER_SET */
624 unsigned int curr; /* index of next available arena-desc */
625 struct arena_desc set[ARENAS_PER_SET];
629 =for apidoc sv_free_arenas
631 Deallocate the memory used by all arenas. Note that all the individual SV
632 heads and bodies within the arenas must already have been freed.
637 Perl_sv_free_arenas(pTHX)
644 /* Free arenas here, but be careful about fake ones. (We assume
645 contiguity of the fake ones with the corresponding real ones.) */
647 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
648 svanext = MUTABLE_SV(SvANY(sva));
649 while (svanext && SvFAKE(svanext))
650 svanext = MUTABLE_SV(SvANY(svanext));
657 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
660 struct arena_set *current = aroot;
663 assert(aroot->set[i].arena);
664 Safefree(aroot->set[i].arena);
672 i = PERL_ARENA_ROOTS_SIZE;
674 PL_body_roots[i] = 0;
676 Safefree(PL_nice_chunk);
677 PL_nice_chunk = NULL;
678 PL_nice_chunk_size = 0;
684 Here are mid-level routines that manage the allocation of bodies out
685 of the various arenas. There are 5 kinds of arenas:
687 1. SV-head arenas, which are discussed and handled above
688 2. regular body arenas
689 3. arenas for reduced-size bodies
692 Arena types 2 & 3 are chained by body-type off an array of
693 arena-root pointers, which is indexed by svtype. Some of the
694 larger/less used body types are malloced singly, since a large
695 unused block of them is wasteful. Also, several svtypes dont have
696 bodies; the data fits into the sv-head itself. The arena-root
697 pointer thus has a few unused root-pointers (which may be hijacked
698 later for arena types 4,5)
700 3 differs from 2 as an optimization; some body types have several
701 unused fields in the front of the structure (which are kept in-place
702 for consistency). These bodies can be allocated in smaller chunks,
703 because the leading fields arent accessed. Pointers to such bodies
704 are decremented to point at the unused 'ghost' memory, knowing that
705 the pointers are used with offsets to the real memory.
707 HE, HEK arenas are managed separately, with separate code, but may
708 be merge-able later..
711 /* get_arena(size): this creates custom-sized arenas
712 TBD: export properly for hv.c: S_more_he().
715 Perl_get_arena(pTHX_ const size_t arena_size, const svtype bodytype)
718 struct arena_desc* adesc;
719 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
722 /* shouldnt need this
723 if (!arena_size) arena_size = PERL_ARENA_SIZE;
726 /* may need new arena-set to hold new arena */
727 if (!aroot || aroot->curr >= aroot->set_size) {
728 struct arena_set *newroot;
729 Newxz(newroot, 1, struct arena_set);
730 newroot->set_size = ARENAS_PER_SET;
731 newroot->next = aroot;
733 PL_body_arenas = (void *) newroot;
734 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
737 /* ok, now have arena-set with at least 1 empty/available arena-desc */
738 curr = aroot->curr++;
739 adesc = &(aroot->set[curr]);
740 assert(!adesc->arena);
742 Newx(adesc->arena, arena_size, char);
743 adesc->size = arena_size;
744 adesc->utype = bodytype;
745 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
746 curr, (void*)adesc->arena, (UV)arena_size));
752 /* return a thing to the free list */
754 #define del_body(thing, root) \
756 void ** const thing_copy = (void **)thing;\
757 *thing_copy = *root; \
758 *root = (void*)thing_copy; \
763 =head1 SV-Body Allocation
765 Allocation of SV-bodies is similar to SV-heads, differing as follows;
766 the allocation mechanism is used for many body types, so is somewhat
767 more complicated, it uses arena-sets, and has no need for still-live
770 At the outermost level, (new|del)_X*V macros return bodies of the
771 appropriate type. These macros call either (new|del)_body_type or
772 (new|del)_body_allocated macro pairs, depending on specifics of the
773 type. Most body types use the former pair, the latter pair is used to
774 allocate body types with "ghost fields".
776 "ghost fields" are fields that are unused in certain types, and
777 consequently don't need to actually exist. They are declared because
778 they're part of a "base type", which allows use of functions as
779 methods. The simplest examples are AVs and HVs, 2 aggregate types
780 which don't use the fields which support SCALAR semantics.
782 For these types, the arenas are carved up into appropriately sized
783 chunks, we thus avoid wasted memory for those unaccessed members.
784 When bodies are allocated, we adjust the pointer back in memory by the
785 size of the part not allocated, so it's as if we allocated the full
786 structure. (But things will all go boom if you write to the part that
787 is "not there", because you'll be overwriting the last members of the
788 preceding structure in memory.)
790 We calculate the correction using the STRUCT_OFFSET macro on the first
791 member present. If the allocated structure is smaller (no initial NV
792 actually allocated) then the net effect is to subtract the size of the NV
793 from the pointer, to return a new pointer as if an initial NV were actually
794 allocated. (We were using structures named *_allocated for this, but
795 this turned out to be a subtle bug, because a structure without an NV
796 could have a lower alignment constraint, but the compiler is allowed to
797 optimised accesses based on the alignment constraint of the actual pointer
798 to the full structure, for example, using a single 64 bit load instruction
799 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
801 This is the same trick as was used for NV and IV bodies. Ironically it
802 doesn't need to be used for NV bodies any more, because NV is now at
803 the start of the structure. IV bodies don't need it either, because
804 they are no longer allocated.
806 In turn, the new_body_* allocators call S_new_body(), which invokes
807 new_body_inline macro, which takes a lock, and takes a body off the
808 linked list at PL_body_roots[sv_type], calling S_more_bodies() if
809 necessary to refresh an empty list. Then the lock is released, and
810 the body is returned.
812 S_more_bodies calls get_arena(), and carves it up into an array of N
813 bodies, which it strings into a linked list. It looks up arena-size
814 and body-size from the body_details table described below, thus
815 supporting the multiple body-types.
817 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
818 the (new|del)_X*V macros are mapped directly to malloc/free.
824 For each sv-type, struct body_details bodies_by_type[] carries
825 parameters which control these aspects of SV handling:
827 Arena_size determines whether arenas are used for this body type, and if
828 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
829 zero, forcing individual mallocs and frees.
831 Body_size determines how big a body is, and therefore how many fit into
832 each arena. Offset carries the body-pointer adjustment needed for
833 "ghost fields", and is used in *_allocated macros.
835 But its main purpose is to parameterize info needed in
836 Perl_sv_upgrade(). The info here dramatically simplifies the function
837 vs the implementation in 5.8.8, making it table-driven. All fields
838 are used for this, except for arena_size.
840 For the sv-types that have no bodies, arenas are not used, so those
841 PL_body_roots[sv_type] are unused, and can be overloaded. In
842 something of a special case, SVt_NULL is borrowed for HE arenas;
843 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
844 bodies_by_type[SVt_NULL] slot is not used, as the table is not
849 struct body_details {
850 U8 body_size; /* Size to allocate */
851 U8 copy; /* Size of structure to copy (may be shorter) */
853 unsigned int type : 4; /* We have space for a sanity check. */
854 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
855 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
856 unsigned int arena : 1; /* Allocated from an arena */
857 size_t arena_size; /* Size of arena to allocate */
865 /* With -DPURFIY we allocate everything directly, and don't use arenas.
866 This seems a rather elegant way to simplify some of the code below. */
867 #define HASARENA FALSE
869 #define HASARENA TRUE
871 #define NOARENA FALSE
873 /* Size the arenas to exactly fit a given number of bodies. A count
874 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
875 simplifying the default. If count > 0, the arena is sized to fit
876 only that many bodies, allowing arenas to be used for large, rare
877 bodies (XPVFM, XPVIO) without undue waste. The arena size is
878 limited by PERL_ARENA_SIZE, so we can safely oversize the
881 #define FIT_ARENA0(body_size) \
882 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
883 #define FIT_ARENAn(count,body_size) \
884 ( count * body_size <= PERL_ARENA_SIZE) \
885 ? count * body_size \
886 : FIT_ARENA0 (body_size)
887 #define FIT_ARENA(count,body_size) \
889 ? FIT_ARENAn (count, body_size) \
890 : FIT_ARENA0 (body_size)
892 /* Calculate the length to copy. Specifically work out the length less any
893 final padding the compiler needed to add. See the comment in sv_upgrade
894 for why copying the padding proved to be a bug. */
896 #define copy_length(type, last_member) \
897 STRUCT_OFFSET(type, last_member) \
898 + sizeof (((type*)SvANY((const SV *)0))->last_member)
900 static const struct body_details bodies_by_type[] = {
901 { sizeof(HE), 0, 0, SVt_NULL,
902 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
904 /* The bind placeholder pretends to be an RV for now.
905 Also it's marked as "can't upgrade" to stop anyone using it before it's
907 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
909 /* IVs are in the head, so the allocation size is 0. */
911 sizeof(IV), /* This is used to copy out the IV body. */
912 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
913 NOARENA /* IVS don't need an arena */, 0
916 /* 8 bytes on most ILP32 with IEEE doubles */
917 { sizeof(NV), sizeof(NV),
918 STRUCT_OFFSET(XPVNV, xnv_u),
919 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
921 /* 8 bytes on most ILP32 with IEEE doubles */
922 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
923 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
924 + STRUCT_OFFSET(XPV, xpv_cur),
925 SVt_PV, FALSE, NONV, HASARENA,
926 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
929 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
930 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
931 + STRUCT_OFFSET(XPV, xpv_cur),
932 SVt_PVIV, FALSE, NONV, HASARENA,
933 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
936 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
937 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
938 + STRUCT_OFFSET(XPV, xpv_cur),
939 SVt_PVNV, FALSE, HADNV, HASARENA,
940 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
943 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
944 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
950 SVt_REGEXP, FALSE, NONV, HASARENA,
951 FIT_ARENA(0, sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur))
955 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
956 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
959 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
960 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
963 copy_length(XPVAV, xav_alloc),
965 SVt_PVAV, TRUE, NONV, HASARENA,
966 FIT_ARENA(0, sizeof(XPVAV)) },
969 copy_length(XPVHV, xhv_max),
971 SVt_PVHV, TRUE, NONV, HASARENA,
972 FIT_ARENA(0, sizeof(XPVHV)) },
978 SVt_PVCV, TRUE, NONV, HASARENA,
979 FIT_ARENA(0, sizeof(XPVCV)) },
984 SVt_PVFM, TRUE, NONV, NOARENA,
985 FIT_ARENA(20, sizeof(XPVFM)) },
987 /* XPVIO is 84 bytes, fits 48x */
991 SVt_PVIO, TRUE, NONV, HASARENA,
992 FIT_ARENA(24, sizeof(XPVIO)) },
995 #define new_body_allocated(sv_type) \
996 (void *)((char *)S_new_body(aTHX_ sv_type) \
997 - bodies_by_type[sv_type].offset)
999 #define del_body_allocated(p, sv_type) \
1000 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
1004 #define new_XNV() safemalloc(sizeof(XPVNV))
1005 #define new_XPVNV() safemalloc(sizeof(XPVNV))
1006 #define new_XPVMG() safemalloc(sizeof(XPVMG))
1008 #define del_XPVGV(p) safefree(p)
1012 #define new_XNV() new_body_allocated(SVt_NV)
1013 #define new_XPVNV() new_body_allocated(SVt_PVNV)
1014 #define new_XPVMG() new_body_allocated(SVt_PVMG)
1016 #define del_XPVGV(p) del_body_allocated(p, SVt_PVGV)
1020 /* no arena for you! */
1022 #define new_NOARENA(details) \
1023 safemalloc((details)->body_size + (details)->offset)
1024 #define new_NOARENAZ(details) \
1025 safecalloc((details)->body_size + (details)->offset, 1)
1028 S_more_bodies (pTHX_ const svtype sv_type)
1031 void ** const root = &PL_body_roots[sv_type];
1032 const struct body_details * const bdp = &bodies_by_type[sv_type];
1033 const size_t body_size = bdp->body_size;
1036 const size_t arena_size = Perl_malloc_good_size(bdp->arena_size);
1037 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1038 static bool done_sanity_check;
1040 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1041 * variables like done_sanity_check. */
1042 if (!done_sanity_check) {
1043 unsigned int i = SVt_LAST;
1045 done_sanity_check = TRUE;
1048 assert (bodies_by_type[i].type == i);
1052 assert(bdp->arena_size);
1054 start = (char*) Perl_get_arena(aTHX_ arena_size, sv_type);
1056 end = start + arena_size - 2 * body_size;
1058 /* computed count doesnt reflect the 1st slot reservation */
1059 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1060 DEBUG_m(PerlIO_printf(Perl_debug_log,
1061 "arena %p end %p arena-size %d (from %d) type %d "
1063 (void*)start, (void*)end, (int)arena_size,
1064 (int)bdp->arena_size, sv_type, (int)body_size,
1065 (int)arena_size / (int)body_size));
1067 DEBUG_m(PerlIO_printf(Perl_debug_log,
1068 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1069 (void*)start, (void*)end,
1070 (int)bdp->arena_size, sv_type, (int)body_size,
1071 (int)bdp->arena_size / (int)body_size));
1073 *root = (void *)start;
1075 while (start <= end) {
1076 char * const next = start + body_size;
1077 *(void**) start = (void *)next;
1080 *(void **)start = 0;
1085 /* grab a new thing from the free list, allocating more if necessary.
1086 The inline version is used for speed in hot routines, and the
1087 function using it serves the rest (unless PURIFY).
1089 #define new_body_inline(xpv, sv_type) \
1091 void ** const r3wt = &PL_body_roots[sv_type]; \
1092 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1093 ? *((void **)(r3wt)) : more_bodies(sv_type)); \
1094 *(r3wt) = *(void**)(xpv); \
1100 S_new_body(pTHX_ const svtype sv_type)
1104 new_body_inline(xpv, sv_type);
1110 static const struct body_details fake_rv =
1111 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1114 =for apidoc sv_upgrade
1116 Upgrade an SV to a more complex form. Generally adds a new body type to the
1117 SV, then copies across as much information as possible from the old body.
1118 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1124 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1129 const svtype old_type = SvTYPE(sv);
1130 const struct body_details *new_type_details;
1131 const struct body_details *old_type_details
1132 = bodies_by_type + old_type;
1133 SV *referant = NULL;
1135 PERL_ARGS_ASSERT_SV_UPGRADE;
1137 if (old_type == new_type)
1140 /* This clause was purposefully added ahead of the early return above to
1141 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1142 inference by Nick I-S that it would fix other troublesome cases. See
1143 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1145 Given that shared hash key scalars are no longer PVIV, but PV, there is
1146 no longer need to unshare so as to free up the IVX slot for its proper
1147 purpose. So it's safe to move the early return earlier. */
1149 if (new_type != SVt_PV && SvIsCOW(sv)) {
1150 sv_force_normal_flags(sv, 0);
1153 old_body = SvANY(sv);
1155 /* Copying structures onto other structures that have been neatly zeroed
1156 has a subtle gotcha. Consider XPVMG
1158 +------+------+------+------+------+-------+-------+
1159 | NV | CUR | LEN | IV | MAGIC | STASH |
1160 +------+------+------+------+------+-------+-------+
1161 0 4 8 12 16 20 24 28
1163 where NVs are aligned to 8 bytes, so that sizeof that structure is
1164 actually 32 bytes long, with 4 bytes of padding at the end:
1166 +------+------+------+------+------+-------+-------+------+
1167 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1168 +------+------+------+------+------+-------+-------+------+
1169 0 4 8 12 16 20 24 28 32
1171 so what happens if you allocate memory for this structure:
1173 +------+------+------+------+------+-------+-------+------+------+...
1174 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1175 +------+------+------+------+------+-------+-------+------+------+...
1176 0 4 8 12 16 20 24 28 32 36
1178 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1179 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1180 started out as zero once, but it's quite possible that it isn't. So now,
1181 rather than a nicely zeroed GP, you have it pointing somewhere random.
1184 (In fact, GP ends up pointing at a previous GP structure, because the
1185 principle cause of the padding in XPVMG getting garbage is a copy of
1186 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1187 this happens to be moot because XPVGV has been re-ordered, with GP
1188 no longer after STASH)
1190 So we are careful and work out the size of used parts of all the
1198 referant = SvRV(sv);
1199 old_type_details = &fake_rv;
1200 if (new_type == SVt_NV)
1201 new_type = SVt_PVNV;
1203 if (new_type < SVt_PVIV) {
1204 new_type = (new_type == SVt_NV)
1205 ? SVt_PVNV : SVt_PVIV;
1210 if (new_type < SVt_PVNV) {
1211 new_type = SVt_PVNV;
1215 assert(new_type > SVt_PV);
1216 assert(SVt_IV < SVt_PV);
1217 assert(SVt_NV < SVt_PV);
1224 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1225 there's no way that it can be safely upgraded, because perl.c
1226 expects to Safefree(SvANY(PL_mess_sv)) */
1227 assert(sv != PL_mess_sv);
1228 /* This flag bit is used to mean other things in other scalar types.
1229 Given that it only has meaning inside the pad, it shouldn't be set
1230 on anything that can get upgraded. */
1231 assert(!SvPAD_TYPED(sv));
1234 if (old_type_details->cant_upgrade)
1235 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1236 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1239 if (old_type > new_type)
1240 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1241 (int)old_type, (int)new_type);
1243 new_type_details = bodies_by_type + new_type;
1245 SvFLAGS(sv) &= ~SVTYPEMASK;
1246 SvFLAGS(sv) |= new_type;
1248 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1249 the return statements above will have triggered. */
1250 assert (new_type != SVt_NULL);
1253 assert(old_type == SVt_NULL);
1254 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1258 assert(old_type == SVt_NULL);
1259 SvANY(sv) = new_XNV();
1264 assert(new_type_details->body_size);
1267 assert(new_type_details->arena);
1268 assert(new_type_details->arena_size);
1269 /* This points to the start of the allocated area. */
1270 new_body_inline(new_body, new_type);
1271 Zero(new_body, new_type_details->body_size, char);
1272 new_body = ((char *)new_body) - new_type_details->offset;
1274 /* We always allocated the full length item with PURIFY. To do this
1275 we fake things so that arena is false for all 16 types.. */
1276 new_body = new_NOARENAZ(new_type_details);
1278 SvANY(sv) = new_body;
1279 if (new_type == SVt_PVAV) {
1283 if (old_type_details->body_size) {
1286 /* It will have been zeroed when the new body was allocated.
1287 Lets not write to it, in case it confuses a write-back
1293 #ifndef NODEFAULT_SHAREKEYS
1294 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1296 HvMAX(sv) = 7; /* (start with 8 buckets) */
1299 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1300 The target created by newSVrv also is, and it can have magic.
1301 However, it never has SvPVX set.
1303 if (old_type == SVt_IV) {
1305 } else if (old_type >= SVt_PV) {
1306 assert(SvPVX_const(sv) == 0);
1309 if (old_type >= SVt_PVMG) {
1310 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1311 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1313 sv->sv_u.svu_array = NULL; /* or svu_hash */
1319 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1320 sv_force_normal_flags(sv) is called. */
1323 /* XXX Is this still needed? Was it ever needed? Surely as there is
1324 no route from NV to PVIV, NOK can never be true */
1325 assert(!SvNOKp(sv));
1336 assert(new_type_details->body_size);
1337 /* We always allocated the full length item with PURIFY. To do this
1338 we fake things so that arena is false for all 16 types.. */
1339 if(new_type_details->arena) {
1340 /* This points to the start of the allocated area. */
1341 new_body_inline(new_body, new_type);
1342 Zero(new_body, new_type_details->body_size, char);
1343 new_body = ((char *)new_body) - new_type_details->offset;
1345 new_body = new_NOARENAZ(new_type_details);
1347 SvANY(sv) = new_body;
1349 if (old_type_details->copy) {
1350 /* There is now the potential for an upgrade from something without
1351 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1352 int offset = old_type_details->offset;
1353 int length = old_type_details->copy;
1355 if (new_type_details->offset > old_type_details->offset) {
1356 const int difference
1357 = new_type_details->offset - old_type_details->offset;
1358 offset += difference;
1359 length -= difference;
1361 assert (length >= 0);
1363 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1367 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1368 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1369 * correct 0.0 for us. Otherwise, if the old body didn't have an
1370 * NV slot, but the new one does, then we need to initialise the
1371 * freshly created NV slot with whatever the correct bit pattern is
1373 if (old_type_details->zero_nv && !new_type_details->zero_nv
1374 && !isGV_with_GP(sv))
1378 if (new_type == SVt_PVIO) {
1379 IO * const io = MUTABLE_IO(sv);
1380 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1383 /* Clear the stashcache because a new IO could overrule a package
1385 hv_clear(PL_stashcache);
1387 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1388 IoPAGE_LEN(sv) = 60;
1390 if (old_type < SVt_PV) {
1391 /* referant will be NULL unless the old type was SVt_IV emulating
1393 sv->sv_u.svu_rv = referant;
1397 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1398 (unsigned long)new_type);
1401 if (old_type > SVt_IV) {
1405 /* Note that there is an assumption that all bodies of types that
1406 can be upgraded came from arenas. Only the more complex non-
1407 upgradable types are allowed to be directly malloc()ed. */
1408 assert(old_type_details->arena);
1409 del_body((void*)((char*)old_body + old_type_details->offset),
1410 &PL_body_roots[old_type]);
1416 =for apidoc sv_backoff
1418 Remove any string offset. You should normally use the C<SvOOK_off> macro
1425 Perl_sv_backoff(pTHX_ register SV *const sv)
1428 const char * const s = SvPVX_const(sv);
1430 PERL_ARGS_ASSERT_SV_BACKOFF;
1431 PERL_UNUSED_CONTEXT;
1434 assert(SvTYPE(sv) != SVt_PVHV);
1435 assert(SvTYPE(sv) != SVt_PVAV);
1437 SvOOK_offset(sv, delta);
1439 SvLEN_set(sv, SvLEN(sv) + delta);
1440 SvPV_set(sv, SvPVX(sv) - delta);
1441 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1442 SvFLAGS(sv) &= ~SVf_OOK;
1449 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1450 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1451 Use the C<SvGROW> wrapper instead.
1457 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1461 PERL_ARGS_ASSERT_SV_GROW;
1463 if (PL_madskills && newlen >= 0x100000) {
1464 PerlIO_printf(Perl_debug_log,
1465 "Allocation too large: %"UVxf"\n", (UV)newlen);
1467 #ifdef HAS_64K_LIMIT
1468 if (newlen >= 0x10000) {
1469 PerlIO_printf(Perl_debug_log,
1470 "Allocation too large: %"UVxf"\n", (UV)newlen);
1473 #endif /* HAS_64K_LIMIT */
1476 if (SvTYPE(sv) < SVt_PV) {
1477 sv_upgrade(sv, SVt_PV);
1478 s = SvPVX_mutable(sv);
1480 else if (SvOOK(sv)) { /* pv is offset? */
1482 s = SvPVX_mutable(sv);
1483 if (newlen > SvLEN(sv))
1484 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1485 #ifdef HAS_64K_LIMIT
1486 if (newlen >= 0x10000)
1491 s = SvPVX_mutable(sv);
1493 if (newlen > SvLEN(sv)) { /* need more room? */
1494 STRLEN minlen = SvCUR(sv);
1495 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1496 if (newlen < minlen)
1498 #ifndef Perl_safesysmalloc_size
1499 newlen = PERL_STRLEN_ROUNDUP(newlen);
1501 if (SvLEN(sv) && s) {
1502 s = (char*)saferealloc(s, newlen);
1505 s = (char*)safemalloc(newlen);
1506 if (SvPVX_const(sv) && SvCUR(sv)) {
1507 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1511 #ifdef Perl_safesysmalloc_size
1512 /* Do this here, do it once, do it right, and then we will never get
1513 called back into sv_grow() unless there really is some growing
1515 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1517 SvLEN_set(sv, newlen);
1524 =for apidoc sv_setiv
1526 Copies an integer into the given SV, upgrading first if necessary.
1527 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1533 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1537 PERL_ARGS_ASSERT_SV_SETIV;
1539 SV_CHECK_THINKFIRST_COW_DROP(sv);
1540 switch (SvTYPE(sv)) {
1543 sv_upgrade(sv, SVt_IV);
1546 sv_upgrade(sv, SVt_PVIV);
1550 if (!isGV_with_GP(sv))
1557 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1561 (void)SvIOK_only(sv); /* validate number */
1567 =for apidoc sv_setiv_mg
1569 Like C<sv_setiv>, but also handles 'set' magic.
1575 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1577 PERL_ARGS_ASSERT_SV_SETIV_MG;
1584 =for apidoc sv_setuv
1586 Copies an unsigned integer into the given SV, upgrading first if necessary.
1587 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1593 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1595 PERL_ARGS_ASSERT_SV_SETUV;
1597 /* With these two if statements:
1598 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1601 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1603 If you wish to remove them, please benchmark to see what the effect is
1605 if (u <= (UV)IV_MAX) {
1606 sv_setiv(sv, (IV)u);
1615 =for apidoc sv_setuv_mg
1617 Like C<sv_setuv>, but also handles 'set' magic.
1623 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1625 PERL_ARGS_ASSERT_SV_SETUV_MG;
1632 =for apidoc sv_setnv
1634 Copies a double into the given SV, upgrading first if necessary.
1635 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1641 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1645 PERL_ARGS_ASSERT_SV_SETNV;
1647 SV_CHECK_THINKFIRST_COW_DROP(sv);
1648 switch (SvTYPE(sv)) {
1651 sv_upgrade(sv, SVt_NV);
1655 sv_upgrade(sv, SVt_PVNV);
1659 if (!isGV_with_GP(sv))
1666 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1671 (void)SvNOK_only(sv); /* validate number */
1676 =for apidoc sv_setnv_mg
1678 Like C<sv_setnv>, but also handles 'set' magic.
1684 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1686 PERL_ARGS_ASSERT_SV_SETNV_MG;
1692 /* Print an "isn't numeric" warning, using a cleaned-up,
1693 * printable version of the offending string
1697 S_not_a_number(pTHX_ SV *const sv)
1704 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1707 dsv = newSVpvs_flags("", SVs_TEMP);
1708 pv = sv_uni_display(dsv, sv, 10, 0);
1711 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1712 /* each *s can expand to 4 chars + "...\0",
1713 i.e. need room for 8 chars */
1715 const char *s = SvPVX_const(sv);
1716 const char * const end = s + SvCUR(sv);
1717 for ( ; s < end && d < limit; s++ ) {
1719 if (ch & 128 && !isPRINT_LC(ch)) {
1728 else if (ch == '\r') {
1732 else if (ch == '\f') {
1736 else if (ch == '\\') {
1740 else if (ch == '\0') {
1744 else if (isPRINT_LC(ch))
1761 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1762 "Argument \"%s\" isn't numeric in %s", pv,
1765 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1766 "Argument \"%s\" isn't numeric", pv);
1770 =for apidoc looks_like_number
1772 Test if the content of an SV looks like a number (or is a number).
1773 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1774 non-numeric warning), even if your atof() doesn't grok them.
1780 Perl_looks_like_number(pTHX_ SV *const sv)
1782 register const char *sbegin;
1785 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1788 sbegin = SvPVX_const(sv);
1791 else if (SvPOKp(sv))
1792 sbegin = SvPV_const(sv, len);
1794 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1795 return grok_number(sbegin, len, NULL);
1799 S_glob_2number(pTHX_ GV * const gv)
1801 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1802 SV *const buffer = sv_newmortal();
1804 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1806 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1809 gv_efullname3(buffer, gv, "*");
1810 SvFLAGS(gv) |= wasfake;
1812 /* We know that all GVs stringify to something that is not-a-number,
1813 so no need to test that. */
1814 if (ckWARN(WARN_NUMERIC))
1815 not_a_number(buffer);
1816 /* We just want something true to return, so that S_sv_2iuv_common
1817 can tail call us and return true. */
1821 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1822 until proven guilty, assume that things are not that bad... */
1827 As 64 bit platforms often have an NV that doesn't preserve all bits of
1828 an IV (an assumption perl has been based on to date) it becomes necessary
1829 to remove the assumption that the NV always carries enough precision to
1830 recreate the IV whenever needed, and that the NV is the canonical form.
1831 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1832 precision as a side effect of conversion (which would lead to insanity
1833 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1834 1) to distinguish between IV/UV/NV slots that have cached a valid
1835 conversion where precision was lost and IV/UV/NV slots that have a
1836 valid conversion which has lost no precision
1837 2) to ensure that if a numeric conversion to one form is requested that
1838 would lose precision, the precise conversion (or differently
1839 imprecise conversion) is also performed and cached, to prevent
1840 requests for different numeric formats on the same SV causing
1841 lossy conversion chains. (lossless conversion chains are perfectly
1846 SvIOKp is true if the IV slot contains a valid value
1847 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1848 SvNOKp is true if the NV slot contains a valid value
1849 SvNOK is true only if the NV value is accurate
1852 while converting from PV to NV, check to see if converting that NV to an
1853 IV(or UV) would lose accuracy over a direct conversion from PV to
1854 IV(or UV). If it would, cache both conversions, return NV, but mark
1855 SV as IOK NOKp (ie not NOK).
1857 While converting from PV to IV, check to see if converting that IV to an
1858 NV would lose accuracy over a direct conversion from PV to NV. If it
1859 would, cache both conversions, flag similarly.
1861 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1862 correctly because if IV & NV were set NV *always* overruled.
1863 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1864 changes - now IV and NV together means that the two are interchangeable:
1865 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1867 The benefit of this is that operations such as pp_add know that if
1868 SvIOK is true for both left and right operands, then integer addition
1869 can be used instead of floating point (for cases where the result won't
1870 overflow). Before, floating point was always used, which could lead to
1871 loss of precision compared with integer addition.
1873 * making IV and NV equal status should make maths accurate on 64 bit
1875 * may speed up maths somewhat if pp_add and friends start to use
1876 integers when possible instead of fp. (Hopefully the overhead in
1877 looking for SvIOK and checking for overflow will not outweigh the
1878 fp to integer speedup)
1879 * will slow down integer operations (callers of SvIV) on "inaccurate"
1880 values, as the change from SvIOK to SvIOKp will cause a call into
1881 sv_2iv each time rather than a macro access direct to the IV slot
1882 * should speed up number->string conversion on integers as IV is
1883 favoured when IV and NV are equally accurate
1885 ####################################################################
1886 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1887 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1888 On the other hand, SvUOK is true iff UV.
1889 ####################################################################
1891 Your mileage will vary depending your CPU's relative fp to integer
1895 #ifndef NV_PRESERVES_UV
1896 # define IS_NUMBER_UNDERFLOW_IV 1
1897 # define IS_NUMBER_UNDERFLOW_UV 2
1898 # define IS_NUMBER_IV_AND_UV 2
1899 # define IS_NUMBER_OVERFLOW_IV 4
1900 # define IS_NUMBER_OVERFLOW_UV 5
1902 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1904 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1906 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1914 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1916 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));
1917 if (SvNVX(sv) < (NV)IV_MIN) {
1918 (void)SvIOKp_on(sv);
1920 SvIV_set(sv, IV_MIN);
1921 return IS_NUMBER_UNDERFLOW_IV;
1923 if (SvNVX(sv) > (NV)UV_MAX) {
1924 (void)SvIOKp_on(sv);
1927 SvUV_set(sv, UV_MAX);
1928 return IS_NUMBER_OVERFLOW_UV;
1930 (void)SvIOKp_on(sv);
1932 /* Can't use strtol etc to convert this string. (See truth table in
1934 if (SvNVX(sv) <= (UV)IV_MAX) {
1935 SvIV_set(sv, I_V(SvNVX(sv)));
1936 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1937 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1939 /* Integer is imprecise. NOK, IOKp */
1941 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1944 SvUV_set(sv, U_V(SvNVX(sv)));
1945 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1946 if (SvUVX(sv) == UV_MAX) {
1947 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1948 possibly be preserved by NV. Hence, it must be overflow.
1950 return IS_NUMBER_OVERFLOW_UV;
1952 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1954 /* Integer is imprecise. NOK, IOKp */
1956 return IS_NUMBER_OVERFLOW_IV;
1958 #endif /* !NV_PRESERVES_UV*/
1961 S_sv_2iuv_common(pTHX_ SV *const sv)
1965 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1968 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1969 * without also getting a cached IV/UV from it at the same time
1970 * (ie PV->NV conversion should detect loss of accuracy and cache
1971 * IV or UV at same time to avoid this. */
1972 /* IV-over-UV optimisation - choose to cache IV if possible */
1974 if (SvTYPE(sv) == SVt_NV)
1975 sv_upgrade(sv, SVt_PVNV);
1977 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1978 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1979 certainly cast into the IV range at IV_MAX, whereas the correct
1980 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1982 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1983 if (Perl_isnan(SvNVX(sv))) {
1989 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
1990 SvIV_set(sv, I_V(SvNVX(sv)));
1991 if (SvNVX(sv) == (NV) SvIVX(sv)
1992 #ifndef NV_PRESERVES_UV
1993 && (((UV)1 << NV_PRESERVES_UV_BITS) >
1994 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
1995 /* Don't flag it as "accurately an integer" if the number
1996 came from a (by definition imprecise) NV operation, and
1997 we're outside the range of NV integer precision */
2001 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2003 /* scalar has trailing garbage, eg "42a" */
2005 DEBUG_c(PerlIO_printf(Perl_debug_log,
2006 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2012 /* IV not precise. No need to convert from PV, as NV
2013 conversion would already have cached IV if it detected
2014 that PV->IV would be better than PV->NV->IV
2015 flags already correct - don't set public IOK. */
2016 DEBUG_c(PerlIO_printf(Perl_debug_log,
2017 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2022 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2023 but the cast (NV)IV_MIN rounds to a the value less (more
2024 negative) than IV_MIN which happens to be equal to SvNVX ??
2025 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2026 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2027 (NV)UVX == NVX are both true, but the values differ. :-(
2028 Hopefully for 2s complement IV_MIN is something like
2029 0x8000000000000000 which will be exact. NWC */
2032 SvUV_set(sv, U_V(SvNVX(sv)));
2034 (SvNVX(sv) == (NV) SvUVX(sv))
2035 #ifndef NV_PRESERVES_UV
2036 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2037 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2038 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2039 /* Don't flag it as "accurately an integer" if the number
2040 came from a (by definition imprecise) NV operation, and
2041 we're outside the range of NV integer precision */
2047 DEBUG_c(PerlIO_printf(Perl_debug_log,
2048 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2054 else if (SvPOKp(sv) && SvLEN(sv)) {
2056 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2057 /* We want to avoid a possible problem when we cache an IV/ a UV which
2058 may be later translated to an NV, and the resulting NV is not
2059 the same as the direct translation of the initial string
2060 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2061 be careful to ensure that the value with the .456 is around if the
2062 NV value is requested in the future).
2064 This means that if we cache such an IV/a UV, we need to cache the
2065 NV as well. Moreover, we trade speed for space, and do not
2066 cache the NV if we are sure it's not needed.
2069 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2070 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2071 == IS_NUMBER_IN_UV) {
2072 /* It's definitely an integer, only upgrade to PVIV */
2073 if (SvTYPE(sv) < SVt_PVIV)
2074 sv_upgrade(sv, SVt_PVIV);
2076 } else if (SvTYPE(sv) < SVt_PVNV)
2077 sv_upgrade(sv, SVt_PVNV);
2079 /* If NVs preserve UVs then we only use the UV value if we know that
2080 we aren't going to call atof() below. If NVs don't preserve UVs
2081 then the value returned may have more precision than atof() will
2082 return, even though value isn't perfectly accurate. */
2083 if ((numtype & (IS_NUMBER_IN_UV
2084 #ifdef NV_PRESERVES_UV
2087 )) == IS_NUMBER_IN_UV) {
2088 /* This won't turn off the public IOK flag if it was set above */
2089 (void)SvIOKp_on(sv);
2091 if (!(numtype & IS_NUMBER_NEG)) {
2093 if (value <= (UV)IV_MAX) {
2094 SvIV_set(sv, (IV)value);
2096 /* it didn't overflow, and it was positive. */
2097 SvUV_set(sv, value);
2101 /* 2s complement assumption */
2102 if (value <= (UV)IV_MIN) {
2103 SvIV_set(sv, -(IV)value);
2105 /* Too negative for an IV. This is a double upgrade, but
2106 I'm assuming it will be rare. */
2107 if (SvTYPE(sv) < SVt_PVNV)
2108 sv_upgrade(sv, SVt_PVNV);
2112 SvNV_set(sv, -(NV)value);
2113 SvIV_set(sv, IV_MIN);
2117 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2118 will be in the previous block to set the IV slot, and the next
2119 block to set the NV slot. So no else here. */
2121 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2122 != IS_NUMBER_IN_UV) {
2123 /* It wasn't an (integer that doesn't overflow the UV). */
2124 SvNV_set(sv, Atof(SvPVX_const(sv)));
2126 if (! numtype && ckWARN(WARN_NUMERIC))
2129 #if defined(USE_LONG_DOUBLE)
2130 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2131 PTR2UV(sv), SvNVX(sv)));
2133 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2134 PTR2UV(sv), SvNVX(sv)));
2137 #ifdef NV_PRESERVES_UV
2138 (void)SvIOKp_on(sv);
2140 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2141 SvIV_set(sv, I_V(SvNVX(sv)));
2142 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2145 NOOP; /* Integer is imprecise. NOK, IOKp */
2147 /* UV will not work better than IV */
2149 if (SvNVX(sv) > (NV)UV_MAX) {
2151 /* Integer is inaccurate. NOK, IOKp, is UV */
2152 SvUV_set(sv, UV_MAX);
2154 SvUV_set(sv, U_V(SvNVX(sv)));
2155 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2156 NV preservse UV so can do correct comparison. */
2157 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2160 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2165 #else /* NV_PRESERVES_UV */
2166 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2167 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2168 /* The IV/UV slot will have been set from value returned by
2169 grok_number above. The NV slot has just been set using
2172 assert (SvIOKp(sv));
2174 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2175 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2176 /* Small enough to preserve all bits. */
2177 (void)SvIOKp_on(sv);
2179 SvIV_set(sv, I_V(SvNVX(sv)));
2180 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2182 /* Assumption: first non-preserved integer is < IV_MAX,
2183 this NV is in the preserved range, therefore: */
2184 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2186 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);
2190 0 0 already failed to read UV.
2191 0 1 already failed to read UV.
2192 1 0 you won't get here in this case. IV/UV
2193 slot set, public IOK, Atof() unneeded.
2194 1 1 already read UV.
2195 so there's no point in sv_2iuv_non_preserve() attempting
2196 to use atol, strtol, strtoul etc. */
2198 sv_2iuv_non_preserve (sv, numtype);
2200 sv_2iuv_non_preserve (sv);
2204 #endif /* NV_PRESERVES_UV */
2205 /* It might be more code efficient to go through the entire logic above
2206 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2207 gets complex and potentially buggy, so more programmer efficient
2208 to do it this way, by turning off the public flags: */
2210 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2214 if (isGV_with_GP(sv))
2215 return glob_2number(MUTABLE_GV(sv));
2217 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2218 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2221 if (SvTYPE(sv) < SVt_IV)
2222 /* Typically the caller expects that sv_any is not NULL now. */
2223 sv_upgrade(sv, SVt_IV);
2224 /* Return 0 from the caller. */
2231 =for apidoc sv_2iv_flags
2233 Return the integer value of an SV, doing any necessary string
2234 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2235 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2241 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2246 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2247 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2248 cache IVs just in case. In practice it seems that they never
2249 actually anywhere accessible by user Perl code, let alone get used
2250 in anything other than a string context. */
2251 if (flags & SV_GMAGIC)
2256 return I_V(SvNVX(sv));
2258 if (SvPOKp(sv) && SvLEN(sv)) {
2261 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2263 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2264 == IS_NUMBER_IN_UV) {
2265 /* It's definitely an integer */
2266 if (numtype & IS_NUMBER_NEG) {
2267 if (value < (UV)IV_MIN)
2270 if (value < (UV)IV_MAX)
2275 if (ckWARN(WARN_NUMERIC))
2278 return I_V(Atof(SvPVX_const(sv)));
2283 assert(SvTYPE(sv) >= SVt_PVMG);
2284 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2285 } else if (SvTHINKFIRST(sv)) {
2290 if (flags & SV_SKIP_OVERLOAD)
2292 tmpstr=AMG_CALLun(sv,numer);
2293 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2294 return SvIV(tmpstr);
2297 return PTR2IV(SvRV(sv));
2300 sv_force_normal_flags(sv, 0);
2302 if (SvREADONLY(sv) && !SvOK(sv)) {
2303 if (ckWARN(WARN_UNINITIALIZED))
2309 if (S_sv_2iuv_common(aTHX_ sv))
2312 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2313 PTR2UV(sv),SvIVX(sv)));
2314 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2318 =for apidoc sv_2uv_flags
2320 Return the unsigned integer value of an SV, doing any necessary string
2321 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2322 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2328 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2333 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2334 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2335 cache IVs just in case. */
2336 if (flags & SV_GMAGIC)
2341 return U_V(SvNVX(sv));
2342 if (SvPOKp(sv) && SvLEN(sv)) {
2345 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2347 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2348 == IS_NUMBER_IN_UV) {
2349 /* It's definitely an integer */
2350 if (!(numtype & IS_NUMBER_NEG))
2354 if (ckWARN(WARN_NUMERIC))
2357 return U_V(Atof(SvPVX_const(sv)));
2362 assert(SvTYPE(sv) >= SVt_PVMG);
2363 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2364 } else if (SvTHINKFIRST(sv)) {
2369 if (flags & SV_SKIP_OVERLOAD)
2371 tmpstr = AMG_CALLun(sv,numer);
2372 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2373 return SvUV(tmpstr);
2376 return PTR2UV(SvRV(sv));
2379 sv_force_normal_flags(sv, 0);
2381 if (SvREADONLY(sv) && !SvOK(sv)) {
2382 if (ckWARN(WARN_UNINITIALIZED))
2388 if (S_sv_2iuv_common(aTHX_ sv))
2392 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2393 PTR2UV(sv),SvUVX(sv)));
2394 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2398 =for apidoc sv_2nv_flags
2400 Return the num value of an SV, doing any necessary string or integer
2401 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2402 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2408 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2413 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2414 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2415 cache IVs just in case. */
2416 if (flags & SV_GMAGIC)
2420 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2421 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2422 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2424 return Atof(SvPVX_const(sv));
2428 return (NV)SvUVX(sv);
2430 return (NV)SvIVX(sv);
2435 assert(SvTYPE(sv) >= SVt_PVMG);
2436 /* This falls through to the report_uninit near the end of the
2438 } else if (SvTHINKFIRST(sv)) {
2443 if (flags & SV_SKIP_OVERLOAD)
2445 tmpstr = AMG_CALLun(sv,numer);
2446 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2447 return SvNV(tmpstr);
2450 return PTR2NV(SvRV(sv));
2453 sv_force_normal_flags(sv, 0);
2455 if (SvREADONLY(sv) && !SvOK(sv)) {
2456 if (ckWARN(WARN_UNINITIALIZED))
2461 if (SvTYPE(sv) < SVt_NV) {
2462 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2463 sv_upgrade(sv, SVt_NV);
2464 #ifdef USE_LONG_DOUBLE
2466 STORE_NUMERIC_LOCAL_SET_STANDARD();
2467 PerlIO_printf(Perl_debug_log,
2468 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2469 PTR2UV(sv), SvNVX(sv));
2470 RESTORE_NUMERIC_LOCAL();
2474 STORE_NUMERIC_LOCAL_SET_STANDARD();
2475 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2476 PTR2UV(sv), SvNVX(sv));
2477 RESTORE_NUMERIC_LOCAL();
2481 else if (SvTYPE(sv) < SVt_PVNV)
2482 sv_upgrade(sv, SVt_PVNV);
2487 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2488 #ifdef NV_PRESERVES_UV
2494 /* Only set the public NV OK flag if this NV preserves the IV */
2495 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2497 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2498 : (SvIVX(sv) == I_V(SvNVX(sv))))
2504 else if (SvPOKp(sv) && SvLEN(sv)) {
2506 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2507 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2509 #ifdef NV_PRESERVES_UV
2510 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2511 == IS_NUMBER_IN_UV) {
2512 /* It's definitely an integer */
2513 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2515 SvNV_set(sv, Atof(SvPVX_const(sv)));
2521 SvNV_set(sv, Atof(SvPVX_const(sv)));
2522 /* Only set the public NV OK flag if this NV preserves the value in
2523 the PV at least as well as an IV/UV would.
2524 Not sure how to do this 100% reliably. */
2525 /* if that shift count is out of range then Configure's test is
2526 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2528 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2529 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2530 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2531 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2532 /* Can't use strtol etc to convert this string, so don't try.
2533 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2536 /* value has been set. It may not be precise. */
2537 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2538 /* 2s complement assumption for (UV)IV_MIN */
2539 SvNOK_on(sv); /* Integer is too negative. */
2544 if (numtype & IS_NUMBER_NEG) {
2545 SvIV_set(sv, -(IV)value);
2546 } else if (value <= (UV)IV_MAX) {
2547 SvIV_set(sv, (IV)value);
2549 SvUV_set(sv, value);
2553 if (numtype & IS_NUMBER_NOT_INT) {
2554 /* I believe that even if the original PV had decimals,
2555 they are lost beyond the limit of the FP precision.
2556 However, neither is canonical, so both only get p
2557 flags. NWC, 2000/11/25 */
2558 /* Both already have p flags, so do nothing */
2560 const NV nv = SvNVX(sv);
2561 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2562 if (SvIVX(sv) == I_V(nv)) {
2565 /* It had no "." so it must be integer. */
2569 /* between IV_MAX and NV(UV_MAX).
2570 Could be slightly > UV_MAX */
2572 if (numtype & IS_NUMBER_NOT_INT) {
2573 /* UV and NV both imprecise. */
2575 const UV nv_as_uv = U_V(nv);
2577 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2586 /* It might be more code efficient to go through the entire logic above
2587 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2588 gets complex and potentially buggy, so more programmer efficient
2589 to do it this way, by turning off the public flags: */
2591 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2592 #endif /* NV_PRESERVES_UV */
2595 if (isGV_with_GP(sv)) {
2596 glob_2number(MUTABLE_GV(sv));
2600 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2602 assert (SvTYPE(sv) >= SVt_NV);
2603 /* Typically the caller expects that sv_any is not NULL now. */
2604 /* XXX Ilya implies that this is a bug in callers that assume this
2605 and ideally should be fixed. */
2608 #if defined(USE_LONG_DOUBLE)
2610 STORE_NUMERIC_LOCAL_SET_STANDARD();
2611 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2612 PTR2UV(sv), SvNVX(sv));
2613 RESTORE_NUMERIC_LOCAL();
2617 STORE_NUMERIC_LOCAL_SET_STANDARD();
2618 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2619 PTR2UV(sv), SvNVX(sv));
2620 RESTORE_NUMERIC_LOCAL();
2629 Return an SV with the numeric value of the source SV, doing any necessary
2630 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2631 access this function.
2637 Perl_sv_2num(pTHX_ register SV *const sv)
2639 PERL_ARGS_ASSERT_SV_2NUM;
2644 SV * const tmpsv = AMG_CALLun(sv,numer);
2645 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2646 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2647 return sv_2num(tmpsv);
2649 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2652 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2653 * UV as a string towards the end of buf, and return pointers to start and
2656 * We assume that buf is at least TYPE_CHARS(UV) long.
2660 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2662 char *ptr = buf + TYPE_CHARS(UV);
2663 char * const ebuf = ptr;
2666 PERL_ARGS_ASSERT_UIV_2BUF;
2678 *--ptr = '0' + (char)(uv % 10);
2687 =for apidoc sv_2pv_flags
2689 Returns a pointer to the string value of an SV, and sets *lp to its length.
2690 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2692 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2693 usually end up here too.
2699 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2709 if (SvGMAGICAL(sv)) {
2710 if (flags & SV_GMAGIC)
2715 if (flags & SV_MUTABLE_RETURN)
2716 return SvPVX_mutable(sv);
2717 if (flags & SV_CONST_RETURN)
2718 return (char *)SvPVX_const(sv);
2721 if (SvIOKp(sv) || SvNOKp(sv)) {
2722 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2727 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2728 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2730 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2737 #ifdef FIXNEGATIVEZERO
2738 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2744 SvUPGRADE(sv, SVt_PV);
2747 s = SvGROW_mutable(sv, len + 1);
2750 return (char*)memcpy(s, tbuf, len + 1);
2756 assert(SvTYPE(sv) >= SVt_PVMG);
2757 /* This falls through to the report_uninit near the end of the
2759 } else if (SvTHINKFIRST(sv)) {
2764 if (flags & SV_SKIP_OVERLOAD)
2766 tmpstr = AMG_CALLun(sv,string);
2767 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2768 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2770 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2774 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2775 if (flags & SV_CONST_RETURN) {
2776 pv = (char *) SvPVX_const(tmpstr);
2778 pv = (flags & SV_MUTABLE_RETURN)
2779 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2782 *lp = SvCUR(tmpstr);
2784 pv = sv_2pv_flags(tmpstr, lp, flags);
2797 SV *const referent = SvRV(sv);
2801 retval = buffer = savepvn("NULLREF", len);
2802 } else if (SvTYPE(referent) == SVt_REGEXP) {
2803 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2808 /* If the regex is UTF-8 we want the containing scalar to
2809 have an UTF-8 flag too */
2815 if ((seen_evals = RX_SEEN_EVALS(re)))
2816 PL_reginterp_cnt += seen_evals;
2819 *lp = RX_WRAPLEN(re);
2821 return RX_WRAPPED(re);
2823 const char *const typestr = sv_reftype(referent, 0);
2824 const STRLEN typelen = strlen(typestr);
2825 UV addr = PTR2UV(referent);
2826 const char *stashname = NULL;
2827 STRLEN stashnamelen = 0; /* hush, gcc */
2828 const char *buffer_end;
2830 if (SvOBJECT(referent)) {
2831 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2834 stashname = HEK_KEY(name);
2835 stashnamelen = HEK_LEN(name);
2837 if (HEK_UTF8(name)) {
2843 stashname = "__ANON__";
2846 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2847 + 2 * sizeof(UV) + 2 /* )\0 */;
2849 len = typelen + 3 /* (0x */
2850 + 2 * sizeof(UV) + 2 /* )\0 */;
2853 Newx(buffer, len, char);
2854 buffer_end = retval = buffer + len;
2856 /* Working backwards */
2860 *--retval = PL_hexdigit[addr & 15];
2861 } while (addr >>= 4);
2867 memcpy(retval, typestr, typelen);
2871 retval -= stashnamelen;
2872 memcpy(retval, stashname, stashnamelen);
2874 /* retval may not neccesarily have reached the start of the
2876 assert (retval >= buffer);
2878 len = buffer_end - retval - 1; /* -1 for that \0 */
2886 if (SvREADONLY(sv) && !SvOK(sv)) {
2889 if (flags & SV_UNDEF_RETURNS_NULL)
2891 if (ckWARN(WARN_UNINITIALIZED))
2896 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2897 /* I'm assuming that if both IV and NV are equally valid then
2898 converting the IV is going to be more efficient */
2899 const U32 isUIOK = SvIsUV(sv);
2900 char buf[TYPE_CHARS(UV)];
2904 if (SvTYPE(sv) < SVt_PVIV)
2905 sv_upgrade(sv, SVt_PVIV);
2906 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2908 /* inlined from sv_setpvn */
2909 s = SvGROW_mutable(sv, len + 1);
2910 Move(ptr, s, len, char);
2914 else if (SvNOKp(sv)) {
2916 if (SvTYPE(sv) < SVt_PVNV)
2917 sv_upgrade(sv, SVt_PVNV);
2918 /* The +20 is pure guesswork. Configure test needed. --jhi */
2919 s = SvGROW_mutable(sv, NV_DIG + 20);
2920 /* some Xenix systems wipe out errno here */
2922 if (SvNVX(sv) == 0.0)
2923 my_strlcpy(s, "0", SvLEN(sv));
2927 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2930 #ifdef FIXNEGATIVEZERO
2931 if (*s == '-' && s[1] == '0' && !s[2]) {
2943 if (isGV_with_GP(sv)) {
2944 GV *const gv = MUTABLE_GV(sv);
2945 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2946 SV *const buffer = sv_newmortal();
2948 /* FAKE globs can get coerced, so need to turn this off temporarily
2951 gv_efullname3(buffer, gv, "*");
2952 SvFLAGS(gv) |= wasfake;
2954 if (SvPOK(buffer)) {
2956 *lp = SvCUR(buffer);
2958 return SvPVX(buffer);
2969 if (flags & SV_UNDEF_RETURNS_NULL)
2971 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2973 if (SvTYPE(sv) < SVt_PV)
2974 /* Typically the caller expects that sv_any is not NULL now. */
2975 sv_upgrade(sv, SVt_PV);
2979 const STRLEN len = s - SvPVX_const(sv);
2985 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2986 PTR2UV(sv),SvPVX_const(sv)));
2987 if (flags & SV_CONST_RETURN)
2988 return (char *)SvPVX_const(sv);
2989 if (flags & SV_MUTABLE_RETURN)
2990 return SvPVX_mutable(sv);
2995 =for apidoc sv_copypv
2997 Copies a stringified representation of the source SV into the
2998 destination SV. Automatically performs any necessary mg_get and
2999 coercion of numeric values into strings. Guaranteed to preserve
3000 UTF8 flag even from overloaded objects. Similar in nature to
3001 sv_2pv[_flags] but operates directly on an SV instead of just the
3002 string. Mostly uses sv_2pv_flags to do its work, except when that
3003 would lose the UTF-8'ness of the PV.
3009 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3012 const char * const s = SvPV_const(ssv,len);
3014 PERL_ARGS_ASSERT_SV_COPYPV;
3016 sv_setpvn(dsv,s,len);
3024 =for apidoc sv_2pvbyte
3026 Return a pointer to the byte-encoded representation of the SV, and set *lp
3027 to its length. May cause the SV to be downgraded from UTF-8 as a
3030 Usually accessed via the C<SvPVbyte> macro.
3036 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3038 PERL_ARGS_ASSERT_SV_2PVBYTE;
3040 sv_utf8_downgrade(sv,0);
3041 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3045 =for apidoc sv_2pvutf8
3047 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3048 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3050 Usually accessed via the C<SvPVutf8> macro.
3056 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3058 PERL_ARGS_ASSERT_SV_2PVUTF8;
3060 sv_utf8_upgrade(sv);
3061 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3066 =for apidoc sv_2bool
3068 This function is only called on magical items, and is only used by
3069 sv_true() or its macro equivalent.
3075 Perl_sv_2bool(pTHX_ register SV *const sv)
3079 PERL_ARGS_ASSERT_SV_2BOOL;
3087 SV * const tmpsv = AMG_CALLun(sv,bool_);
3088 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3089 return cBOOL(SvTRUE(tmpsv));
3091 return SvRV(sv) != 0;
3094 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3096 (*sv->sv_u.svu_pv > '0' ||
3097 Xpvtmp->xpv_cur > 1 ||
3098 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3105 return SvIVX(sv) != 0;
3108 return SvNVX(sv) != 0.0;
3110 if (isGV_with_GP(sv))
3120 =for apidoc sv_utf8_upgrade
3122 Converts the PV of an SV to its UTF-8-encoded form.
3123 Forces the SV to string form if it is not already.
3124 Will C<mg_get> on C<sv> if appropriate.
3125 Always sets the SvUTF8 flag to avoid future validity checks even
3126 if the whole string is the same in UTF-8 as not.
3127 Returns the number of bytes in the converted string
3129 This is not as a general purpose byte encoding to Unicode interface:
3130 use the Encode extension for that.
3132 =for apidoc sv_utf8_upgrade_nomg
3134 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3136 =for apidoc sv_utf8_upgrade_flags
3138 Converts the PV of an SV to its UTF-8-encoded form.
3139 Forces the SV to string form if it is not already.
3140 Always sets the SvUTF8 flag to avoid future validity checks even
3141 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3142 will C<mg_get> on C<sv> if appropriate, else not.
3143 Returns the number of bytes in the converted string
3144 C<sv_utf8_upgrade> and
3145 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3147 This is not as a general purpose byte encoding to Unicode interface:
3148 use the Encode extension for that.
3152 The grow version is currently not externally documented. It adds a parameter,
3153 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3154 have free after it upon return. This allows the caller to reserve extra space
3155 that it intends to fill, to avoid extra grows.
3157 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3158 which can be used to tell this function to not first check to see if there are
3159 any characters that are different in UTF-8 (variant characters) which would
3160 force it to allocate a new string to sv, but to assume there are. Typically
3161 this flag is used by a routine that has already parsed the string to find that
3162 there are such characters, and passes this information on so that the work
3163 doesn't have to be repeated.
3165 (One might think that the calling routine could pass in the position of the
3166 first such variant, so it wouldn't have to be found again. But that is not the
3167 case, because typically when the caller is likely to use this flag, it won't be
3168 calling this routine unless it finds something that won't fit into a byte.
3169 Otherwise it tries to not upgrade and just use bytes. But some things that
3170 do fit into a byte are variants in utf8, and the caller may not have been
3171 keeping track of these.)
3173 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3174 isn't guaranteed due to having other routines do the work in some input cases,
3175 or if the input is already flagged as being in utf8.
3177 The speed of this could perhaps be improved for many cases if someone wanted to
3178 write a fast function that counts the number of variant characters in a string,
3179 especially if it could return the position of the first one.
3184 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3188 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3190 if (sv == &PL_sv_undef)
3194 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3195 (void) sv_2pv_flags(sv,&len, flags);
3197 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3201 (void) SvPV_force(sv,len);
3206 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3211 sv_force_normal_flags(sv, 0);
3214 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3215 sv_recode_to_utf8(sv, PL_encoding);
3216 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3220 if (SvCUR(sv) == 0) {
3221 if (extra) SvGROW(sv, extra);
3222 } else { /* Assume Latin-1/EBCDIC */
3223 /* This function could be much more efficient if we
3224 * had a FLAG in SVs to signal if there are any variant
3225 * chars in the PV. Given that there isn't such a flag
3226 * make the loop as fast as possible (although there are certainly ways
3227 * to speed this up, eg. through vectorization) */
3228 U8 * s = (U8 *) SvPVX_const(sv);
3229 U8 * e = (U8 *) SvEND(sv);
3231 STRLEN two_byte_count = 0;
3233 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3235 /* See if really will need to convert to utf8. We mustn't rely on our
3236 * incoming SV being well formed and having a trailing '\0', as certain
3237 * code in pp_formline can send us partially built SVs. */
3241 if (NATIVE_IS_INVARIANT(ch)) continue;
3243 t--; /* t already incremented; re-point to first variant */
3248 /* utf8 conversion not needed because all are invariants. Mark as
3249 * UTF-8 even if no variant - saves scanning loop */
3255 /* Here, the string should be converted to utf8, either because of an
3256 * input flag (two_byte_count = 0), or because a character that
3257 * requires 2 bytes was found (two_byte_count = 1). t points either to
3258 * the beginning of the string (if we didn't examine anything), or to
3259 * the first variant. In either case, everything from s to t - 1 will
3260 * occupy only 1 byte each on output.
3262 * There are two main ways to convert. One is to create a new string
3263 * and go through the input starting from the beginning, appending each
3264 * converted value onto the new string as we go along. It's probably
3265 * best to allocate enough space in the string for the worst possible
3266 * case rather than possibly running out of space and having to
3267 * reallocate and then copy what we've done so far. Since everything
3268 * from s to t - 1 is invariant, the destination can be initialized
3269 * with these using a fast memory copy
3271 * The other way is to figure out exactly how big the string should be
3272 * by parsing the entire input. Then you don't have to make it big
3273 * enough to handle the worst possible case, and more importantly, if
3274 * the string you already have is large enough, you don't have to
3275 * allocate a new string, you can copy the last character in the input
3276 * string to the final position(s) that will be occupied by the
3277 * converted string and go backwards, stopping at t, since everything
3278 * before that is invariant.
3280 * There are advantages and disadvantages to each method.
3282 * In the first method, we can allocate a new string, do the memory
3283 * copy from the s to t - 1, and then proceed through the rest of the
3284 * string byte-by-byte.
3286 * In the second method, we proceed through the rest of the input
3287 * string just calculating how big the converted string will be. Then
3288 * there are two cases:
3289 * 1) if the string has enough extra space to handle the converted
3290 * value. We go backwards through the string, converting until we
3291 * get to the position we are at now, and then stop. If this
3292 * position is far enough along in the string, this method is
3293 * faster than the other method. If the memory copy were the same
3294 * speed as the byte-by-byte loop, that position would be about
3295 * half-way, as at the half-way mark, parsing to the end and back
3296 * is one complete string's parse, the same amount as starting
3297 * over and going all the way through. Actually, it would be
3298 * somewhat less than half-way, as it's faster to just count bytes
3299 * than to also copy, and we don't have the overhead of allocating
3300 * a new string, changing the scalar to use it, and freeing the
3301 * existing one. But if the memory copy is fast, the break-even
3302 * point is somewhere after half way. The counting loop could be
3303 * sped up by vectorization, etc, to move the break-even point
3304 * further towards the beginning.
3305 * 2) if the string doesn't have enough space to handle the converted
3306 * value. A new string will have to be allocated, and one might
3307 * as well, given that, start from the beginning doing the first
3308 * method. We've spent extra time parsing the string and in
3309 * exchange all we've gotten is that we know precisely how big to
3310 * make the new one. Perl is more optimized for time than space,
3311 * so this case is a loser.
3312 * So what I've decided to do is not use the 2nd method unless it is
3313 * guaranteed that a new string won't have to be allocated, assuming
3314 * the worst case. I also decided not to put any more conditions on it
3315 * than this, for now. It seems likely that, since the worst case is
3316 * twice as big as the unknown portion of the string (plus 1), we won't
3317 * be guaranteed enough space, causing us to go to the first method,
3318 * unless the string is short, or the first variant character is near
3319 * the end of it. In either of these cases, it seems best to use the
3320 * 2nd method. The only circumstance I can think of where this would
3321 * be really slower is if the string had once had much more data in it
3322 * than it does now, but there is still a substantial amount in it */
3325 STRLEN invariant_head = t - s;
3326 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3327 if (SvLEN(sv) < size) {
3329 /* Here, have decided to allocate a new string */
3334 Newx(dst, size, U8);
3336 /* If no known invariants at the beginning of the input string,
3337 * set so starts from there. Otherwise, can use memory copy to
3338 * get up to where we are now, and then start from here */
3340 if (invariant_head <= 0) {
3343 Copy(s, dst, invariant_head, char);
3344 d = dst + invariant_head;
3348 const UV uv = NATIVE8_TO_UNI(*t++);
3349 if (UNI_IS_INVARIANT(uv))
3350 *d++ = (U8)UNI_TO_NATIVE(uv);
3352 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3353 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3357 SvPV_free(sv); /* No longer using pre-existing string */
3358 SvPV_set(sv, (char*)dst);
3359 SvCUR_set(sv, d - dst);
3360 SvLEN_set(sv, size);
3363 /* Here, have decided to get the exact size of the string.
3364 * Currently this happens only when we know that there is
3365 * guaranteed enough space to fit the converted string, so
3366 * don't have to worry about growing. If two_byte_count is 0,
3367 * then t points to the first byte of the string which hasn't
3368 * been examined yet. Otherwise two_byte_count is 1, and t
3369 * points to the first byte in the string that will expand to
3370 * two. Depending on this, start examining at t or 1 after t.
3373 U8 *d = t + two_byte_count;
3376 /* Count up the remaining bytes that expand to two */
3379 const U8 chr = *d++;
3380 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3383 /* The string will expand by just the number of bytes that
3384 * occupy two positions. But we are one afterwards because of
3385 * the increment just above. This is the place to put the
3386 * trailing NUL, and to set the length before we decrement */
3388 d += two_byte_count;
3389 SvCUR_set(sv, d - s);
3393 /* Having decremented d, it points to the position to put the
3394 * very last byte of the expanded string. Go backwards through
3395 * the string, copying and expanding as we go, stopping when we
3396 * get to the part that is invariant the rest of the way down */
3400 const U8 ch = NATIVE8_TO_UNI(*e--);
3401 if (UNI_IS_INVARIANT(ch)) {
3402 *d-- = UNI_TO_NATIVE(ch);
3404 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3405 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3412 /* Mark as UTF-8 even if no variant - saves scanning loop */
3418 =for apidoc sv_utf8_downgrade
3420 Attempts to convert the PV of an SV from characters to bytes.
3421 If the PV contains a character that cannot fit
3422 in a byte, this conversion will fail;
3423 in this case, either returns false or, if C<fail_ok> is not
3426 This is not as a general purpose Unicode to byte encoding interface:
3427 use the Encode extension for that.
3433 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3437 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3439 if (SvPOKp(sv) && SvUTF8(sv)) {
3445 sv_force_normal_flags(sv, 0);
3447 s = (U8 *) SvPV(sv, len);
3448 if (!utf8_to_bytes(s, &len)) {
3453 Perl_croak(aTHX_ "Wide character in %s",
3456 Perl_croak(aTHX_ "Wide character");
3467 =for apidoc sv_utf8_encode
3469 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3470 flag off so that it looks like octets again.
3476 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3478 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3481 sv_force_normal_flags(sv, 0);
3483 if (SvREADONLY(sv)) {
3484 Perl_croak_no_modify(aTHX);
3486 (void) sv_utf8_upgrade(sv);
3491 =for apidoc sv_utf8_decode
3493 If the PV of the SV is an octet sequence in UTF-8
3494 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3495 so that it looks like a character. If the PV contains only single-byte
3496 characters, the C<SvUTF8> flag stays being off.
3497 Scans PV for validity and returns false if the PV is invalid UTF-8.
3503 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3505 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3511 /* The octets may have got themselves encoded - get them back as
3514 if (!sv_utf8_downgrade(sv, TRUE))
3517 /* it is actually just a matter of turning the utf8 flag on, but
3518 * we want to make sure everything inside is valid utf8 first.
3520 c = (const U8 *) SvPVX_const(sv);
3521 if (!is_utf8_string(c, SvCUR(sv)+1))
3523 e = (const U8 *) SvEND(sv);
3526 if (!UTF8_IS_INVARIANT(ch)) {
3536 =for apidoc sv_setsv
3538 Copies the contents of the source SV C<ssv> into the destination SV
3539 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3540 function if the source SV needs to be reused. Does not handle 'set' magic.
3541 Loosely speaking, it performs a copy-by-value, obliterating any previous
3542 content of the destination.
3544 You probably want to use one of the assortment of wrappers, such as
3545 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3546 C<SvSetMagicSV_nosteal>.
3548 =for apidoc sv_setsv_flags
3550 Copies the contents of the source SV C<ssv> into the destination SV
3551 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3552 function if the source SV needs to be reused. Does not handle 'set' magic.
3553 Loosely speaking, it performs a copy-by-value, obliterating any previous
3554 content of the destination.
3555 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3556 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3557 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3558 and C<sv_setsv_nomg> are implemented in terms of this function.
3560 You probably want to use one of the assortment of wrappers, such as
3561 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3562 C<SvSetMagicSV_nosteal>.
3564 This is the primary function for copying scalars, and most other
3565 copy-ish functions and macros use this underneath.
3571 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3573 I32 mro_changes = 0; /* 1 = method, 2 = isa */
3575 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3577 if (dtype != SVt_PVGV) {
3578 const char * const name = GvNAME(sstr);
3579 const STRLEN len = GvNAMELEN(sstr);
3581 if (dtype >= SVt_PV) {
3587 SvUPGRADE(dstr, SVt_PVGV);
3588 (void)SvOK_off(dstr);
3589 /* FIXME - why are we doing this, then turning it off and on again
3591 isGV_with_GP_on(dstr);
3593 GvSTASH(dstr) = GvSTASH(sstr);
3595 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3596 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3597 SvFAKE_on(dstr); /* can coerce to non-glob */
3600 if(GvGP(MUTABLE_GV(sstr))) {
3601 /* If source has method cache entry, clear it */
3603 SvREFCNT_dec(GvCV(sstr));
3607 /* If source has a real method, then a method is
3609 else if(GvCV((const GV *)sstr)) {
3614 /* If dest already had a real method, that's a change as well */
3615 if(!mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)) {
3619 if(strEQ(GvNAME((const GV *)dstr),"ISA"))
3622 gp_free(MUTABLE_GV(dstr));
3623 isGV_with_GP_off(dstr);
3624 (void)SvOK_off(dstr);
3625 isGV_with_GP_on(dstr);
3626 GvINTRO_off(dstr); /* one-shot flag */
3627 GvGP(dstr) = gp_ref(GvGP(sstr));
3628 if (SvTAINTED(sstr))
3630 if (GvIMPORTED(dstr) != GVf_IMPORTED
3631 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3633 GvIMPORTED_on(dstr);
3636 if(mro_changes == 2) mro_isa_changed_in(GvSTASH(dstr));
3637 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3642 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3644 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3646 const int intro = GvINTRO(dstr);
3649 const U32 stype = SvTYPE(sref);
3651 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3654 GvINTRO_off(dstr); /* one-shot flag */
3655 GvLINE(dstr) = CopLINE(PL_curcop);
3656 GvEGV(dstr) = MUTABLE_GV(dstr);
3661 location = (SV **) &GvCV(dstr);
3662 import_flag = GVf_IMPORTED_CV;
3665 location = (SV **) &GvHV(dstr);
3666 import_flag = GVf_IMPORTED_HV;
3669 location = (SV **) &GvAV(dstr);
3670 import_flag = GVf_IMPORTED_AV;
3673 location = (SV **) &GvIOp(dstr);
3676 location = (SV **) &GvFORM(dstr);
3679 location = &GvSV(dstr);
3680 import_flag = GVf_IMPORTED_SV;
3683 if (stype == SVt_PVCV) {
3684 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3685 if (GvCVGEN(dstr)) {
3686 SvREFCNT_dec(GvCV(dstr));
3688 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3691 SAVEGENERICSV(*location);
3695 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3696 CV* const cv = MUTABLE_CV(*location);
3698 if (!GvCVGEN((const GV *)dstr) &&
3699 (CvROOT(cv) || CvXSUB(cv)))
3701 /* Redefining a sub - warning is mandatory if
3702 it was a const and its value changed. */
3703 if (CvCONST(cv) && CvCONST((const CV *)sref)
3705 == cv_const_sv((const CV *)sref)) {
3707 /* They are 2 constant subroutines generated from
3708 the same constant. This probably means that
3709 they are really the "same" proxy subroutine
3710 instantiated in 2 places. Most likely this is
3711 when a constant is exported twice. Don't warn.
3714 else if (ckWARN(WARN_REDEFINE)
3716 && (!CvCONST((const CV *)sref)
3717 || sv_cmp(cv_const_sv(cv),
3718 cv_const_sv((const CV *)
3720 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3723 ? "Constant subroutine %s::%s redefined"
3724 : "Subroutine %s::%s redefined"),
3725 HvNAME_get(GvSTASH((const GV *)dstr)),
3726 GvENAME(MUTABLE_GV(dstr)));
3730 cv_ckproto_len(cv, (const GV *)dstr,
3731 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3732 SvPOK(sref) ? SvCUR(sref) : 0);
3734 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3735 GvASSUMECV_on(dstr);
3736 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3739 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3740 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3741 GvFLAGS(dstr) |= import_flag;
3743 if (stype == SVt_PVAV && strEQ(GvNAME((GV*)dstr), "ISA")) {
3744 sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3745 mro_isa_changed_in(GvSTASH(dstr));
3750 if (SvTAINTED(sstr))
3756 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3759 register U32 sflags;
3761 register svtype stype;
3763 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3768 if (SvIS_FREED(dstr)) {
3769 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3770 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3772 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3774 sstr = &PL_sv_undef;
3775 if (SvIS_FREED(sstr)) {
3776 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3777 (void*)sstr, (void*)dstr);
3779 stype = SvTYPE(sstr);
3780 dtype = SvTYPE(dstr);
3782 (void)SvAMAGIC_off(dstr);
3785 /* need to nuke the magic */
3789 /* There's a lot of redundancy below but we're going for speed here */
3794 if (dtype != SVt_PVGV) {
3795 (void)SvOK_off(dstr);
3803 sv_upgrade(dstr, SVt_IV);
3807 sv_upgrade(dstr, SVt_PVIV);
3810 goto end_of_first_switch;
3812 (void)SvIOK_only(dstr);
3813 SvIV_set(dstr, SvIVX(sstr));
3816 /* SvTAINTED can only be true if the SV has taint magic, which in
3817 turn means that the SV type is PVMG (or greater). This is the
3818 case statement for SVt_IV, so this cannot be true (whatever gcov
3820 assert(!SvTAINTED(sstr));
3825 if (dtype < SVt_PV && dtype != SVt_IV)
3826 sv_upgrade(dstr, SVt_IV);
3834 sv_upgrade(dstr, SVt_NV);
3838 sv_upgrade(dstr, SVt_PVNV);
3841 goto end_of_first_switch;
3843 SvNV_set(dstr, SvNVX(sstr));
3844 (void)SvNOK_only(dstr);
3845 /* SvTAINTED can only be true if the SV has taint magic, which in
3846 turn means that the SV type is PVMG (or greater). This is the
3847 case statement for SVt_NV, so this cannot be true (whatever gcov
3849 assert(!SvTAINTED(sstr));
3855 #ifdef PERL_OLD_COPY_ON_WRITE
3856 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3857 if (dtype < SVt_PVIV)
3858 sv_upgrade(dstr, SVt_PVIV);
3865 sv_upgrade(dstr, SVt_PV);
3868 if (dtype < SVt_PVIV)
3869 sv_upgrade(dstr, SVt_PVIV);
3872 if (dtype < SVt_PVNV)
3873 sv_upgrade(dstr, SVt_PVNV);
3877 const char * const type = sv_reftype(sstr,0);
3879 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
3881 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3886 if (dtype < SVt_REGEXP)
3887 sv_upgrade(dstr, SVt_REGEXP);
3890 /* case SVt_BIND: */
3893 if (isGV_with_GP(sstr) && dtype <= SVt_PVGV) {
3894 glob_assign_glob(dstr, sstr, dtype);
3897 /* SvVALID means that this PVGV is playing at being an FBM. */
3901 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
3903 if (SvTYPE(sstr) != stype) {
3904 stype = SvTYPE(sstr);
3905 if (isGV_with_GP(sstr) && stype == SVt_PVGV && dtype <= SVt_PVGV) {
3906 glob_assign_glob(dstr, sstr, dtype);
3911 if (stype == SVt_PVLV)
3912 SvUPGRADE(dstr, SVt_PVNV);
3914 SvUPGRADE(dstr, (svtype)stype);
3916 end_of_first_switch:
3918 /* dstr may have been upgraded. */
3919 dtype = SvTYPE(dstr);
3920 sflags = SvFLAGS(sstr);
3922 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
3923 /* Assigning to a subroutine sets the prototype. */
3926 const char *const ptr = SvPV_const(sstr, len);
3928 SvGROW(dstr, len + 1);
3929 Copy(ptr, SvPVX(dstr), len + 1, char);
3930 SvCUR_set(dstr, len);
3932 SvFLAGS(dstr) |= sflags & SVf_UTF8;
3936 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3937 const char * const type = sv_reftype(dstr,0);
3939 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
3941 Perl_croak(aTHX_ "Cannot copy to %s", type);
3942 } else if (sflags & SVf_ROK) {
3943 if (isGV_with_GP(dstr) && dtype == SVt_PVGV
3944 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
3947 if (GvIMPORTED(dstr) != GVf_IMPORTED
3948 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3950 GvIMPORTED_on(dstr);
3955 glob_assign_glob(dstr, sstr, dtype);
3959 if (dtype >= SVt_PV) {
3960 if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
3961 glob_assign_ref(dstr, sstr);
3964 if (SvPVX_const(dstr)) {
3970 (void)SvOK_off(dstr);
3971 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
3972 SvFLAGS(dstr) |= sflags & SVf_ROK;
3973 assert(!(sflags & SVp_NOK));
3974 assert(!(sflags & SVp_IOK));
3975 assert(!(sflags & SVf_NOK));
3976 assert(!(sflags & SVf_IOK));
3978 else if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
3979 if (!(sflags & SVf_OK)) {
3980 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3981 "Undefined value assigned to typeglob");
3984 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
3985 if (dstr != (const SV *)gv) {
3987 gp_free(MUTABLE_GV(dstr));
3988 GvGP(dstr) = gp_ref(GvGP(gv));
3992 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
3993 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
3995 else if (sflags & SVp_POK) {
3999 * Check to see if we can just swipe the string. If so, it's a
4000 * possible small lose on short strings, but a big win on long ones.
4001 * It might even be a win on short strings if SvPVX_const(dstr)
4002 * has to be allocated and SvPVX_const(sstr) has to be freed.
4003 * Likewise if we can set up COW rather than doing an actual copy, we
4004 * drop to the else clause, as the swipe code and the COW setup code
4005 * have much in common.
4008 /* Whichever path we take through the next code, we want this true,
4009 and doing it now facilitates the COW check. */
4010 (void)SvPOK_only(dstr);
4013 /* If we're already COW then this clause is not true, and if COW
4014 is allowed then we drop down to the else and make dest COW
4015 with us. If caller hasn't said that we're allowed to COW
4016 shared hash keys then we don't do the COW setup, even if the
4017 source scalar is a shared hash key scalar. */
4018 (((flags & SV_COW_SHARED_HASH_KEYS)
4019 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4020 : 1 /* If making a COW copy is forbidden then the behaviour we
4021 desire is as if the source SV isn't actually already
4022 COW, even if it is. So we act as if the source flags
4023 are not COW, rather than actually testing them. */
4025 #ifndef PERL_OLD_COPY_ON_WRITE
4026 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4027 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4028 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4029 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4030 but in turn, it's somewhat dead code, never expected to go
4031 live, but more kept as a placeholder on how to do it better
4032 in a newer implementation. */
4033 /* If we are COW and dstr is a suitable target then we drop down
4034 into the else and make dest a COW of us. */
4035 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4040 (sflags & SVs_TEMP) && /* slated for free anyway? */
4041 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4042 (!(flags & SV_NOSTEAL)) &&
4043 /* and we're allowed to steal temps */
4044 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4045 SvLEN(sstr)) /* and really is a string */
4046 #ifdef PERL_OLD_COPY_ON_WRITE
4047 && ((flags & SV_COW_SHARED_HASH_KEYS)
4048 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4049 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4050 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4054 /* Failed the swipe test, and it's not a shared hash key either.
4055 Have to copy the string. */
4056 STRLEN len = SvCUR(sstr);
4057 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4058 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4059 SvCUR_set(dstr, len);
4060 *SvEND(dstr) = '\0';
4062 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4064 /* Either it's a shared hash key, or it's suitable for
4065 copy-on-write or we can swipe the string. */
4067 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4071 #ifdef PERL_OLD_COPY_ON_WRITE
4073 if ((sflags & (SVf_FAKE | SVf_READONLY))
4074 != (SVf_FAKE | SVf_READONLY)) {
4075 SvREADONLY_on(sstr);
4077 /* Make the source SV into a loop of 1.
4078 (about to become 2) */
4079 SV_COW_NEXT_SV_SET(sstr, sstr);
4083 /* Initial code is common. */
4084 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4089 /* making another shared SV. */
4090 STRLEN cur = SvCUR(sstr);
4091 STRLEN len = SvLEN(sstr);
4092 #ifdef PERL_OLD_COPY_ON_WRITE
4094 assert (SvTYPE(dstr) >= SVt_PVIV);
4095 /* SvIsCOW_normal */
4096 /* splice us in between source and next-after-source. */
4097 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4098 SV_COW_NEXT_SV_SET(sstr, dstr);
4099 SvPV_set(dstr, SvPVX_mutable(sstr));
4103 /* SvIsCOW_shared_hash */
4104 DEBUG_C(PerlIO_printf(Perl_debug_log,
4105 "Copy on write: Sharing hash\n"));
4107 assert (SvTYPE(dstr) >= SVt_PV);
4109 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4111 SvLEN_set(dstr, len);
4112 SvCUR_set(dstr, cur);
4113 SvREADONLY_on(dstr);
4117 { /* Passes the swipe test. */
4118 SvPV_set(dstr, SvPVX_mutable(sstr));
4119 SvLEN_set(dstr, SvLEN(sstr));
4120 SvCUR_set(dstr, SvCUR(sstr));
4123 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4124 SvPV_set(sstr, NULL);
4130 if (sflags & SVp_NOK) {
4131 SvNV_set(dstr, SvNVX(sstr));
4133 if (sflags & SVp_IOK) {
4134 SvIV_set(dstr, SvIVX(sstr));
4135 /* Must do this otherwise some other overloaded use of 0x80000000
4136 gets confused. I guess SVpbm_VALID */
4137 if (sflags & SVf_IVisUV)
4140 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4142 const MAGIC * const smg = SvVSTRING_mg(sstr);
4144 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4145 smg->mg_ptr, smg->mg_len);
4146 SvRMAGICAL_on(dstr);
4150 else if (sflags & (SVp_IOK|SVp_NOK)) {
4151 (void)SvOK_off(dstr);
4152 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4153 if (sflags & SVp_IOK) {
4154 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4155 SvIV_set(dstr, SvIVX(sstr));
4157 if (sflags & SVp_NOK) {
4158 SvNV_set(dstr, SvNVX(sstr));
4162 if (isGV_with_GP(sstr)) {
4163 /* This stringification rule for globs is spread in 3 places.
4164 This feels bad. FIXME. */
4165 const U32 wasfake = sflags & SVf_FAKE;
4167 /* FAKE globs can get coerced, so need to turn this off
4168 temporarily if it is on. */
4170 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4171 SvFLAGS(sstr) |= wasfake;
4174 (void)SvOK_off(dstr);
4176 if (SvTAINTED(sstr))
4181 =for apidoc sv_setsv_mg
4183 Like C<sv_setsv>, but also handles 'set' magic.
4189 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4191 PERL_ARGS_ASSERT_SV_SETSV_MG;
4193 sv_setsv(dstr,sstr);
4197 #ifdef PERL_OLD_COPY_ON_WRITE
4199 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4201 STRLEN cur = SvCUR(sstr);
4202 STRLEN len = SvLEN(sstr);
4203 register char *new_pv;
4205 PERL_ARGS_ASSERT_SV_SETSV_COW;
4208 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4209 (void*)sstr, (void*)dstr);
4216 if (SvTHINKFIRST(dstr))
4217 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4218 else if (SvPVX_const(dstr))
4219 Safefree(SvPVX_const(dstr));
4223 SvUPGRADE(dstr, SVt_PVIV);
4225 assert (SvPOK(sstr));
4226 assert (SvPOKp(sstr));
4227 assert (!SvIOK(sstr));
4228 assert (!SvIOKp(sstr));
4229 assert (!SvNOK(sstr));
4230 assert (!SvNOKp(sstr));
4232 if (SvIsCOW(sstr)) {
4234 if (SvLEN(sstr) == 0) {
4235 /* source is a COW shared hash key. */
4236 DEBUG_C(PerlIO_printf(Perl_debug_log,
4237 "Fast copy on write: Sharing hash\n"));
4238 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4241 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4243 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4244 SvUPGRADE(sstr, SVt_PVIV);
4245 SvREADONLY_on(sstr);
4247 DEBUG_C(PerlIO_printf(Perl_debug_log,
4248 "Fast copy on write: Converting sstr to COW\n"));
4249 SV_COW_NEXT_SV_SET(dstr, sstr);
4251 SV_COW_NEXT_SV_SET(sstr, dstr);
4252 new_pv = SvPVX_mutable(sstr);
4255 SvPV_set(dstr, new_pv);
4256 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4259 SvLEN_set(dstr, len);
4260 SvCUR_set(dstr, cur);
4269 =for apidoc sv_setpvn
4271 Copies a string into an SV. The C<len> parameter indicates the number of
4272 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4273 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4279 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4282 register char *dptr;
4284 PERL_ARGS_ASSERT_SV_SETPVN;
4286 SV_CHECK_THINKFIRST_COW_DROP(sv);
4292 /* len is STRLEN which is unsigned, need to copy to signed */
4295 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4297 SvUPGRADE(sv, SVt_PV);
4299 dptr = SvGROW(sv, len + 1);
4300 Move(ptr,dptr,len,char);
4303 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4308 =for apidoc sv_setpvn_mg
4310 Like C<sv_setpvn>, but also handles 'set' magic.
4316 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4318 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4320 sv_setpvn(sv,ptr,len);
4325 =for apidoc sv_setpv
4327 Copies a string into an SV. The string must be null-terminated. Does not
4328 handle 'set' magic. See C<sv_setpv_mg>.
4334 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4337 register STRLEN len;
4339 PERL_ARGS_ASSERT_SV_SETPV;
4341 SV_CHECK_THINKFIRST_COW_DROP(sv);
4347 SvUPGRADE(sv, SVt_PV);
4349 SvGROW(sv, len + 1);
4350 Move(ptr,SvPVX(sv),len+1,char);
4352 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4357 =for apidoc sv_setpv_mg
4359 Like C<sv_setpv>, but also handles 'set' magic.
4365 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4367 PERL_ARGS_ASSERT_SV_SETPV_MG;
4374 =for apidoc sv_usepvn_flags
4376 Tells an SV to use C<ptr> to find its string value. Normally the
4377 string is stored inside the SV but sv_usepvn allows the SV to use an
4378 outside string. The C<ptr> should point to memory that was allocated
4379 by C<malloc>. The string length, C<len>, must be supplied. By default
4380 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4381 so that pointer should not be freed or used by the programmer after
4382 giving it to sv_usepvn, and neither should any pointers from "behind"
4383 that pointer (e.g. ptr + 1) be used.
4385 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4386 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4387 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4388 C<len>, and already meets the requirements for storing in C<SvPVX>)
4394 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4399 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4401 SV_CHECK_THINKFIRST_COW_DROP(sv);
4402 SvUPGRADE(sv, SVt_PV);
4405 if (flags & SV_SMAGIC)
4409 if (SvPVX_const(sv))
4413 if (flags & SV_HAS_TRAILING_NUL)
4414 assert(ptr[len] == '\0');
4417 allocate = (flags & SV_HAS_TRAILING_NUL)
4419 #ifdef Perl_safesysmalloc_size
4422 PERL_STRLEN_ROUNDUP(len + 1);
4424 if (flags & SV_HAS_TRAILING_NUL) {
4425 /* It's long enough - do nothing.
4426 Specfically Perl_newCONSTSUB is relying on this. */
4429 /* Force a move to shake out bugs in callers. */
4430 char *new_ptr = (char*)safemalloc(allocate);
4431 Copy(ptr, new_ptr, len, char);
4432 PoisonFree(ptr,len,char);
4436 ptr = (char*) saferealloc (ptr, allocate);
4439 #ifdef Perl_safesysmalloc_size
4440 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4442 SvLEN_set(sv, allocate);
4446 if (!(flags & SV_HAS_TRAILING_NUL)) {
4449 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4451 if (flags & SV_SMAGIC)
4455 #ifdef PERL_OLD_COPY_ON_WRITE
4456 /* Need to do this *after* making the SV normal, as we need the buffer
4457 pointer to remain valid until after we've copied it. If we let go too early,
4458 another thread could invalidate it by unsharing last of the same hash key
4459 (which it can do by means other than releasing copy-on-write Svs)
4460 or by changing the other copy-on-write SVs in the loop. */
4462 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4464 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4466 { /* this SV was SvIsCOW_normal(sv) */
4467 /* we need to find the SV pointing to us. */
4468 SV *current = SV_COW_NEXT_SV(after);
4470 if (current == sv) {
4471 /* The SV we point to points back to us (there were only two of us
4473 Hence other SV is no longer copy on write either. */
4475 SvREADONLY_off(after);
4477 /* We need to follow the pointers around the loop. */
4479 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4482 /* don't loop forever if the structure is bust, and we have
4483 a pointer into a closed loop. */
4484 assert (current != after);
4485 assert (SvPVX_const(current) == pvx);
4487 /* Make the SV before us point to the SV after us. */
4488 SV_COW_NEXT_SV_SET(current, after);
4494 =for apidoc sv_force_normal_flags
4496 Undo various types of fakery on an SV: if the PV is a shared string, make
4497 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4498 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4499 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4500 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4501 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4502 set to some other value.) In addition, the C<flags> parameter gets passed to
4503 C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4504 with flags set to 0.
4510 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4514 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4516 #ifdef PERL_OLD_COPY_ON_WRITE
4517 if (SvREADONLY(sv)) {
4519 const char * const pvx = SvPVX_const(sv);
4520 const STRLEN len = SvLEN(sv);
4521 const STRLEN cur = SvCUR(sv);
4522 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4523 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4524 we'll fail an assertion. */
4525 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4528 PerlIO_printf(Perl_debug_log,
4529 "Copy on write: Force normal %ld\n",
4535 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4538 if (flags & SV_COW_DROP_PV) {
4539 /* OK, so we don't need to copy our buffer. */
4542 SvGROW(sv, cur + 1);
4543 Move(pvx,SvPVX(sv),cur,char);
4548 sv_release_COW(sv, pvx, next);
4550 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4556 else if (IN_PERL_RUNTIME)
4557 Perl_croak_no_modify(aTHX);
4560 if (SvREADONLY(sv)) {
4562 const char * const pvx = SvPVX_const(sv);
4563 const STRLEN len = SvCUR(sv);
4568 SvGROW(sv, len + 1);
4569 Move(pvx,SvPVX(sv),len,char);
4571 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4573 else if (IN_PERL_RUNTIME)
4574 Perl_croak_no_modify(aTHX);
4578 sv_unref_flags(sv, flags);
4579 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4581 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4582 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analagous
4583 to sv_unglob. We only need it here, so inline it. */
4584 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4585 SV *const temp = newSV_type(new_type);
4586 void *const temp_p = SvANY(sv);
4588 if (new_type == SVt_PVMG) {
4589 SvMAGIC_set(temp, SvMAGIC(sv));
4590 SvMAGIC_set(sv, NULL);
4591 SvSTASH_set(temp, SvSTASH(sv));
4592 SvSTASH_set(sv, NULL);
4594 SvCUR_set(temp, SvCUR(sv));
4595 /* Remember that SvPVX is in the head, not the body. */
4597 SvLEN_set(temp, SvLEN(sv));
4598 /* This signals "buffer is owned by someone else" in sv_clear,
4599 which is the least effort way to stop it freeing the buffer.
4601 SvLEN_set(sv, SvLEN(sv)+1);
4603 /* Their buffer is already owned by someone else. */
4604 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4605 SvLEN_set(temp, SvCUR(sv)+1);
4608 /* Now swap the rest of the bodies. */
4610 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4611 SvFLAGS(sv) |= new_type;
4612 SvANY(sv) = SvANY(temp);
4614 SvFLAGS(temp) &= ~(SVTYPEMASK);
4615 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4616 SvANY(temp) = temp_p;
4625 Efficient removal of characters from the beginning of the string buffer.
4626 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4627 the string buffer. The C<ptr> becomes the first character of the adjusted
4628 string. Uses the "OOK hack".
4629 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4630 refer to the same chunk of data.
4636 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4642 const U8 *real_start;
4646 PERL_ARGS_ASSERT_SV_CHOP;
4648 if (!ptr || !SvPOKp(sv))
4650 delta = ptr - SvPVX_const(sv);
4652 /* Nothing to do. */
4655 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4656 nothing uses the value of ptr any more. */
4657 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4658 if (ptr <= SvPVX_const(sv))
4659 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4660 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4661 SV_CHECK_THINKFIRST(sv);
4662 if (delta > max_delta)
4663 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4664 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4665 SvPVX_const(sv) + max_delta);
4668 if (!SvLEN(sv)) { /* make copy of shared string */
4669 const char *pvx = SvPVX_const(sv);
4670 const STRLEN len = SvCUR(sv);
4671 SvGROW(sv, len + 1);
4672 Move(pvx,SvPVX(sv),len,char);
4675 SvFLAGS(sv) |= SVf_OOK;
4678 SvOOK_offset(sv, old_delta);
4680 SvLEN_set(sv, SvLEN(sv) - delta);
4681 SvCUR_set(sv, SvCUR(sv) - delta);
4682 SvPV_set(sv, SvPVX(sv) + delta);
4684 p = (U8 *)SvPVX_const(sv);
4689 real_start = p - delta;
4693 if (delta < 0x100) {
4697 p -= sizeof(STRLEN);
4698 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4702 /* Fill the preceding buffer with sentinals to verify that no-one is
4704 while (p > real_start) {
4712 =for apidoc sv_catpvn
4714 Concatenates the string onto the end of the string which is in the SV. The
4715 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4716 status set, then the bytes appended should be valid UTF-8.
4717 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4719 =for apidoc sv_catpvn_flags
4721 Concatenates the string onto the end of the string which is in the SV. The
4722 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4723 status set, then the bytes appended should be valid UTF-8.
4724 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4725 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4726 in terms of this function.
4732 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4736 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4738 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4740 SvGROW(dsv, dlen + slen + 1);
4742 sstr = SvPVX_const(dsv);
4743 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4744 SvCUR_set(dsv, SvCUR(dsv) + slen);
4746 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4748 if (flags & SV_SMAGIC)
4753 =for apidoc sv_catsv
4755 Concatenates the string from SV C<ssv> onto the end of the string in
4756 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4757 not 'set' magic. See C<sv_catsv_mg>.
4759 =for apidoc sv_catsv_flags
4761 Concatenates the string from SV C<ssv> onto the end of the string in
4762 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4763 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4764 and C<sv_catsv_nomg> are implemented in terms of this function.
4769 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4773 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4777 const char *spv = SvPV_const(ssv, slen);
4779 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4780 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4781 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4782 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4783 dsv->sv_flags doesn't have that bit set.
4784 Andy Dougherty 12 Oct 2001
4786 const I32 sutf8 = DO_UTF8(ssv);
4789 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4791 dutf8 = DO_UTF8(dsv);
4793 if (dutf8 != sutf8) {
4795 /* Not modifying source SV, so taking a temporary copy. */
4796 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
4798 sv_utf8_upgrade(csv);
4799 spv = SvPV_const(csv, slen);
4802 /* Leave enough space for the cat that's about to happen */
4803 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
4805 sv_catpvn_nomg(dsv, spv, slen);
4808 if (flags & SV_SMAGIC)
4813 =for apidoc sv_catpv
4815 Concatenates the string onto the end of the string which is in the SV.
4816 If the SV has the UTF-8 status set, then the bytes appended should be
4817 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
4822 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
4825 register STRLEN len;
4829 PERL_ARGS_ASSERT_SV_CATPV;
4833 junk = SvPV_force(sv, tlen);
4835 SvGROW(sv, tlen + len + 1);
4837 ptr = SvPVX_const(sv);
4838 Move(ptr,SvPVX(sv)+tlen,len+1,char);
4839 SvCUR_set(sv, SvCUR(sv) + len);
4840 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4845 =for apidoc sv_catpv_mg
4847 Like C<sv_catpv>, but also handles 'set' magic.
4853 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4855 PERL_ARGS_ASSERT_SV_CATPV_MG;
4864 Creates a new SV. A non-zero C<len> parameter indicates the number of
4865 bytes of preallocated string space the SV should have. An extra byte for a
4866 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4867 space is allocated.) The reference count for the new SV is set to 1.
4869 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4870 parameter, I<x>, a debug aid which allowed callers to identify themselves.
4871 This aid has been superseded by a new build option, PERL_MEM_LOG (see
4872 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4873 modules supporting older perls.
4879 Perl_newSV(pTHX_ const STRLEN len)
4886 sv_upgrade(sv, SVt_PV);
4887 SvGROW(sv, len + 1);
4892 =for apidoc sv_magicext
4894 Adds magic to an SV, upgrading it if necessary. Applies the
4895 supplied vtable and returns a pointer to the magic added.
4897 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4898 In particular, you can add magic to SvREADONLY SVs, and add more than
4899 one instance of the same 'how'.
4901 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4902 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4903 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4904 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
4906 (This is now used as a subroutine by C<sv_magic>.)
4911 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
4912 const MGVTBL *const vtable, const char *const name, const I32 namlen)
4917 PERL_ARGS_ASSERT_SV_MAGICEXT;
4919 SvUPGRADE(sv, SVt_PVMG);
4920 Newxz(mg, 1, MAGIC);
4921 mg->mg_moremagic = SvMAGIC(sv);
4922 SvMAGIC_set(sv, mg);
4924 /* Sometimes a magic contains a reference loop, where the sv and
4925 object refer to each other. To prevent a reference loop that
4926 would prevent such objects being freed, we look for such loops
4927 and if we find one we avoid incrementing the object refcount.
4929 Note we cannot do this to avoid self-tie loops as intervening RV must
4930 have its REFCNT incremented to keep it in existence.
4933 if (!obj || obj == sv ||
4934 how == PERL_MAGIC_arylen ||
4935 how == PERL_MAGIC_symtab ||
4936 (SvTYPE(obj) == SVt_PVGV &&
4937 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
4938 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
4939 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
4944 mg->mg_obj = SvREFCNT_inc_simple(obj);
4945 mg->mg_flags |= MGf_REFCOUNTED;
4948 /* Normal self-ties simply pass a null object, and instead of
4949 using mg_obj directly, use the SvTIED_obj macro to produce a
4950 new RV as needed. For glob "self-ties", we are tieing the PVIO
4951 with an RV obj pointing to the glob containing the PVIO. In
4952 this case, to avoid a reference loop, we need to weaken the
4956 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4957 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
4963 mg->mg_len = namlen;
4966 mg->mg_ptr = savepvn(name, namlen);
4967 else if (namlen == HEf_SVKEY) {
4968 /* Yes, this is casting away const. This is only for the case of
4969 HEf_SVKEY. I think we need to document this abberation of the
4970 constness of the API, rather than making name non-const, as
4971 that change propagating outwards a long way. */
4972 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
4974 mg->mg_ptr = (char *) name;
4976 mg->mg_virtual = (MGVTBL *) vtable;
4980 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4985 =for apidoc sv_magic
4987 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4988 then adds a new magic item of type C<how> to the head of the magic list.
4990 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4991 handling of the C<name> and C<namlen> arguments.
4993 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4994 to add more than one instance of the same 'how'.
5000 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5001 const char *const name, const I32 namlen)
5004 const MGVTBL *vtable;
5007 PERL_ARGS_ASSERT_SV_MAGIC;
5009 #ifdef PERL_OLD_COPY_ON_WRITE
5011 sv_force_normal_flags(sv, 0);
5013 if (SvREADONLY(sv)) {
5015 /* its okay to attach magic to shared strings; the subsequent
5016 * upgrade to PVMG will unshare the string */
5017 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5020 && how != PERL_MAGIC_regex_global
5021 && how != PERL_MAGIC_bm
5022 && how != PERL_MAGIC_fm
5023 && how != PERL_MAGIC_sv
5024 && how != PERL_MAGIC_backref
5027 Perl_croak_no_modify(aTHX);
5030 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5031 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5032 /* sv_magic() refuses to add a magic of the same 'how' as an
5035 if (how == PERL_MAGIC_taint) {
5037 /* Any scalar which already had taint magic on which someone
5038 (erroneously?) did SvIOK_on() or similar will now be
5039 incorrectly sporting public "OK" flags. */
5040 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5048 vtable = &PL_vtbl_sv;
5050 case PERL_MAGIC_overload:
5051 vtable = &PL_vtbl_amagic;
5053 case PERL_MAGIC_overload_elem:
5054 vtable = &PL_vtbl_amagicelem;
5056 case PERL_MAGIC_overload_table:
5057 vtable = &PL_vtbl_ovrld;
5060 vtable = &PL_vtbl_bm;
5062 case PERL_MAGIC_regdata:
5063 vtable = &PL_vtbl_regdata;
5065 case PERL_MAGIC_regdatum:
5066 vtable = &PL_vtbl_regdatum;
5068 case PERL_MAGIC_env:
5069 vtable = &PL_vtbl_env;
5072 vtable = &PL_vtbl_fm;
5074 case PERL_MAGIC_envelem:
5075 vtable = &PL_vtbl_envelem;
5077 case PERL_MAGIC_regex_global:
5078 vtable = &PL_vtbl_mglob;
5080 case PERL_MAGIC_isa:
5081 vtable = &PL_vtbl_isa;
5083 case PERL_MAGIC_isaelem:
5084 vtable = &PL_vtbl_isaelem;
5086 case PERL_MAGIC_nkeys:
5087 vtable = &PL_vtbl_nkeys;
5089 case PERL_MAGIC_dbfile:
5092 case PERL_MAGIC_dbline:
5093 vtable = &PL_vtbl_dbline;
5095 #ifdef USE_LOCALE_COLLATE
5096 case PERL_MAGIC_collxfrm:
5097 vtable = &PL_vtbl_collxfrm;
5099 #endif /* USE_LOCALE_COLLATE */
5100 case PERL_MAGIC_tied:
5101 vtable = &PL_vtbl_pack;
5103 case PERL_MAGIC_tiedelem:
5104 case PERL_MAGIC_tiedscalar:
5105 vtable = &PL_vtbl_packelem;
5108 vtable = &PL_vtbl_regexp;
5110 case PERL_MAGIC_sig:
5111 vtable = &PL_vtbl_sig;
5113 case PERL_MAGIC_sigelem:
5114 vtable = &PL_vtbl_sigelem;
5116 case PERL_MAGIC_taint:
5117 vtable = &PL_vtbl_taint;
5119 case PERL_MAGIC_uvar:
5120 vtable = &PL_vtbl_uvar;
5122 case PERL_MAGIC_vec:
5123 vtable = &PL_vtbl_vec;
5125 case PERL_MAGIC_arylen_p:
5126 case PERL_MAGIC_rhash:
5127 case PERL_MAGIC_symtab:
5128 case PERL_MAGIC_vstring:
5131 case PERL_MAGIC_utf8:
5132 vtable = &PL_vtbl_utf8;
5134 case PERL_MAGIC_substr:
5135 vtable = &PL_vtbl_substr;
5137 case PERL_MAGIC_defelem:
5138 vtable = &PL_vtbl_defelem;
5140 case PERL_MAGIC_arylen:
5141 vtable = &PL_vtbl_arylen;
5143 case PERL_MAGIC_pos:
5144 vtable = &PL_vtbl_pos;
5146 case PERL_MAGIC_backref:
5147 vtable = &PL_vtbl_backref;
5149 case PERL_MAGIC_hintselem:
5150 vtable = &PL_vtbl_hintselem;
5152 case PERL_MAGIC_hints:
5153 vtable = &PL_vtbl_hints;
5155 case PERL_MAGIC_ext:
5156 /* Reserved for use by extensions not perl internals. */
5157 /* Useful for attaching extension internal data to perl vars. */
5158 /* Note that multiple extensions may clash if magical scalars */
5159 /* etc holding private data from one are passed to another. */
5163 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5166 /* Rest of work is done else where */
5167 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5170 case PERL_MAGIC_taint:
5173 case PERL_MAGIC_ext:
5174 case PERL_MAGIC_dbfile:
5181 =for apidoc sv_unmagic
5183 Removes all magic of type C<type> from an SV.
5189 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5194 PERL_ARGS_ASSERT_SV_UNMAGIC;
5196 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5198 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5199 for (mg = *mgp; mg; mg = *mgp) {
5200 if (mg->mg_type == type) {
5201 const MGVTBL* const vtbl = mg->mg_virtual;
5202 *mgp = mg->mg_moremagic;
5203 if (vtbl && vtbl->svt_free)
5204 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
5205 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5207 Safefree(mg->mg_ptr);
5208 else if (mg->mg_len == HEf_SVKEY)
5209 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5210 else if (mg->mg_type == PERL_MAGIC_utf8)
5211 Safefree(mg->mg_ptr);
5213 if (mg->mg_flags & MGf_REFCOUNTED)
5214 SvREFCNT_dec(mg->mg_obj);
5218 mgp = &mg->mg_moremagic;
5221 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5222 mg_magical(sv); /* else fix the flags now */
5226 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5232 =for apidoc sv_rvweaken
5234 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5235 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5236 push a back-reference to this RV onto the array of backreferences
5237 associated with that magic. If the RV is magical, set magic will be
5238 called after the RV is cleared.
5244 Perl_sv_rvweaken(pTHX_ SV *const sv)
5248 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5250 if (!SvOK(sv)) /* let undefs pass */
5253 Perl_croak(aTHX_ "Can't weaken a nonreference");
5254 else if (SvWEAKREF(sv)) {
5255 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5259 Perl_sv_add_backref(aTHX_ tsv, sv);
5265 /* Give tsv backref magic if it hasn't already got it, then push a
5266 * back-reference to sv onto the array associated with the backref magic.
5268 * As an optimisation, if there's only one backref and it's not an AV,
5269 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5270 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5273 * If an HV's backref is stored in magic, it is moved back to HvAUX.
5276 /* A discussion about the backreferences array and its refcount:
5278 * The AV holding the backreferences is pointed to either as the mg_obj of
5279 * PERL_MAGIC_backref, or in the specific case of a HV that has the hv_aux
5280 * structure, from the xhv_backreferences field. (A HV without hv_aux will
5281 * have the standard magic instead.) The array is created with a refcount
5282 * of 2. This means that if during global destruction the array gets
5283 * picked on before its parent to have its refcount decremented by the
5284 * random zapper, it won't actually be freed, meaning it's still there for
5285 * when its parent gets freed.
5287 * When the parent SV is freed, the extra ref is killed by
5288 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5289 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5291 * When a single backref SV is stored directly, it is not reference
5296 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5303 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5305 /* find slot to store array or singleton backref */
5307 if (SvTYPE(tsv) == SVt_PVHV) {
5308 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5311 if ((mg = mg_find(tsv, PERL_MAGIC_backref))) {
5312 /* Aha. They've got it stowed in magic instead.
5313 * Move it back to xhv_backreferences */
5315 /* Stop mg_free decreasing the reference count. */
5317 /* Stop mg_free even calling the destructor, given that
5318 there's no AV to free up. */
5320 sv_unmagic(tsv, PERL_MAGIC_backref);
5326 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5328 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5329 mg = mg_find(tsv, PERL_MAGIC_backref);
5331 svp = &(mg->mg_obj);
5334 /* create or retrieve the array */
5336 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5337 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5342 SvREFCNT_inc_simple_void(av);
5343 /* av now has a refcnt of 2; see discussion above */
5345 /* move single existing backref to the array */
5347 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5351 mg->mg_flags |= MGf_REFCOUNTED;
5354 av = MUTABLE_AV(*svp);
5357 /* optimisation: store single backref directly in HvAUX or mg_obj */
5361 /* push new backref */
5362 assert(SvTYPE(av) == SVt_PVAV);
5363 if (AvFILLp(av) >= AvMAX(av)) {
5364 av_extend(av, AvFILLp(av)+1);
5366 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5369 /* delete a back-reference to ourselves from the backref magic associated
5370 * with the SV we point to.
5374 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5380 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5382 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
5383 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5385 if (!svp || !*svp) {
5387 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5388 svp = mg ? &(mg->mg_obj) : NULL;
5392 Perl_croak(aTHX_ "panic: del_backref");
5394 if (SvTYPE(*svp) == SVt_PVAV) {
5396 AV * const av = (AV*)*svp;
5397 assert(!SvIS_FREED(av));
5399 for (i = AvFILLp(av); i >= 0; i--) {
5401 const SSize_t fill = AvFILLp(av);
5403 /* We weren't the last entry.
5404 An unordered list has this property that you can take the
5405 last element off the end to fill the hole, and it's still
5406 an unordered list :-)
5411 AvFILLp(av) = fill - 1;
5414 break; /* should only be one */
5421 /* optimisation: only a single backref, stored directly */
5423 Perl_croak(aTHX_ "panic: del_backref");
5430 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5436 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5441 is_array = (SvTYPE(av) == SVt_PVAV);
5443 assert(!SvIS_FREED(av));
5446 last = svp + AvFILLp(av);
5449 /* optimisation: only a single backref, stored directly */
5455 while (svp <= last) {
5457 SV *const referrer = *svp;
5458 if (SvWEAKREF(referrer)) {
5459 /* XXX Should we check that it hasn't changed? */
5460 assert(SvROK(referrer));
5461 SvRV_set(referrer, 0);
5463 SvWEAKREF_off(referrer);
5464 SvSETMAGIC(referrer);
5465 } else if (SvTYPE(referrer) == SVt_PVGV ||
5466 SvTYPE(referrer) == SVt_PVLV) {
5467 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5468 /* You lookin' at me? */
5469 assert(GvSTASH(referrer));
5470 assert(GvSTASH(referrer) == (const HV *)sv);
5471 GvSTASH(referrer) = 0;
5472 } else if (SvTYPE(referrer) == SVt_PVCV ||
5473 SvTYPE(referrer) == SVt_PVFM) {
5474 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5475 /* You lookin' at me? */
5476 assert(CvSTASH(referrer));
5477 assert(CvSTASH(referrer) == (const HV *)sv);
5478 CvSTASH(referrer) = 0;
5481 assert(SvTYPE(sv) == SVt_PVGV);
5482 /* You lookin' at me? */
5483 assert(CvGV(referrer));
5484 assert(CvGV(referrer) == (const GV *)sv);
5485 anonymise_cv_maybe(MUTABLE_GV(sv),
5486 MUTABLE_CV(referrer));
5491 "panic: magic_killbackrefs (flags=%"UVxf")",
5492 (UV)SvFLAGS(referrer));
5503 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5509 =for apidoc sv_insert
5511 Inserts a string at the specified offset/length within the SV. Similar to
5512 the Perl substr() function. Handles get magic.
5514 =for apidoc sv_insert_flags
5516 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5522 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5527 register char *midend;
5528 register char *bigend;
5532 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5535 Perl_croak(aTHX_ "Can't modify non-existent substring");
5536 SvPV_force_flags(bigstr, curlen, flags);
5537 (void)SvPOK_only_UTF8(bigstr);
5538 if (offset + len > curlen) {
5539 SvGROW(bigstr, offset+len+1);
5540 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5541 SvCUR_set(bigstr, offset+len);
5545 i = littlelen - len;
5546 if (i > 0) { /* string might grow */
5547 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5548 mid = big + offset + len;
5549 midend = bigend = big + SvCUR(bigstr);
5552 while (midend > mid) /* shove everything down */
5553 *--bigend = *--midend;
5554 Move(little,big+offset,littlelen,char);
5555 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5560 Move(little,SvPVX(bigstr)+offset,len,char);
5565 big = SvPVX(bigstr);
5568 bigend = big + SvCUR(bigstr);
5570 if (midend > bigend)
5571 Perl_croak(aTHX_ "panic: sv_insert");
5573 if (mid - big > bigend - midend) { /* faster to shorten from end */
5575 Move(little, mid, littlelen,char);
5578 i = bigend - midend;
5580 Move(midend, mid, i,char);
5584 SvCUR_set(bigstr, mid - big);
5586 else if ((i = mid - big)) { /* faster from front */
5587 midend -= littlelen;
5589 Move(big, midend - i, i, char);
5590 sv_chop(bigstr,midend-i);
5592 Move(little, mid, littlelen,char);
5594 else if (littlelen) {
5595 midend -= littlelen;
5596 sv_chop(bigstr,midend);
5597 Move(little,midend,littlelen,char);
5600 sv_chop(bigstr,midend);
5606 =for apidoc sv_replace
5608 Make the first argument a copy of the second, then delete the original.
5609 The target SV physically takes over ownership of the body of the source SV
5610 and inherits its flags; however, the target keeps any magic it owns,
5611 and any magic in the source is discarded.
5612 Note that this is a rather specialist SV copying operation; most of the
5613 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5619 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5622 const U32 refcnt = SvREFCNT(sv);
5624 PERL_ARGS_ASSERT_SV_REPLACE;
5626 SV_CHECK_THINKFIRST_COW_DROP(sv);
5627 if (SvREFCNT(nsv) != 1) {
5628 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5629 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5631 if (SvMAGICAL(sv)) {
5635 sv_upgrade(nsv, SVt_PVMG);
5636 SvMAGIC_set(nsv, SvMAGIC(sv));
5637 SvFLAGS(nsv) |= SvMAGICAL(sv);
5639 SvMAGIC_set(sv, NULL);
5643 assert(!SvREFCNT(sv));
5644 #ifdef DEBUG_LEAKING_SCALARS
5645 sv->sv_flags = nsv->sv_flags;
5646 sv->sv_any = nsv->sv_any;
5647 sv->sv_refcnt = nsv->sv_refcnt;
5648 sv->sv_u = nsv->sv_u;
5650 StructCopy(nsv,sv,SV);
5652 if(SvTYPE(sv) == SVt_IV) {
5654 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5658 #ifdef PERL_OLD_COPY_ON_WRITE
5659 if (SvIsCOW_normal(nsv)) {
5660 /* We need to follow the pointers around the loop to make the
5661 previous SV point to sv, rather than nsv. */
5664 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5667 assert(SvPVX_const(current) == SvPVX_const(nsv));
5669 /* Make the SV before us point to the SV after us. */
5671 PerlIO_printf(Perl_debug_log, "previous is\n");
5673 PerlIO_printf(Perl_debug_log,
5674 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5675 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5677 SV_COW_NEXT_SV_SET(current, sv);
5680 SvREFCNT(sv) = refcnt;
5681 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5686 /* We're about to free a GV which has a CV that refers back to us.
5687 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5691 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5697 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5700 assert(SvREFCNT(gv) == 0);
5701 assert(isGV(gv) && isGV_with_GP(gv));
5703 assert(!CvANON(cv));
5704 assert(CvGV(cv) == gv);
5706 /* will the CV shortly be freed by gp_free() ? */
5707 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5708 SvANY(cv)->xcv_gv = NULL;
5712 /* if not, anonymise: */
5713 stash = GvSTASH(gv) ? HvNAME(GvSTASH(gv)) : NULL;
5714 gvname = Perl_newSVpvf(aTHX_ "%s::__ANON__",
5715 stash ? stash : "__ANON__");
5716 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5717 SvREFCNT_dec(gvname);
5721 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5726 =for apidoc sv_clear
5728 Clear an SV: call any destructors, free up any memory used by the body,
5729 and free the body itself. The SV's head is I<not> freed, although
5730 its type is set to all 1's so that it won't inadvertently be assumed
5731 to be live during global destruction etc.
5732 This function should only be called when REFCNT is zero. Most of the time
5733 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5740 Perl_sv_clear(pTHX_ register SV *const sv)
5743 const U32 type = SvTYPE(sv);
5744 const struct body_details *const sv_type_details
5745 = bodies_by_type + type;
5748 PERL_ARGS_ASSERT_SV_CLEAR;
5749 assert(SvREFCNT(sv) == 0);
5750 assert(SvTYPE(sv) != SVTYPEMASK);
5752 if (type <= SVt_IV) {
5753 /* See the comment in sv.h about the collusion between this early
5754 return and the overloading of the NULL slots in the size table. */
5757 SvFLAGS(sv) &= SVf_BREAK;
5758 SvFLAGS(sv) |= SVTYPEMASK;
5763 if (PL_defstash && /* Still have a symbol table? */
5770 stash = SvSTASH(sv);
5771 destructor = StashHANDLER(stash,DESTROY);
5773 /* A constant subroutine can have no side effects, so
5774 don't bother calling it. */
5775 && !CvCONST(destructor)
5776 /* Don't bother calling an empty destructor */
5777 && (CvISXSUB(destructor)
5778 || (CvSTART(destructor)
5779 && (CvSTART(destructor)->op_next->op_type != OP_LEAVESUB))))
5781 SV* const tmpref = newRV(sv);
5782 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
5784 PUSHSTACKi(PERLSI_DESTROY);
5789 call_sv(MUTABLE_SV(destructor), G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
5795 if(SvREFCNT(tmpref) < 2) {
5796 /* tmpref is not kept alive! */
5798 SvRV_set(tmpref, NULL);
5801 SvREFCNT_dec(tmpref);
5803 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
5807 if (PL_in_clean_objs)
5808 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
5810 /* DESTROY gave object new lease on life */
5816 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
5817 SvOBJECT_off(sv); /* Curse the object. */
5818 if (type != SVt_PVIO)
5819 --PL_sv_objcount; /* XXX Might want something more general */
5822 if (type >= SVt_PVMG) {
5823 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5824 SvREFCNT_dec(SvOURSTASH(sv));
5825 } else if (SvMAGIC(sv))
5827 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5828 SvREFCNT_dec(SvSTASH(sv));
5831 /* case SVt_BIND: */
5834 IoIFP(sv) != PerlIO_stdin() &&
5835 IoIFP(sv) != PerlIO_stdout() &&
5836 IoIFP(sv) != PerlIO_stderr() &&
5837 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5839 io_close(MUTABLE_IO(sv), FALSE);
5841 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5842 PerlDir_close(IoDIRP(sv));
5843 IoDIRP(sv) = (DIR*)NULL;
5844 Safefree(IoTOP_NAME(sv));
5845 Safefree(IoFMT_NAME(sv));
5846 Safefree(IoBOTTOM_NAME(sv));
5849 /* FIXME for plugins */
5850 pregfree2((REGEXP*) sv);
5854 cv_undef(MUTABLE_CV(sv));
5855 /* If we're in a stash, we don't own a reference to it. However it does
5856 have a back reference to us, which needs to be cleared. */
5857 if ((stash = CvSTASH(sv)))
5858 sv_del_backref(MUTABLE_SV(stash), sv);
5861 if (PL_last_swash_hv == (const HV *)sv) {
5862 PL_last_swash_hv = NULL;
5864 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5865 hv_undef(MUTABLE_HV(sv));
5868 if (PL_comppad == MUTABLE_AV(sv)) {
5872 av_undef(MUTABLE_AV(sv));
5875 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5876 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5877 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5878 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5880 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5881 SvREFCNT_dec(LvTARG(sv));
5883 if (isGV_with_GP(sv)) {
5884 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
5885 && HvNAME_get(stash))
5886 mro_method_changed_in(stash);
5887 gp_free(MUTABLE_GV(sv));
5889 unshare_hek(GvNAME_HEK(sv));
5890 /* If we're in a stash, we don't own a reference to it. However it does
5891 have a back reference to us, which needs to be cleared. */
5892 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
5893 sv_del_backref(MUTABLE_SV(stash), sv);
5895 /* FIXME. There are probably more unreferenced pointers to SVs in the
5896 interpreter struct that we should check and tidy in a similar
5898 if ((const GV *)sv == PL_last_in_gv)
5899 PL_last_in_gv = NULL;
5905 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5908 SvOOK_offset(sv, offset);
5909 SvPV_set(sv, SvPVX_mutable(sv) - offset);
5910 /* Don't even bother with turning off the OOK flag. */
5915 SV * const target = SvRV(sv);
5917 sv_del_backref(target, sv);
5919 SvREFCNT_dec(target);
5922 #ifdef PERL_OLD_COPY_ON_WRITE
5923 else if (SvPVX_const(sv)
5924 && !(SvTYPE(sv) == SVt_PVIO && !(IoFLAGS(sv) & IOf_FAKE_DIRP))) {
5927 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
5931 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
5933 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5937 } else if (SvLEN(sv)) {
5938 Safefree(SvPVX_const(sv));
5942 else if (SvPVX_const(sv) && SvLEN(sv)
5943 && !(SvTYPE(sv) == SVt_PVIO && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
5944 Safefree(SvPVX_mutable(sv));
5945 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
5946 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5955 SvFLAGS(sv) &= SVf_BREAK;
5956 SvFLAGS(sv) |= SVTYPEMASK;
5958 if (sv_type_details->arena) {
5959 del_body(((char *)SvANY(sv) + sv_type_details->offset),
5960 &PL_body_roots[type]);
5962 else if (sv_type_details->body_size) {
5963 safefree(SvANY(sv));
5968 =for apidoc sv_newref
5970 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5977 Perl_sv_newref(pTHX_ SV *const sv)
5979 PERL_UNUSED_CONTEXT;
5988 Decrement an SV's reference count, and if it drops to zero, call
5989 C<sv_clear> to invoke destructors and free up any memory used by
5990 the body; finally, deallocate the SV's head itself.
5991 Normally called via a wrapper macro C<SvREFCNT_dec>.
5997 Perl_sv_free(pTHX_ SV *const sv)
6002 if (SvREFCNT(sv) == 0) {
6003 if (SvFLAGS(sv) & SVf_BREAK)
6004 /* this SV's refcnt has been artificially decremented to
6005 * trigger cleanup */
6007 if (PL_in_clean_all) /* All is fair */
6009 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6010 /* make sure SvREFCNT(sv)==0 happens very seldom */
6011 SvREFCNT(sv) = (~(U32)0)/2;
6014 if (ckWARN_d(WARN_INTERNAL)) {
6015 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6016 Perl_dump_sv_child(aTHX_ sv);
6018 #ifdef DEBUG_LEAKING_SCALARS
6021 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6022 if (PL_warnhook == PERL_WARNHOOK_FATAL
6023 || ckDEAD(packWARN(WARN_INTERNAL))) {
6024 /* Don't let Perl_warner cause us to escape our fate: */
6028 /* This may not return: */
6029 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6030 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6031 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6034 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6039 if (--(SvREFCNT(sv)) > 0)
6041 Perl_sv_free2(aTHX_ sv);
6045 Perl_sv_free2(pTHX_ SV *const sv)
6049 PERL_ARGS_ASSERT_SV_FREE2;
6053 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6054 "Attempt to free temp prematurely: SV 0x%"UVxf
6055 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6059 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6060 /* make sure SvREFCNT(sv)==0 happens very seldom */
6061 SvREFCNT(sv) = (~(U32)0)/2;
6072 Returns the length of the string in the SV. Handles magic and type
6073 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6079 Perl_sv_len(pTHX_ register SV *const sv)
6087 len = mg_length(sv);
6089 (void)SvPV_const(sv, len);
6094 =for apidoc sv_len_utf8
6096 Returns the number of characters in the string in an SV, counting wide
6097 UTF-8 bytes as a single character. Handles magic and type coercion.
6103 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6104 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6105 * (Note that the mg_len is not the length of the mg_ptr field.
6106 * This allows the cache to store the character length of the string without
6107 * needing to malloc() extra storage to attach to the mg_ptr.)
6112 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6118 return mg_length(sv);
6122 const U8 *s = (U8*)SvPV_const(sv, len);
6126 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6128 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6129 if (mg->mg_len != -1)
6132 /* We can use the offset cache for a headstart.
6133 The longer value is stored in the first pair. */
6134 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6136 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6140 if (PL_utf8cache < 0) {
6141 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6142 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6146 ulen = Perl_utf8_length(aTHX_ s, s + len);
6147 utf8_mg_len_cache_update(sv, &mg, ulen);
6151 return Perl_utf8_length(aTHX_ s, s + len);
6155 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6158 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6159 STRLEN *const uoffset_p, bool *const at_end)
6161 const U8 *s = start;
6162 STRLEN uoffset = *uoffset_p;
6164 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6166 while (s < send && uoffset) {
6173 else if (s > send) {
6175 /* This is the existing behaviour. Possibly it should be a croak, as
6176 it's actually a bounds error */
6179 *uoffset_p -= uoffset;
6183 /* Given the length of the string in both bytes and UTF-8 characters, decide
6184 whether to walk forwards or backwards to find the byte corresponding to
6185 the passed in UTF-8 offset. */
6187 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6188 STRLEN uoffset, const STRLEN uend)
6190 STRLEN backw = uend - uoffset;
6192 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6194 if (uoffset < 2 * backw) {
6195 /* The assumption is that going forwards is twice the speed of going
6196 forward (that's where the 2 * backw comes from).
6197 (The real figure of course depends on the UTF-8 data.) */
6198 const U8 *s = start;
6200 while (s < send && uoffset--)
6210 while (UTF8_IS_CONTINUATION(*send))
6213 return send - start;
6216 /* For the string representation of the given scalar, find the byte
6217 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6218 give another position in the string, *before* the sought offset, which
6219 (which is always true, as 0, 0 is a valid pair of positions), which should
6220 help reduce the amount of linear searching.
6221 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6222 will be used to reduce the amount of linear searching. The cache will be
6223 created if necessary, and the found value offered to it for update. */
6225 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6226 const U8 *const send, STRLEN uoffset,
6227 STRLEN uoffset0, STRLEN boffset0)
6229 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6231 bool at_end = FALSE;
6233 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6235 assert (uoffset >= uoffset0);
6242 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6243 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6244 if ((*mgp)->mg_ptr) {
6245 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6246 if (cache[0] == uoffset) {
6247 /* An exact match. */
6250 if (cache[2] == uoffset) {
6251 /* An exact match. */
6255 if (cache[0] < uoffset) {
6256 /* The cache already knows part of the way. */
6257 if (cache[0] > uoffset0) {
6258 /* The cache knows more than the passed in pair */
6259 uoffset0 = cache[0];
6260 boffset0 = cache[1];
6262 if ((*mgp)->mg_len != -1) {
6263 /* And we know the end too. */
6265 + sv_pos_u2b_midway(start + boffset0, send,
6267 (*mgp)->mg_len - uoffset0);
6269 uoffset -= uoffset0;
6271 + sv_pos_u2b_forwards(start + boffset0,
6272 send, &uoffset, &at_end);
6273 uoffset += uoffset0;
6276 else if (cache[2] < uoffset) {
6277 /* We're between the two cache entries. */
6278 if (cache[2] > uoffset0) {
6279 /* and the cache knows more than the passed in pair */
6280 uoffset0 = cache[2];
6281 boffset0 = cache[3];
6285 + sv_pos_u2b_midway(start + boffset0,
6288 cache[0] - uoffset0);
6291 + sv_pos_u2b_midway(start + boffset0,
6294 cache[2] - uoffset0);
6298 else if ((*mgp)->mg_len != -1) {
6299 /* If we can take advantage of a passed in offset, do so. */
6300 /* In fact, offset0 is either 0, or less than offset, so don't
6301 need to worry about the other possibility. */
6303 + sv_pos_u2b_midway(start + boffset0, send,
6305 (*mgp)->mg_len - uoffset0);
6310 if (!found || PL_utf8cache < 0) {
6311 STRLEN real_boffset;
6312 uoffset -= uoffset0;
6313 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6314 send, &uoffset, &at_end);
6315 uoffset += uoffset0;
6317 if (found && PL_utf8cache < 0)
6318 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6320 boffset = real_boffset;
6325 utf8_mg_len_cache_update(sv, mgp, uoffset);
6327 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6334 =for apidoc sv_pos_u2b_flags
6336 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6337 the start of the string, to a count of the equivalent number of bytes; if
6338 lenp is non-zero, it does the same to lenp, but this time starting from
6339 the offset, rather than from the start of the string. Handles type coercion.
6340 I<flags> is passed to C<SvPV_flags>, and usually should be
6341 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6347 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6348 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6349 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6354 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6361 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6363 start = (U8*)SvPV_flags(sv, len, flags);
6365 const U8 * const send = start + len;
6367 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6370 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6371 is 0, and *lenp is already set to that. */) {
6372 /* Convert the relative offset to absolute. */
6373 const STRLEN uoffset2 = uoffset + *lenp;
6374 const STRLEN boffset2
6375 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6376 uoffset, boffset) - boffset;
6390 =for apidoc sv_pos_u2b
6392 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6393 the start of the string, to a count of the equivalent number of bytes; if
6394 lenp is non-zero, it does the same to lenp, but this time starting from
6395 the offset, rather than from the start of the string. Handles magic and
6398 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6405 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6406 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6407 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6411 /* This function is subject to size and sign problems */
6414 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6416 PERL_ARGS_ASSERT_SV_POS_U2B;
6419 STRLEN ulen = (STRLEN)*lenp;
6420 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6421 SV_GMAGIC|SV_CONST_RETURN);
6424 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6425 SV_GMAGIC|SV_CONST_RETURN);
6430 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6433 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6437 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6438 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6439 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6443 (*mgp)->mg_len = ulen;
6444 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6445 if (ulen != (STRLEN) (*mgp)->mg_len)
6446 (*mgp)->mg_len = -1;
6449 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6450 byte length pairing. The (byte) length of the total SV is passed in too,
6451 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6452 may not have updated SvCUR, so we can't rely on reading it directly.
6454 The proffered utf8/byte length pairing isn't used if the cache already has
6455 two pairs, and swapping either for the proffered pair would increase the
6456 RMS of the intervals between known byte offsets.
6458 The cache itself consists of 4 STRLEN values
6459 0: larger UTF-8 offset
6460 1: corresponding byte offset
6461 2: smaller UTF-8 offset
6462 3: corresponding byte offset
6464 Unused cache pairs have the value 0, 0.
6465 Keeping the cache "backwards" means that the invariant of
6466 cache[0] >= cache[2] is maintained even with empty slots, which means that
6467 the code that uses it doesn't need to worry if only 1 entry has actually
6468 been set to non-zero. It also makes the "position beyond the end of the
6469 cache" logic much simpler, as the first slot is always the one to start
6473 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6474 const STRLEN utf8, const STRLEN blen)
6478 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6483 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6484 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6485 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6487 (*mgp)->mg_len = -1;
6491 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6492 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6493 (*mgp)->mg_ptr = (char *) cache;
6497 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6498 /* SvPOKp() because it's possible that sv has string overloading, and
6499 therefore is a reference, hence SvPVX() is actually a pointer.
6500 This cures the (very real) symptoms of RT 69422, but I'm not actually
6501 sure whether we should even be caching the results of UTF-8
6502 operations on overloading, given that nothing stops overloading
6503 returning a different value every time it's called. */
6504 const U8 *start = (const U8 *) SvPVX_const(sv);
6505 const STRLEN realutf8 = utf8_length(start, start + byte);
6507 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6511 /* Cache is held with the later position first, to simplify the code
6512 that deals with unbounded ends. */
6514 ASSERT_UTF8_CACHE(cache);
6515 if (cache[1] == 0) {
6516 /* Cache is totally empty */
6519 } else if (cache[3] == 0) {
6520 if (byte > cache[1]) {
6521 /* New one is larger, so goes first. */
6522 cache[2] = cache[0];
6523 cache[3] = cache[1];
6531 #define THREEWAY_SQUARE(a,b,c,d) \
6532 ((float)((d) - (c))) * ((float)((d) - (c))) \
6533 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6534 + ((float)((b) - (a))) * ((float)((b) - (a)))
6536 /* Cache has 2 slots in use, and we know three potential pairs.
6537 Keep the two that give the lowest RMS distance. Do the
6538 calcualation in bytes simply because we always know the byte
6539 length. squareroot has the same ordering as the positive value,
6540 so don't bother with the actual square root. */
6541 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6542 if (byte > cache[1]) {
6543 /* New position is after the existing pair of pairs. */
6544 const float keep_earlier
6545 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6546 const float keep_later
6547 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6549 if (keep_later < keep_earlier) {
6550 if (keep_later < existing) {
6551 cache[2] = cache[0];
6552 cache[3] = cache[1];
6558 if (keep_earlier < existing) {
6564 else if (byte > cache[3]) {
6565 /* New position is between the existing pair of pairs. */
6566 const float keep_earlier
6567 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6568 const float keep_later
6569 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6571 if (keep_later < keep_earlier) {
6572 if (keep_later < existing) {
6578 if (keep_earlier < existing) {
6585 /* New position is before the existing pair of pairs. */
6586 const float keep_earlier
6587 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6588 const float keep_later
6589 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6591 if (keep_later < keep_earlier) {
6592 if (keep_later < existing) {
6598 if (keep_earlier < existing) {
6599 cache[0] = cache[2];
6600 cache[1] = cache[3];
6607 ASSERT_UTF8_CACHE(cache);
6610 /* We already know all of the way, now we may be able to walk back. The same
6611 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6612 backward is half the speed of walking forward. */
6614 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6615 const U8 *end, STRLEN endu)
6617 const STRLEN forw = target - s;
6618 STRLEN backw = end - target;
6620 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6622 if (forw < 2 * backw) {
6623 return utf8_length(s, target);
6626 while (end > target) {
6628 while (UTF8_IS_CONTINUATION(*end)) {
6637 =for apidoc sv_pos_b2u
6639 Converts the value pointed to by offsetp from a count of bytes from the
6640 start of the string, to a count of the equivalent number of UTF-8 chars.
6641 Handles magic and type coercion.
6647 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6648 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6653 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6656 const STRLEN byte = *offsetp;
6657 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6663 PERL_ARGS_ASSERT_SV_POS_B2U;
6668 s = (const U8*)SvPV_const(sv, blen);
6671 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
6677 && SvTYPE(sv) >= SVt_PVMG
6678 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
6681 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
6682 if (cache[1] == byte) {
6683 /* An exact match. */
6684 *offsetp = cache[0];
6687 if (cache[3] == byte) {
6688 /* An exact match. */
6689 *offsetp = cache[2];
6693 if (cache[1] < byte) {
6694 /* We already know part of the way. */
6695 if (mg->mg_len != -1) {
6696 /* Actually, we know the end too. */
6698 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
6699 s + blen, mg->mg_len - cache[0]);
6701 len = cache[0] + utf8_length(s + cache[1], send);
6704 else if (cache[3] < byte) {
6705 /* We're between the two cached pairs, so we do the calculation
6706 offset by the byte/utf-8 positions for the earlier pair,
6707 then add the utf-8 characters from the string start to
6709 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
6710 s + cache[1], cache[0] - cache[2])
6714 else { /* cache[3] > byte */
6715 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
6719 ASSERT_UTF8_CACHE(cache);
6721 } else if (mg->mg_len != -1) {
6722 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
6726 if (!found || PL_utf8cache < 0) {
6727 const STRLEN real_len = utf8_length(s, send);
6729 if (found && PL_utf8cache < 0)
6730 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
6737 utf8_mg_len_cache_update(sv, &mg, len);
6739 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
6744 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
6745 STRLEN real, SV *const sv)
6747 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
6749 /* As this is debugging only code, save space by keeping this test here,
6750 rather than inlining it in all the callers. */
6751 if (from_cache == real)
6754 /* Need to turn the assertions off otherwise we may recurse infinitely
6755 while printing error messages. */
6756 SAVEI8(PL_utf8cache);
6758 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
6759 func, (UV) from_cache, (UV) real, SVfARG(sv));
6765 Returns a boolean indicating whether the strings in the two SVs are
6766 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6767 coerce its args to strings if necessary.
6773 Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
6782 SV* svrecode = NULL;
6789 /* if pv1 and pv2 are the same, second SvPV_const call may
6790 * invalidate pv1, so we may need to make a copy */
6791 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
6792 pv1 = SvPV_const(sv1, cur1);
6793 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
6795 pv1 = SvPV_const(sv1, cur1);
6803 pv2 = SvPV_const(sv2, cur2);
6805 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6806 /* Differing utf8ness.
6807 * Do not UTF8size the comparands as a side-effect. */
6810 svrecode = newSVpvn(pv2, cur2);
6811 sv_recode_to_utf8(svrecode, PL_encoding);
6812 pv2 = SvPV_const(svrecode, cur2);
6815 svrecode = newSVpvn(pv1, cur1);
6816 sv_recode_to_utf8(svrecode, PL_encoding);
6817 pv1 = SvPV_const(svrecode, cur1);
6819 /* Now both are in UTF-8. */
6821 SvREFCNT_dec(svrecode);
6826 bool is_utf8 = TRUE;
6829 /* sv1 is the UTF-8 one,
6830 * if is equal it must be downgrade-able */
6831 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
6837 /* sv2 is the UTF-8 one,
6838 * if is equal it must be downgrade-able */
6839 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
6845 /* Downgrade not possible - cannot be eq */
6853 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
6855 SvREFCNT_dec(svrecode);
6865 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6866 string in C<sv1> is less than, equal to, or greater than the string in
6867 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6868 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
6874 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
6878 const char *pv1, *pv2;
6881 SV *svrecode = NULL;
6888 pv1 = SvPV_const(sv1, cur1);
6895 pv2 = SvPV_const(sv2, cur2);
6897 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6898 /* Differing utf8ness.
6899 * Do not UTF8size the comparands as a side-effect. */
6902 svrecode = newSVpvn(pv2, cur2);
6903 sv_recode_to_utf8(svrecode, PL_encoding);
6904 pv2 = SvPV_const(svrecode, cur2);
6907 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
6912 svrecode = newSVpvn(pv1, cur1);
6913 sv_recode_to_utf8(svrecode, PL_encoding);
6914 pv1 = SvPV_const(svrecode, cur1);
6917 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
6923 cmp = cur2 ? -1 : 0;
6927 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
6930 cmp = retval < 0 ? -1 : 1;
6931 } else if (cur1 == cur2) {
6934 cmp = cur1 < cur2 ? -1 : 1;
6938 SvREFCNT_dec(svrecode);
6946 =for apidoc sv_cmp_locale
6948 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6949 'use bytes' aware, handles get magic, and will coerce its args to strings
6950 if necessary. See also C<sv_cmp>.
6956 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
6959 #ifdef USE_LOCALE_COLLATE
6965 if (PL_collation_standard)
6969 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
6971 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
6973 if (!pv1 || !len1) {
6984 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
6987 return retval < 0 ? -1 : 1;
6990 * When the result of collation is equality, that doesn't mean
6991 * that there are no differences -- some locales exclude some
6992 * characters from consideration. So to avoid false equalities,
6993 * we use the raw string as a tiebreaker.
6999 #endif /* USE_LOCALE_COLLATE */
7001 return sv_cmp(sv1, sv2);
7005 #ifdef USE_LOCALE_COLLATE
7008 =for apidoc sv_collxfrm
7010 Add Collate Transform magic to an SV if it doesn't already have it.
7012 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7013 scalar data of the variable, but transformed to such a format that a normal
7014 memory comparison can be used to compare the data according to the locale
7021 Perl_sv_collxfrm(pTHX_ SV *const sv, STRLEN *const nxp)
7026 PERL_ARGS_ASSERT_SV_COLLXFRM;
7028 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7029 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7035 Safefree(mg->mg_ptr);
7036 s = SvPV_const(sv, len);
7037 if ((xf = mem_collxfrm(s, len, &xlen))) {
7039 #ifdef PERL_OLD_COPY_ON_WRITE
7041 sv_force_normal_flags(sv, 0);
7043 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7057 if (mg && mg->mg_ptr) {
7059 return mg->mg_ptr + sizeof(PL_collation_ix);
7067 #endif /* USE_LOCALE_COLLATE */
7072 Get a line from the filehandle and store it into the SV, optionally
7073 appending to the currently-stored string.
7079 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7084 register STDCHAR rslast;
7085 register STDCHAR *bp;
7090 PERL_ARGS_ASSERT_SV_GETS;
7092 if (SvTHINKFIRST(sv))
7093 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7094 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7096 However, perlbench says it's slower, because the existing swipe code
7097 is faster than copy on write.
7098 Swings and roundabouts. */
7099 SvUPGRADE(sv, SVt_PV);
7104 if (PerlIO_isutf8(fp)) {
7106 sv_utf8_upgrade_nomg(sv);
7107 sv_pos_u2b(sv,&append,0);
7109 } else if (SvUTF8(sv)) {
7110 SV * const tsv = newSV(0);
7111 sv_gets(tsv, fp, 0);
7112 sv_utf8_upgrade_nomg(tsv);
7113 SvCUR_set(sv,append);
7116 goto return_string_or_null;
7124 if (PerlIO_isutf8(fp))
7127 if (IN_PERL_COMPILETIME) {
7128 /* we always read code in line mode */
7132 else if (RsSNARF(PL_rs)) {
7133 /* If it is a regular disk file use size from stat() as estimate
7134 of amount we are going to read -- may result in mallocing
7135 more memory than we really need if the layers below reduce
7136 the size we read (e.g. CRLF or a gzip layer).
7139 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7140 const Off_t offset = PerlIO_tell(fp);
7141 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7142 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7148 else if (RsRECORD(PL_rs)) {
7156 /* Grab the size of the record we're getting */
7157 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7158 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7161 /* VMS wants read instead of fread, because fread doesn't respect */
7162 /* RMS record boundaries. This is not necessarily a good thing to be */
7163 /* doing, but we've got no other real choice - except avoid stdio
7164 as implementation - perhaps write a :vms layer ?
7166 fd = PerlIO_fileno(fp);
7167 if (fd == -1) { /* in-memory file from PerlIO::Scalar */
7168 bytesread = PerlIO_read(fp, buffer, recsize);
7171 bytesread = PerlLIO_read(fd, buffer, recsize);
7174 bytesread = PerlIO_read(fp, buffer, recsize);
7178 SvCUR_set(sv, bytesread + append);
7179 buffer[bytesread] = '\0';
7180 goto return_string_or_null;
7182 else if (RsPARA(PL_rs)) {
7188 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7189 if (PerlIO_isutf8(fp)) {
7190 rsptr = SvPVutf8(PL_rs, rslen);
7193 if (SvUTF8(PL_rs)) {
7194 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7195 Perl_croak(aTHX_ "Wide character in $/");
7198 rsptr = SvPV_const(PL_rs, rslen);
7202 rslast = rslen ? rsptr[rslen - 1] : '\0';
7204 if (rspara) { /* have to do this both before and after */
7205 do { /* to make sure file boundaries work right */
7208 i = PerlIO_getc(fp);
7212 PerlIO_ungetc(fp,i);
7218 /* See if we know enough about I/O mechanism to cheat it ! */
7220 /* This used to be #ifdef test - it is made run-time test for ease
7221 of abstracting out stdio interface. One call should be cheap
7222 enough here - and may even be a macro allowing compile
7226 if (PerlIO_fast_gets(fp)) {
7229 * We're going to steal some values from the stdio struct
7230 * and put EVERYTHING in the innermost loop into registers.
7232 register STDCHAR *ptr;
7236 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7237 /* An ungetc()d char is handled separately from the regular
7238 * buffer, so we getc() it back out and stuff it in the buffer.
7240 i = PerlIO_getc(fp);
7241 if (i == EOF) return 0;
7242 *(--((*fp)->_ptr)) = (unsigned char) i;
7246 /* Here is some breathtakingly efficient cheating */
7248 cnt = PerlIO_get_cnt(fp); /* get count into register */
7249 /* make sure we have the room */
7250 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7251 /* Not room for all of it
7252 if we are looking for a separator and room for some
7254 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7255 /* just process what we have room for */
7256 shortbuffered = cnt - SvLEN(sv) + append + 1;
7257 cnt -= shortbuffered;
7261 /* remember that cnt can be negative */
7262 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7267 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7268 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7269 DEBUG_P(PerlIO_printf(Perl_debug_log,
7270 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7271 DEBUG_P(PerlIO_printf(Perl_debug_log,
7272 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7273 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7274 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7279 while (cnt > 0) { /* this | eat */
7281 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7282 goto thats_all_folks; /* screams | sed :-) */
7286 Copy(ptr, bp, cnt, char); /* this | eat */
7287 bp += cnt; /* screams | dust */
7288 ptr += cnt; /* louder | sed :-) */
7293 if (shortbuffered) { /* oh well, must extend */
7294 cnt = shortbuffered;
7296 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7298 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7299 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7303 DEBUG_P(PerlIO_printf(Perl_debug_log,
7304 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7305 PTR2UV(ptr),(long)cnt));
7306 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7308 DEBUG_P(PerlIO_printf(Perl_debug_log,
7309 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7310 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7311 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7313 /* This used to call 'filbuf' in stdio form, but as that behaves like
7314 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7315 another abstraction. */
7316 i = PerlIO_getc(fp); /* get more characters */
7318 DEBUG_P(PerlIO_printf(Perl_debug_log,
7319 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7320 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7321 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7323 cnt = PerlIO_get_cnt(fp);
7324 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7325 DEBUG_P(PerlIO_printf(Perl_debug_log,
7326 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7328 if (i == EOF) /* all done for ever? */
7329 goto thats_really_all_folks;
7331 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7333 SvGROW(sv, bpx + cnt + 2);
7334 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7336 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7338 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7339 goto thats_all_folks;
7343 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7344 memNE((char*)bp - rslen, rsptr, rslen))
7345 goto screamer; /* go back to the fray */
7346 thats_really_all_folks:
7348 cnt += shortbuffered;
7349 DEBUG_P(PerlIO_printf(Perl_debug_log,
7350 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7351 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7352 DEBUG_P(PerlIO_printf(Perl_debug_log,
7353 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7354 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7355 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7357 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7358 DEBUG_P(PerlIO_printf(Perl_debug_log,
7359 "Screamer: done, len=%ld, string=|%.*s|\n",
7360 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7364 /*The big, slow, and stupid way. */
7365 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7366 STDCHAR *buf = NULL;
7367 Newx(buf, 8192, STDCHAR);
7375 register const STDCHAR * const bpe = buf + sizeof(buf);
7377 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7378 ; /* keep reading */
7382 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7383 /* Accomodate broken VAXC compiler, which applies U8 cast to
7384 * both args of ?: operator, causing EOF to change into 255
7387 i = (U8)buf[cnt - 1];
7393 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7395 sv_catpvn(sv, (char *) buf, cnt);
7397 sv_setpvn(sv, (char *) buf, cnt);
7399 if (i != EOF && /* joy */
7401 SvCUR(sv) < rslen ||
7402 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7406 * If we're reading from a TTY and we get a short read,
7407 * indicating that the user hit his EOF character, we need
7408 * to notice it now, because if we try to read from the TTY
7409 * again, the EOF condition will disappear.
7411 * The comparison of cnt to sizeof(buf) is an optimization
7412 * that prevents unnecessary calls to feof().
7416 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7420 #ifdef USE_HEAP_INSTEAD_OF_STACK
7425 if (rspara) { /* have to do this both before and after */
7426 while (i != EOF) { /* to make sure file boundaries work right */
7427 i = PerlIO_getc(fp);
7429 PerlIO_ungetc(fp,i);
7435 return_string_or_null:
7436 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7442 Auto-increment of the value in the SV, doing string to numeric conversion
7443 if necessary. Handles 'get' magic and operator overloading.
7449 Perl_sv_inc(pTHX_ register SV *const sv)
7458 =for apidoc sv_inc_nomg
7460 Auto-increment of the value in the SV, doing string to numeric conversion
7461 if necessary. Handles operator overloading. Skips handling 'get' magic.
7467 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7475 if (SvTHINKFIRST(sv)) {
7477 sv_force_normal_flags(sv, 0);
7478 if (SvREADONLY(sv)) {
7479 if (IN_PERL_RUNTIME)
7480 Perl_croak_no_modify(aTHX);
7484 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
7486 i = PTR2IV(SvRV(sv));
7491 flags = SvFLAGS(sv);
7492 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7493 /* It's (privately or publicly) a float, but not tested as an
7494 integer, so test it to see. */
7496 flags = SvFLAGS(sv);
7498 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7499 /* It's publicly an integer, or privately an integer-not-float */
7500 #ifdef PERL_PRESERVE_IVUV
7504 if (SvUVX(sv) == UV_MAX)
7505 sv_setnv(sv, UV_MAX_P1);
7507 (void)SvIOK_only_UV(sv);
7508 SvUV_set(sv, SvUVX(sv) + 1);
7510 if (SvIVX(sv) == IV_MAX)
7511 sv_setuv(sv, (UV)IV_MAX + 1);
7513 (void)SvIOK_only(sv);
7514 SvIV_set(sv, SvIVX(sv) + 1);
7519 if (flags & SVp_NOK) {
7520 const NV was = SvNVX(sv);
7521 if (NV_OVERFLOWS_INTEGERS_AT &&
7522 was >= NV_OVERFLOWS_INTEGERS_AT) {
7523 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7524 "Lost precision when incrementing %" NVff " by 1",
7527 (void)SvNOK_only(sv);
7528 SvNV_set(sv, was + 1.0);
7532 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7533 if ((flags & SVTYPEMASK) < SVt_PVIV)
7534 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7535 (void)SvIOK_only(sv);
7540 while (isALPHA(*d)) d++;
7541 while (isDIGIT(*d)) d++;
7542 if (d < SvEND(sv)) {
7543 #ifdef PERL_PRESERVE_IVUV
7544 /* Got to punt this as an integer if needs be, but we don't issue
7545 warnings. Probably ought to make the sv_iv_please() that does
7546 the conversion if possible, and silently. */
7547 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7548 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7549 /* Need to try really hard to see if it's an integer.
7550 9.22337203685478e+18 is an integer.
7551 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7552 so $a="9.22337203685478e+18"; $a+0; $a++
7553 needs to be the same as $a="9.22337203685478e+18"; $a++
7560 /* sv_2iv *should* have made this an NV */
7561 if (flags & SVp_NOK) {
7562 (void)SvNOK_only(sv);
7563 SvNV_set(sv, SvNVX(sv) + 1.0);
7566 /* I don't think we can get here. Maybe I should assert this
7567 And if we do get here I suspect that sv_setnv will croak. NWC
7569 #if defined(USE_LONG_DOUBLE)
7570 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",
7571 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7573 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7574 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7577 #endif /* PERL_PRESERVE_IVUV */
7578 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7582 while (d >= SvPVX_const(sv)) {
7590 /* MKS: The original code here died if letters weren't consecutive.
7591 * at least it didn't have to worry about non-C locales. The
7592 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7593 * arranged in order (although not consecutively) and that only
7594 * [A-Za-z] are accepted by isALPHA in the C locale.
7596 if (*d != 'z' && *d != 'Z') {
7597 do { ++*d; } while (!isALPHA(*d));
7600 *(d--) -= 'z' - 'a';
7605 *(d--) -= 'z' - 'a' + 1;
7609 /* oh,oh, the number grew */
7610 SvGROW(sv, SvCUR(sv) + 2);
7611 SvCUR_set(sv, SvCUR(sv) + 1);
7612 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7623 Auto-decrement of the value in the SV, doing string to numeric conversion
7624 if necessary. Handles 'get' magic and operator overloading.
7630 Perl_sv_dec(pTHX_ register SV *const sv)
7640 =for apidoc sv_dec_nomg
7642 Auto-decrement of the value in the SV, doing string to numeric conversion
7643 if necessary. Handles operator overloading. Skips handling 'get' magic.
7649 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
7656 if (SvTHINKFIRST(sv)) {
7658 sv_force_normal_flags(sv, 0);
7659 if (SvREADONLY(sv)) {
7660 if (IN_PERL_RUNTIME)
7661 Perl_croak_no_modify(aTHX);
7665 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
7667 i = PTR2IV(SvRV(sv));
7672 /* Unlike sv_inc we don't have to worry about string-never-numbers
7673 and keeping them magic. But we mustn't warn on punting */
7674 flags = SvFLAGS(sv);
7675 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7676 /* It's publicly an integer, or privately an integer-not-float */
7677 #ifdef PERL_PRESERVE_IVUV
7681 if (SvUVX(sv) == 0) {
7682 (void)SvIOK_only(sv);
7686 (void)SvIOK_only_UV(sv);
7687 SvUV_set(sv, SvUVX(sv) - 1);
7690 if (SvIVX(sv) == IV_MIN) {
7691 sv_setnv(sv, (NV)IV_MIN);
7695 (void)SvIOK_only(sv);
7696 SvIV_set(sv, SvIVX(sv) - 1);
7701 if (flags & SVp_NOK) {
7704 const NV was = SvNVX(sv);
7705 if (NV_OVERFLOWS_INTEGERS_AT &&
7706 was <= -NV_OVERFLOWS_INTEGERS_AT) {
7707 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7708 "Lost precision when decrementing %" NVff " by 1",
7711 (void)SvNOK_only(sv);
7712 SvNV_set(sv, was - 1.0);
7716 if (!(flags & SVp_POK)) {
7717 if ((flags & SVTYPEMASK) < SVt_PVIV)
7718 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
7720 (void)SvIOK_only(sv);
7723 #ifdef PERL_PRESERVE_IVUV
7725 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7726 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7727 /* Need to try really hard to see if it's an integer.
7728 9.22337203685478e+18 is an integer.
7729 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7730 so $a="9.22337203685478e+18"; $a+0; $a--
7731 needs to be the same as $a="9.22337203685478e+18"; $a--
7738 /* sv_2iv *should* have made this an NV */
7739 if (flags & SVp_NOK) {
7740 (void)SvNOK_only(sv);
7741 SvNV_set(sv, SvNVX(sv) - 1.0);
7744 /* I don't think we can get here. Maybe I should assert this
7745 And if we do get here I suspect that sv_setnv will croak. NWC
7747 #if defined(USE_LONG_DOUBLE)
7748 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",
7749 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7751 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7752 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7756 #endif /* PERL_PRESERVE_IVUV */
7757 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
7760 /* this define is used to eliminate a chunk of duplicated but shared logic
7761 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
7762 * used anywhere but here - yves
7764 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
7767 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
7771 =for apidoc sv_mortalcopy
7773 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
7774 The new SV is marked as mortal. It will be destroyed "soon", either by an
7775 explicit call to FREETMPS, or by an implicit call at places such as
7776 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
7781 /* Make a string that will exist for the duration of the expression
7782 * evaluation. Actually, it may have to last longer than that, but
7783 * hopefully we won't free it until it has been assigned to a
7784 * permanent location. */
7787 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
7793 sv_setsv(sv,oldstr);
7794 PUSH_EXTEND_MORTAL__SV_C(sv);
7800 =for apidoc sv_newmortal
7802 Creates a new null SV which is mortal. The reference count of the SV is
7803 set to 1. It will be destroyed "soon", either by an explicit call to
7804 FREETMPS, or by an implicit call at places such as statement boundaries.
7805 See also C<sv_mortalcopy> and C<sv_2mortal>.
7811 Perl_sv_newmortal(pTHX)
7817 SvFLAGS(sv) = SVs_TEMP;
7818 PUSH_EXTEND_MORTAL__SV_C(sv);
7824 =for apidoc newSVpvn_flags
7826 Creates a new SV and copies a string into it. The reference count for the
7827 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7828 string. You are responsible for ensuring that the source string is at least
7829 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7830 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
7831 If C<SVs_TEMP> is set, then C<sv2mortal()> is called on the result before
7832 returning. If C<SVf_UTF8> is set, C<s> is considered to be in UTF-8 and the
7833 C<SVf_UTF8> flag will be set on the new SV.
7834 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
7836 #define newSVpvn_utf8(s, len, u) \
7837 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
7843 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
7848 /* All the flags we don't support must be zero.
7849 And we're new code so I'm going to assert this from the start. */
7850 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
7852 sv_setpvn(sv,s,len);
7854 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
7855 * and do what it does outselves here.
7856 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
7857 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
7858 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
7859 * eleminate quite a few steps than it looks - Yves (explaining patch by gfx)
7862 SvFLAGS(sv) |= flags;
7864 if(flags & SVs_TEMP){
7865 PUSH_EXTEND_MORTAL__SV_C(sv);
7872 =for apidoc sv_2mortal
7874 Marks an existing SV as mortal. The SV will be destroyed "soon", either
7875 by an explicit call to FREETMPS, or by an implicit call at places such as
7876 statement boundaries. SvTEMP() is turned on which means that the SV's
7877 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
7878 and C<sv_mortalcopy>.
7884 Perl_sv_2mortal(pTHX_ register SV *const sv)
7889 if (SvREADONLY(sv) && SvIMMORTAL(sv))
7891 PUSH_EXTEND_MORTAL__SV_C(sv);
7899 Creates a new SV and copies a string into it. The reference count for the
7900 SV is set to 1. If C<len> is zero, Perl will compute the length using
7901 strlen(). For efficiency, consider using C<newSVpvn> instead.
7907 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
7913 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
7918 =for apidoc newSVpvn
7920 Creates a new SV and copies a string into it. The reference count for the
7921 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7922 string. You are responsible for ensuring that the source string is at least
7923 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7929 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
7935 sv_setpvn(sv,s,len);
7940 =for apidoc newSVhek
7942 Creates a new SV from the hash key structure. It will generate scalars that
7943 point to the shared string table where possible. Returns a new (undefined)
7944 SV if the hek is NULL.
7950 Perl_newSVhek(pTHX_ const HEK *const hek)
7960 if (HEK_LEN(hek) == HEf_SVKEY) {
7961 return newSVsv(*(SV**)HEK_KEY(hek));
7963 const int flags = HEK_FLAGS(hek);
7964 if (flags & HVhek_WASUTF8) {
7966 Andreas would like keys he put in as utf8 to come back as utf8
7968 STRLEN utf8_len = HEK_LEN(hek);
7969 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
7970 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
7973 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
7975 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
7976 /* We don't have a pointer to the hv, so we have to replicate the
7977 flag into every HEK. This hv is using custom a hasing
7978 algorithm. Hence we can't return a shared string scalar, as
7979 that would contain the (wrong) hash value, and might get passed
7980 into an hv routine with a regular hash.
7981 Similarly, a hash that isn't using shared hash keys has to have
7982 the flag in every key so that we know not to try to call
7983 share_hek_kek on it. */
7985 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
7990 /* This will be overwhelminly the most common case. */
7992 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
7993 more efficient than sharepvn(). */
7997 sv_upgrade(sv, SVt_PV);
7998 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7999 SvCUR_set(sv, HEK_LEN(hek));
8012 =for apidoc newSVpvn_share
8014 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8015 table. If the string does not already exist in the table, it is created
8016 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
8017 value is used; otherwise the hash is computed. The string's hash can be later
8018 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
8019 that as the string table is used for shared hash keys these strings will have
8020 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8026 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8030 bool is_utf8 = FALSE;
8031 const char *const orig_src = src;
8034 STRLEN tmplen = -len;
8036 /* See the note in hv.c:hv_fetch() --jhi */
8037 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8041 PERL_HASH(hash, src, len);
8043 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8044 changes here, update it there too. */
8045 sv_upgrade(sv, SVt_PV);
8046 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8054 if (src != orig_src)
8060 #if defined(PERL_IMPLICIT_CONTEXT)
8062 /* pTHX_ magic can't cope with varargs, so this is a no-context
8063 * version of the main function, (which may itself be aliased to us).
8064 * Don't access this version directly.
8068 Perl_newSVpvf_nocontext(const char *const pat, ...)
8074 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8076 va_start(args, pat);
8077 sv = vnewSVpvf(pat, &args);
8084 =for apidoc newSVpvf
8086 Creates a new SV and initializes it with the string formatted like
8093 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8098 PERL_ARGS_ASSERT_NEWSVPVF;
8100 va_start(args, pat);
8101 sv = vnewSVpvf(pat, &args);
8106 /* backend for newSVpvf() and newSVpvf_nocontext() */
8109 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8114 PERL_ARGS_ASSERT_VNEWSVPVF;
8117 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8124 Creates a new SV and copies a floating point value into it.
8125 The reference count for the SV is set to 1.
8131 Perl_newSVnv(pTHX_ const NV n)
8144 Creates a new SV and copies an integer into it. The reference count for the
8151 Perl_newSViv(pTHX_ const IV i)
8164 Creates a new SV and copies an unsigned integer into it.
8165 The reference count for the SV is set to 1.
8171 Perl_newSVuv(pTHX_ const UV u)
8182 =for apidoc newSV_type
8184 Creates a new SV, of the type specified. The reference count for the new SV
8191 Perl_newSV_type(pTHX_ const svtype type)
8196 sv_upgrade(sv, type);
8201 =for apidoc newRV_noinc
8203 Creates an RV wrapper for an SV. The reference count for the original
8204 SV is B<not> incremented.
8210 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8213 register SV *sv = newSV_type(SVt_IV);
8215 PERL_ARGS_ASSERT_NEWRV_NOINC;
8218 SvRV_set(sv, tmpRef);
8223 /* newRV_inc is the official function name to use now.
8224 * newRV_inc is in fact #defined to newRV in sv.h
8228 Perl_newRV(pTHX_ SV *const sv)
8232 PERL_ARGS_ASSERT_NEWRV;
8234 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8240 Creates a new SV which is an exact duplicate of the original SV.
8247 Perl_newSVsv(pTHX_ register SV *const old)
8254 if (SvTYPE(old) == SVTYPEMASK) {
8255 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8259 /* SV_GMAGIC is the default for sv_setv()
8260 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8261 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8262 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8267 =for apidoc sv_reset
8269 Underlying implementation for the C<reset> Perl function.
8270 Note that the perl-level function is vaguely deprecated.
8276 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8279 char todo[PERL_UCHAR_MAX+1];
8281 PERL_ARGS_ASSERT_SV_RESET;
8286 if (!*s) { /* reset ?? searches */
8287 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8289 const U32 count = mg->mg_len / sizeof(PMOP**);
8290 PMOP **pmp = (PMOP**) mg->mg_ptr;
8291 PMOP *const *const end = pmp + count;
8295 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8297 (*pmp)->op_pmflags &= ~PMf_USED;
8305 /* reset variables */
8307 if (!HvARRAY(stash))
8310 Zero(todo, 256, char);
8313 I32 i = (unsigned char)*s;
8317 max = (unsigned char)*s++;
8318 for ( ; i <= max; i++) {
8321 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8323 for (entry = HvARRAY(stash)[i];
8325 entry = HeNEXT(entry))
8330 if (!todo[(U8)*HeKEY(entry)])
8332 gv = MUTABLE_GV(HeVAL(entry));
8335 if (SvTHINKFIRST(sv)) {
8336 if (!SvREADONLY(sv) && SvROK(sv))
8338 /* XXX Is this continue a bug? Why should THINKFIRST
8339 exempt us from resetting arrays and hashes? */
8343 if (SvTYPE(sv) >= SVt_PV) {
8345 if (SvPVX_const(sv) != NULL)
8353 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8355 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8358 # if defined(USE_ENVIRON_ARRAY)
8361 # endif /* USE_ENVIRON_ARRAY */
8372 Using various gambits, try to get an IO from an SV: the IO slot if its a
8373 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8374 named after the PV if we're a string.
8380 Perl_sv_2io(pTHX_ SV *const sv)
8385 PERL_ARGS_ASSERT_SV_2IO;
8387 switch (SvTYPE(sv)) {
8389 io = MUTABLE_IO(sv);
8392 if (isGV_with_GP(sv)) {
8393 gv = MUTABLE_GV(sv);
8396 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8402 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8404 return sv_2io(SvRV(sv));
8405 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8411 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8420 Using various gambits, try to get a CV from an SV; in addition, try if
8421 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8422 The flags in C<lref> are passed to gv_fetchsv.
8428 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8434 PERL_ARGS_ASSERT_SV_2CV;
8441 switch (SvTYPE(sv)) {
8445 return MUTABLE_CV(sv);
8452 if (isGV_with_GP(sv)) {
8453 gv = MUTABLE_GV(sv);
8462 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
8464 tryAMAGICunDEREF(to_cv);
8467 if (SvTYPE(sv) == SVt_PVCV) {
8468 cv = MUTABLE_CV(sv);
8473 else if(isGV_with_GP(sv))
8474 gv = MUTABLE_GV(sv);
8476 Perl_croak(aTHX_ "Not a subroutine reference");
8478 else if (isGV_with_GP(sv)) {
8480 gv = MUTABLE_GV(sv);
8483 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8489 /* Some flags to gv_fetchsv mean don't really create the GV */
8490 if (!isGV_with_GP(gv)) {
8496 if (lref && !GvCVu(gv)) {
8500 gv_efullname3(tmpsv, gv, NULL);
8501 /* XXX this is probably not what they think they're getting.
8502 * It has the same effect as "sub name;", i.e. just a forward
8504 newSUB(start_subparse(FALSE, 0),
8505 newSVOP(OP_CONST, 0, tmpsv),
8509 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8510 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8519 Returns true if the SV has a true value by Perl's rules.
8520 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8521 instead use an in-line version.
8527 Perl_sv_true(pTHX_ register SV *const sv)
8532 register const XPV* const tXpv = (XPV*)SvANY(sv);
8534 (tXpv->xpv_cur > 1 ||
8535 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8542 return SvIVX(sv) != 0;
8545 return SvNVX(sv) != 0.0;
8547 return sv_2bool(sv);
8553 =for apidoc sv_pvn_force
8555 Get a sensible string out of the SV somehow.
8556 A private implementation of the C<SvPV_force> macro for compilers which
8557 can't cope with complex macro expressions. Always use the macro instead.
8559 =for apidoc sv_pvn_force_flags
8561 Get a sensible string out of the SV somehow.
8562 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8563 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8564 implemented in terms of this function.
8565 You normally want to use the various wrapper macros instead: see
8566 C<SvPV_force> and C<SvPV_force_nomg>
8572 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8576 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8578 if (SvTHINKFIRST(sv) && !SvROK(sv))
8579 sv_force_normal_flags(sv, 0);
8589 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8590 const char * const ref = sv_reftype(sv,0);
8592 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8593 ref, OP_DESC(PL_op));
8595 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8597 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8598 || isGV_with_GP(sv))
8599 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8601 s = sv_2pv_flags(sv, &len, flags);
8605 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8608 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8609 SvGROW(sv, len + 1);
8610 Move(s,SvPVX(sv),len,char);
8612 SvPVX(sv)[len] = '\0';
8615 SvPOK_on(sv); /* validate pointer */
8617 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
8618 PTR2UV(sv),SvPVX_const(sv)));
8621 return SvPVX_mutable(sv);
8625 =for apidoc sv_pvbyten_force
8627 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
8633 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
8635 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
8637 sv_pvn_force(sv,lp);
8638 sv_utf8_downgrade(sv,0);
8644 =for apidoc sv_pvutf8n_force
8646 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
8652 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
8654 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
8656 sv_pvn_force(sv,lp);
8657 sv_utf8_upgrade(sv);
8663 =for apidoc sv_reftype
8665 Returns a string describing what the SV is a reference to.
8671 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
8673 PERL_ARGS_ASSERT_SV_REFTYPE;
8675 /* The fact that I don't need to downcast to char * everywhere, only in ?:
8676 inside return suggests a const propagation bug in g++. */
8677 if (ob && SvOBJECT(sv)) {
8678 char * const name = HvNAME_get(SvSTASH(sv));
8679 return name ? name : (char *) "__ANON__";
8682 switch (SvTYPE(sv)) {
8697 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
8698 /* tied lvalues should appear to be
8699 * scalars for backwards compatitbility */
8700 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
8701 ? "SCALAR" : "LVALUE");
8702 case SVt_PVAV: return "ARRAY";
8703 case SVt_PVHV: return "HASH";
8704 case SVt_PVCV: return "CODE";
8705 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
8706 ? "GLOB" : "SCALAR");
8707 case SVt_PVFM: return "FORMAT";
8708 case SVt_PVIO: return "IO";
8709 case SVt_BIND: return "BIND";
8710 case SVt_REGEXP: return "REGEXP";
8711 default: return "UNKNOWN";
8717 =for apidoc sv_isobject
8719 Returns a boolean indicating whether the SV is an RV pointing to a blessed
8720 object. If the SV is not an RV, or if the object is not blessed, then this
8727 Perl_sv_isobject(pTHX_ SV *sv)
8743 Returns a boolean indicating whether the SV is blessed into the specified
8744 class. This does not check for subtypes; use C<sv_derived_from> to verify
8745 an inheritance relationship.
8751 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
8755 PERL_ARGS_ASSERT_SV_ISA;
8765 hvname = HvNAME_get(SvSTASH(sv));
8769 return strEQ(hvname, name);
8775 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
8776 it will be upgraded to one. If C<classname> is non-null then the new SV will
8777 be blessed in the specified package. The new SV is returned and its
8778 reference count is 1.
8784 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
8789 PERL_ARGS_ASSERT_NEWSVRV;
8793 SV_CHECK_THINKFIRST_COW_DROP(rv);
8794 (void)SvAMAGIC_off(rv);
8796 if (SvTYPE(rv) >= SVt_PVMG) {
8797 const U32 refcnt = SvREFCNT(rv);
8801 SvREFCNT(rv) = refcnt;
8803 sv_upgrade(rv, SVt_IV);
8804 } else if (SvROK(rv)) {
8805 SvREFCNT_dec(SvRV(rv));
8807 prepare_SV_for_RV(rv);
8815 HV* const stash = gv_stashpv(classname, GV_ADD);
8816 (void)sv_bless(rv, stash);
8822 =for apidoc sv_setref_pv
8824 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
8825 argument will be upgraded to an RV. That RV will be modified to point to
8826 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
8827 into the SV. The C<classname> argument indicates the package for the
8828 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8829 will have a reference count of 1, and the RV will be returned.
8831 Do not use with other Perl types such as HV, AV, SV, CV, because those
8832 objects will become corrupted by the pointer copy process.
8834 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
8840 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
8844 PERL_ARGS_ASSERT_SV_SETREF_PV;
8847 sv_setsv(rv, &PL_sv_undef);
8851 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
8856 =for apidoc sv_setref_iv
8858 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
8859 argument will be upgraded to an RV. That RV will be modified to point to
8860 the new SV. The C<classname> argument indicates the package for the
8861 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8862 will have a reference count of 1, and the RV will be returned.
8868 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
8870 PERL_ARGS_ASSERT_SV_SETREF_IV;
8872 sv_setiv(newSVrv(rv,classname), iv);
8877 =for apidoc sv_setref_uv
8879 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
8880 argument will be upgraded to an RV. That RV will be modified to point to
8881 the new SV. The C<classname> argument indicates the package for the
8882 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8883 will have a reference count of 1, and the RV will be returned.
8889 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
8891 PERL_ARGS_ASSERT_SV_SETREF_UV;
8893 sv_setuv(newSVrv(rv,classname), uv);
8898 =for apidoc sv_setref_nv
8900 Copies a double into a new SV, optionally blessing the SV. The C<rv>
8901 argument will be upgraded to an RV. That RV will be modified to point to
8902 the new SV. The C<classname> argument indicates the package for the
8903 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8904 will have a reference count of 1, and the RV will be returned.
8910 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
8912 PERL_ARGS_ASSERT_SV_SETREF_NV;
8914 sv_setnv(newSVrv(rv,classname), nv);
8919 =for apidoc sv_setref_pvn
8921 Copies a string into a new SV, optionally blessing the SV. The length of the
8922 string must be specified with C<n>. The C<rv> argument will be upgraded to
8923 an RV. That RV will be modified to point to the new SV. The C<classname>
8924 argument indicates the package for the blessing. Set C<classname> to
8925 C<NULL> to avoid the blessing. The new SV will have a reference count
8926 of 1, and the RV will be returned.
8928 Note that C<sv_setref_pv> copies the pointer while this copies the string.
8934 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
8935 const char *const pv, const STRLEN n)
8937 PERL_ARGS_ASSERT_SV_SETREF_PVN;
8939 sv_setpvn(newSVrv(rv,classname), pv, n);
8944 =for apidoc sv_bless
8946 Blesses an SV into a specified package. The SV must be an RV. The package
8947 must be designated by its stash (see C<gv_stashpv()>). The reference count
8948 of the SV is unaffected.
8954 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
8959 PERL_ARGS_ASSERT_SV_BLESS;
8962 Perl_croak(aTHX_ "Can't bless non-reference value");
8964 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
8965 if (SvIsCOW(tmpRef))
8966 sv_force_normal_flags(tmpRef, 0);
8967 if (SvREADONLY(tmpRef))
8968 Perl_croak_no_modify(aTHX);
8969 if (SvOBJECT(tmpRef)) {
8970 if (SvTYPE(tmpRef) != SVt_PVIO)
8972 SvREFCNT_dec(SvSTASH(tmpRef));
8975 SvOBJECT_on(tmpRef);
8976 if (SvTYPE(tmpRef) != SVt_PVIO)
8978 SvUPGRADE(tmpRef, SVt_PVMG);
8979 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
8984 (void)SvAMAGIC_off(sv);
8986 if(SvSMAGICAL(tmpRef))
8987 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
8995 /* Downgrades a PVGV to a PVMG.
8999 S_sv_unglob(pTHX_ SV *const sv)
9004 SV * const temp = sv_newmortal();
9006 PERL_ARGS_ASSERT_SV_UNGLOB;
9008 assert(SvTYPE(sv) == SVt_PVGV);
9010 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9013 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9014 && HvNAME_get(stash))
9015 mro_method_changed_in(stash);
9016 gp_free(MUTABLE_GV(sv));
9019 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9023 if (GvNAME_HEK(sv)) {
9024 unshare_hek(GvNAME_HEK(sv));
9026 isGV_with_GP_off(sv);
9028 /* need to keep SvANY(sv) in the right arena */
9029 xpvmg = new_XPVMG();
9030 StructCopy(SvANY(sv), xpvmg, XPVMG);
9031 del_XPVGV(SvANY(sv));
9034 SvFLAGS(sv) &= ~SVTYPEMASK;
9035 SvFLAGS(sv) |= SVt_PVMG;
9037 /* Intentionally not calling any local SET magic, as this isn't so much a
9038 set operation as merely an internal storage change. */
9039 sv_setsv_flags(sv, temp, 0);
9043 =for apidoc sv_unref_flags
9045 Unsets the RV status of the SV, and decrements the reference count of
9046 whatever was being referenced by the RV. This can almost be thought of
9047 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9048 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9049 (otherwise the decrementing is conditional on the reference count being
9050 different from one or the reference being a readonly SV).
9057 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9059 SV* const target = SvRV(ref);
9061 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9063 if (SvWEAKREF(ref)) {
9064 sv_del_backref(target, ref);
9066 SvRV_set(ref, NULL);
9069 SvRV_set(ref, NULL);
9071 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9072 assigned to as BEGIN {$a = \"Foo"} will fail. */
9073 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9074 SvREFCNT_dec(target);
9075 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9076 sv_2mortal(target); /* Schedule for freeing later */
9080 =for apidoc sv_untaint
9082 Untaint an SV. Use C<SvTAINTED_off> instead.
9087 Perl_sv_untaint(pTHX_ SV *const sv)
9089 PERL_ARGS_ASSERT_SV_UNTAINT;
9091 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9092 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9099 =for apidoc sv_tainted
9101 Test an SV for taintedness. Use C<SvTAINTED> instead.
9106 Perl_sv_tainted(pTHX_ SV *const sv)
9108 PERL_ARGS_ASSERT_SV_TAINTED;
9110 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9111 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9112 if (mg && (mg->mg_len & 1) )
9119 =for apidoc sv_setpviv
9121 Copies an integer into the given SV, also updating its string value.
9122 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9128 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9130 char buf[TYPE_CHARS(UV)];
9132 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9134 PERL_ARGS_ASSERT_SV_SETPVIV;
9136 sv_setpvn(sv, ptr, ebuf - ptr);
9140 =for apidoc sv_setpviv_mg
9142 Like C<sv_setpviv>, but also handles 'set' magic.
9148 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9150 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9156 #if defined(PERL_IMPLICIT_CONTEXT)
9158 /* pTHX_ magic can't cope with varargs, so this is a no-context
9159 * version of the main function, (which may itself be aliased to us).
9160 * Don't access this version directly.
9164 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9169 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9171 va_start(args, pat);
9172 sv_vsetpvf(sv, pat, &args);
9176 /* pTHX_ magic can't cope with varargs, so this is a no-context
9177 * version of the main function, (which may itself be aliased to us).
9178 * Don't access this version directly.
9182 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9187 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9189 va_start(args, pat);
9190 sv_vsetpvf_mg(sv, pat, &args);
9196 =for apidoc sv_setpvf
9198 Works like C<sv_catpvf> but copies the text into the SV instead of
9199 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9205 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9209 PERL_ARGS_ASSERT_SV_SETPVF;
9211 va_start(args, pat);
9212 sv_vsetpvf(sv, pat, &args);
9217 =for apidoc sv_vsetpvf
9219 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9220 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9222 Usually used via its frontend C<sv_setpvf>.
9228 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9230 PERL_ARGS_ASSERT_SV_VSETPVF;
9232 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9236 =for apidoc sv_setpvf_mg
9238 Like C<sv_setpvf>, but also handles 'set' magic.
9244 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9248 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9250 va_start(args, pat);
9251 sv_vsetpvf_mg(sv, pat, &args);
9256 =for apidoc sv_vsetpvf_mg
9258 Like C<sv_vsetpvf>, but also handles 'set' magic.
9260 Usually used via its frontend C<sv_setpvf_mg>.
9266 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9268 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9270 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9274 #if defined(PERL_IMPLICIT_CONTEXT)
9276 /* pTHX_ magic can't cope with varargs, so this is a no-context
9277 * version of the main function, (which may itself be aliased to us).
9278 * Don't access this version directly.
9282 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9287 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9289 va_start(args, pat);
9290 sv_vcatpvf(sv, pat, &args);
9294 /* pTHX_ magic can't cope with varargs, so this is a no-context
9295 * version of the main function, (which may itself be aliased to us).
9296 * Don't access this version directly.
9300 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9305 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9307 va_start(args, pat);
9308 sv_vcatpvf_mg(sv, pat, &args);
9314 =for apidoc sv_catpvf
9316 Processes its arguments like C<sprintf> and appends the formatted
9317 output to an SV. If the appended data contains "wide" characters
9318 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9319 and characters >255 formatted with %c), the original SV might get
9320 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9321 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9322 valid UTF-8; if the original SV was bytes, the pattern should be too.
9327 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9331 PERL_ARGS_ASSERT_SV_CATPVF;
9333 va_start(args, pat);
9334 sv_vcatpvf(sv, pat, &args);
9339 =for apidoc sv_vcatpvf
9341 Processes its arguments like C<vsprintf> and appends the formatted output
9342 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9344 Usually used via its frontend C<sv_catpvf>.
9350 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9352 PERL_ARGS_ASSERT_SV_VCATPVF;
9354 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9358 =for apidoc sv_catpvf_mg
9360 Like C<sv_catpvf>, but also handles 'set' magic.
9366 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9370 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9372 va_start(args, pat);
9373 sv_vcatpvf_mg(sv, pat, &args);
9378 =for apidoc sv_vcatpvf_mg
9380 Like C<sv_vcatpvf>, but also handles 'set' magic.
9382 Usually used via its frontend C<sv_catpvf_mg>.
9388 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9390 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9392 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9397 =for apidoc sv_vsetpvfn
9399 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9402 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9408 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9409 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9411 PERL_ARGS_ASSERT_SV_VSETPVFN;
9414 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9419 * Warn of missing argument to sprintf, and then return a defined value
9420 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9422 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9424 S_vcatpvfn_missing_argument(pTHX) {
9425 if (ckWARN(WARN_MISSING)) {
9426 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9427 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9434 S_expect_number(pTHX_ char **const pattern)
9439 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9441 switch (**pattern) {
9442 case '1': case '2': case '3':
9443 case '4': case '5': case '6':
9444 case '7': case '8': case '9':
9445 var = *(*pattern)++ - '0';
9446 while (isDIGIT(**pattern)) {
9447 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9449 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9457 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9459 const int neg = nv < 0;
9462 PERL_ARGS_ASSERT_F0CONVERT;
9470 if (uv & 1 && uv == nv)
9471 uv--; /* Round to even */
9473 const unsigned dig = uv % 10;
9486 =for apidoc sv_vcatpvfn
9488 Processes its arguments like C<vsprintf> and appends the formatted output
9489 to an SV. Uses an array of SVs if the C style variable argument list is
9490 missing (NULL). When running with taint checks enabled, indicates via
9491 C<maybe_tainted> if results are untrustworthy (often due to the use of
9494 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9500 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9501 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9502 vec_utf8 = DO_UTF8(vecsv);
9504 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9507 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9508 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9516 static const char nullstr[] = "(null)";
9518 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9519 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9521 /* Times 4: a decimal digit takes more than 3 binary digits.
9522 * NV_DIG: mantissa takes than many decimal digits.
9523 * Plus 32: Playing safe. */
9524 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9525 /* large enough for "%#.#f" --chip */
9526 /* what about long double NVs? --jhi */
9528 PERL_ARGS_ASSERT_SV_VCATPVFN;
9529 PERL_UNUSED_ARG(maybe_tainted);
9531 /* no matter what, this is a string now */
9532 (void)SvPV_force(sv, origlen);
9534 /* special-case "", "%s", and "%-p" (SVf - see below) */
9537 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9539 const char * const s = va_arg(*args, char*);
9540 sv_catpv(sv, s ? s : nullstr);
9542 else if (svix < svmax) {
9543 sv_catsv(sv, *svargs);
9546 S_vcatpvfn_missing_argument(aTHX);
9549 if (args && patlen == 3 && pat[0] == '%' &&
9550 pat[1] == '-' && pat[2] == 'p') {
9551 argsv = MUTABLE_SV(va_arg(*args, void*));
9552 sv_catsv(sv, argsv);
9556 #ifndef USE_LONG_DOUBLE
9557 /* special-case "%.<number>[gf]" */
9558 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9559 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9560 unsigned digits = 0;
9564 while (*pp >= '0' && *pp <= '9')
9565 digits = 10 * digits + (*pp++ - '0');
9566 if (pp - pat == (int)patlen - 1 && svix < svmax) {
9567 const NV nv = SvNV(*svargs);
9569 /* Add check for digits != 0 because it seems that some
9570 gconverts are buggy in this case, and we don't yet have
9571 a Configure test for this. */
9572 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9573 /* 0, point, slack */
9574 Gconvert(nv, (int)digits, 0, ebuf);
9576 if (*ebuf) /* May return an empty string for digits==0 */
9579 } else if (!digits) {
9582 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9583 sv_catpvn(sv, p, l);
9589 #endif /* !USE_LONG_DOUBLE */
9591 if (!args && svix < svmax && DO_UTF8(*svargs))
9594 patend = (char*)pat + patlen;
9595 for (p = (char*)pat; p < patend; p = q) {
9598 bool vectorize = FALSE;
9599 bool vectorarg = FALSE;
9600 bool vec_utf8 = FALSE;
9606 bool has_precis = FALSE;
9608 const I32 osvix = svix;
9609 bool is_utf8 = FALSE; /* is this item utf8? */
9610 #ifdef HAS_LDBL_SPRINTF_BUG
9611 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
9612 with sfio - Allen <allens@cpan.org> */
9613 bool fix_ldbl_sprintf_bug = FALSE;
9617 U8 utf8buf[UTF8_MAXBYTES+1];
9618 STRLEN esignlen = 0;
9620 const char *eptr = NULL;
9621 const char *fmtstart;
9624 const U8 *vecstr = NULL;
9631 /* we need a long double target in case HAS_LONG_DOUBLE but
9634 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9642 const char *dotstr = ".";
9643 STRLEN dotstrlen = 1;
9644 I32 efix = 0; /* explicit format parameter index */
9645 I32 ewix = 0; /* explicit width index */
9646 I32 epix = 0; /* explicit precision index */
9647 I32 evix = 0; /* explicit vector index */
9648 bool asterisk = FALSE;
9650 /* echo everything up to the next format specification */
9651 for (q = p; q < patend && *q != '%'; ++q) ;
9653 if (has_utf8 && !pat_utf8)
9654 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
9656 sv_catpvn(sv, p, q - p);
9665 We allow format specification elements in this order:
9666 \d+\$ explicit format parameter index
9668 v|\*(\d+\$)?v vector with optional (optionally specified) arg
9669 0 flag (as above): repeated to allow "v02"
9670 \d+|\*(\d+\$)? width using optional (optionally specified) arg
9671 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
9673 [%bcdefginopsuxDFOUX] format (mandatory)
9678 As of perl5.9.3, printf format checking is on by default.
9679 Internally, perl uses %p formats to provide an escape to
9680 some extended formatting. This block deals with those
9681 extensions: if it does not match, (char*)q is reset and
9682 the normal format processing code is used.
9684 Currently defined extensions are:
9685 %p include pointer address (standard)
9686 %-p (SVf) include an SV (previously %_)
9687 %-<num>p include an SV with precision <num>
9688 %<num>p reserved for future extensions
9690 Robin Barker 2005-07-14
9692 %1p (VDf) removed. RMB 2007-10-19
9699 n = expect_number(&q);
9706 argsv = MUTABLE_SV(va_arg(*args, void*));
9707 eptr = SvPV_const(argsv, elen);
9713 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
9714 "internal %%<num>p might conflict with future printf extensions");
9720 if ( (width = expect_number(&q)) ) {
9735 if (plus == '+' && *q == ' ') /* '+' over ' ' */
9764 if ( (ewix = expect_number(&q)) )
9773 if ((vectorarg = asterisk)) {
9786 width = expect_number(&q);
9792 vecsv = va_arg(*args, SV*);
9794 vecsv = (evix > 0 && evix <= svmax)
9795 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
9797 vecsv = svix < svmax
9798 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
9800 dotstr = SvPV_const(vecsv, dotstrlen);
9801 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
9802 bad with tied or overloaded values that return UTF8. */
9805 else if (has_utf8) {
9806 vecsv = sv_mortalcopy(vecsv);
9807 sv_utf8_upgrade(vecsv);
9808 dotstr = SvPV_const(vecsv, dotstrlen);
9815 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
9816 vecsv = svargs[efix ? efix-1 : svix++];
9817 vecstr = (U8*)SvPV_const(vecsv,veclen);
9818 vec_utf8 = DO_UTF8(vecsv);
9820 /* if this is a version object, we need to convert
9821 * back into v-string notation and then let the
9822 * vectorize happen normally
9824 if (sv_derived_from(vecsv, "version")) {
9825 char *version = savesvpv(vecsv);
9826 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
9827 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9828 "vector argument not supported with alpha versions");
9831 vecsv = sv_newmortal();
9832 scan_vstring(version, version + veclen, vecsv);
9833 vecstr = (U8*)SvPV_const(vecsv, veclen);
9834 vec_utf8 = DO_UTF8(vecsv);
9846 i = va_arg(*args, int);
9848 i = (ewix ? ewix <= svmax : svix < svmax) ?
9849 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9851 width = (i < 0) ? -i : i;
9861 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
9863 /* XXX: todo, support specified precision parameter */
9867 i = va_arg(*args, int);
9869 i = (ewix ? ewix <= svmax : svix < svmax)
9870 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9872 has_precis = !(i < 0);
9877 precis = precis * 10 + (*q++ - '0');
9886 case 'I': /* Ix, I32x, and I64x */
9888 if (q[1] == '6' && q[2] == '4') {
9894 if (q[1] == '3' && q[2] == '2') {
9904 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9915 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9916 if (*(q + 1) == 'l') { /* lld, llf */
9942 if (!vectorize && !args) {
9944 const I32 i = efix-1;
9945 argsv = (i >= 0 && i < svmax)
9946 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
9948 argsv = (svix >= 0 && svix < svmax)
9949 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
9960 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
9962 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
9964 eptr = (char*)utf8buf;
9965 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
9979 eptr = va_arg(*args, char*);
9981 elen = strlen(eptr);
9983 eptr = (char *)nullstr;
9984 elen = sizeof nullstr - 1;
9988 eptr = SvPV_const(argsv, elen);
9989 if (DO_UTF8(argsv)) {
9990 STRLEN old_precis = precis;
9991 if (has_precis && precis < elen) {
9992 STRLEN ulen = sv_len_utf8(argsv);
9993 I32 p = precis > ulen ? ulen : precis;
9994 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
9997 if (width) { /* fudge width (can't fudge elen) */
9998 if (has_precis && precis < elen)
9999 width += precis - old_precis;
10001 width += elen - sv_len_utf8(argsv);
10008 if (has_precis && precis < elen)
10015 if (alt || vectorize)
10017 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10038 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10047 esignbuf[esignlen++] = plus;
10051 case 'h': iv = (short)va_arg(*args, int); break;
10052 case 'l': iv = va_arg(*args, long); break;
10053 case 'V': iv = va_arg(*args, IV); break;
10054 default: iv = va_arg(*args, int); break;
10057 iv = va_arg(*args, Quad_t); break;
10064 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10066 case 'h': iv = (short)tiv; break;
10067 case 'l': iv = (long)tiv; break;
10069 default: iv = tiv; break;
10072 iv = (Quad_t)tiv; break;
10078 if ( !vectorize ) /* we already set uv above */
10083 esignbuf[esignlen++] = plus;
10087 esignbuf[esignlen++] = '-';
10131 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10142 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10143 case 'l': uv = va_arg(*args, unsigned long); break;
10144 case 'V': uv = va_arg(*args, UV); break;
10145 default: uv = va_arg(*args, unsigned); break;
10148 uv = va_arg(*args, Uquad_t); break;
10155 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10157 case 'h': uv = (unsigned short)tuv; break;
10158 case 'l': uv = (unsigned long)tuv; break;
10160 default: uv = tuv; break;
10163 uv = (Uquad_t)tuv; break;
10172 char *ptr = ebuf + sizeof ebuf;
10173 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10179 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10183 } while (uv >>= 4);
10185 esignbuf[esignlen++] = '0';
10186 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10192 *--ptr = '0' + dig;
10193 } while (uv >>= 3);
10194 if (alt && *ptr != '0')
10200 *--ptr = '0' + dig;
10201 } while (uv >>= 1);
10203 esignbuf[esignlen++] = '0';
10204 esignbuf[esignlen++] = c;
10207 default: /* it had better be ten or less */
10210 *--ptr = '0' + dig;
10211 } while (uv /= base);
10214 elen = (ebuf + sizeof ebuf) - ptr;
10218 zeros = precis - elen;
10219 else if (precis == 0 && elen == 1 && *eptr == '0'
10220 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10223 /* a precision nullifies the 0 flag. */
10230 /* FLOATING POINT */
10233 c = 'f'; /* maybe %F isn't supported here */
10235 case 'e': case 'E':
10237 case 'g': case 'G':
10241 /* This is evil, but floating point is even more evil */
10243 /* for SV-style calling, we can only get NV
10244 for C-style calling, we assume %f is double;
10245 for simplicity we allow any of %Lf, %llf, %qf for long double
10249 #if defined(USE_LONG_DOUBLE)
10253 /* [perl #20339] - we should accept and ignore %lf rather than die */
10257 #if defined(USE_LONG_DOUBLE)
10258 intsize = args ? 0 : 'q';
10262 #if defined(HAS_LONG_DOUBLE)
10271 /* now we need (long double) if intsize == 'q', else (double) */
10273 #if LONG_DOUBLESIZE > DOUBLESIZE
10275 va_arg(*args, long double) :
10276 va_arg(*args, double)
10278 va_arg(*args, double)
10283 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10284 else. frexp() has some unspecified behaviour for those three */
10285 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10287 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10288 will cast our (long double) to (double) */
10289 (void)Perl_frexp(nv, &i);
10290 if (i == PERL_INT_MIN)
10291 Perl_die(aTHX_ "panic: frexp");
10293 need = BIT_DIGITS(i);
10295 need += has_precis ? precis : 6; /* known default */
10300 #ifdef HAS_LDBL_SPRINTF_BUG
10301 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10302 with sfio - Allen <allens@cpan.org> */
10305 # define MY_DBL_MAX DBL_MAX
10306 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10307 # if DOUBLESIZE >= 8
10308 # define MY_DBL_MAX 1.7976931348623157E+308L
10310 # define MY_DBL_MAX 3.40282347E+38L
10314 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10315 # define MY_DBL_MAX_BUG 1L
10317 # define MY_DBL_MAX_BUG MY_DBL_MAX
10321 # define MY_DBL_MIN DBL_MIN
10322 # else /* XXX guessing! -Allen */
10323 # if DOUBLESIZE >= 8
10324 # define MY_DBL_MIN 2.2250738585072014E-308L
10326 # define MY_DBL_MIN 1.17549435E-38L
10330 if ((intsize == 'q') && (c == 'f') &&
10331 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10332 (need < DBL_DIG)) {
10333 /* it's going to be short enough that
10334 * long double precision is not needed */
10336 if ((nv <= 0L) && (nv >= -0L))
10337 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10339 /* would use Perl_fp_class as a double-check but not
10340 * functional on IRIX - see perl.h comments */
10342 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10343 /* It's within the range that a double can represent */
10344 #if defined(DBL_MAX) && !defined(DBL_MIN)
10345 if ((nv >= ((long double)1/DBL_MAX)) ||
10346 (nv <= (-(long double)1/DBL_MAX)))
10348 fix_ldbl_sprintf_bug = TRUE;
10351 if (fix_ldbl_sprintf_bug == TRUE) {
10361 # undef MY_DBL_MAX_BUG
10364 #endif /* HAS_LDBL_SPRINTF_BUG */
10366 need += 20; /* fudge factor */
10367 if (PL_efloatsize < need) {
10368 Safefree(PL_efloatbuf);
10369 PL_efloatsize = need + 20; /* more fudge */
10370 Newx(PL_efloatbuf, PL_efloatsize, char);
10371 PL_efloatbuf[0] = '\0';
10374 if ( !(width || left || plus || alt) && fill != '0'
10375 && has_precis && intsize != 'q' ) { /* Shortcuts */
10376 /* See earlier comment about buggy Gconvert when digits,
10378 if ( c == 'g' && precis) {
10379 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10380 /* May return an empty string for digits==0 */
10381 if (*PL_efloatbuf) {
10382 elen = strlen(PL_efloatbuf);
10383 goto float_converted;
10385 } else if ( c == 'f' && !precis) {
10386 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10391 char *ptr = ebuf + sizeof ebuf;
10394 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10395 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10396 if (intsize == 'q') {
10397 /* Copy the one or more characters in a long double
10398 * format before the 'base' ([efgEFG]) character to
10399 * the format string. */
10400 static char const prifldbl[] = PERL_PRIfldbl;
10401 char const *p = prifldbl + sizeof(prifldbl) - 3;
10402 while (p >= prifldbl) { *--ptr = *p--; }
10407 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10412 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10424 /* No taint. Otherwise we are in the strange situation
10425 * where printf() taints but print($float) doesn't.
10427 #if defined(HAS_LONG_DOUBLE)
10428 elen = ((intsize == 'q')
10429 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10430 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10432 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10436 eptr = PL_efloatbuf;
10444 i = SvCUR(sv) - origlen;
10447 case 'h': *(va_arg(*args, short*)) = i; break;
10448 default: *(va_arg(*args, int*)) = i; break;
10449 case 'l': *(va_arg(*args, long*)) = i; break;
10450 case 'V': *(va_arg(*args, IV*)) = i; break;
10453 *(va_arg(*args, Quad_t*)) = i; break;
10460 sv_setuv_mg(argsv, (UV)i);
10461 continue; /* not "break" */
10468 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10469 && ckWARN(WARN_PRINTF))
10471 SV * const msg = sv_newmortal();
10472 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10473 (PL_op->op_type == OP_PRTF) ? "" : "s");
10474 if (fmtstart < patend) {
10475 const char * const fmtend = q < patend ? q : patend;
10477 sv_catpvs(msg, "\"%");
10478 for (f = fmtstart; f < fmtend; f++) {
10480 sv_catpvn(msg, f, 1);
10482 Perl_sv_catpvf(aTHX_ msg,
10483 "\\%03"UVof, (UV)*f & 0xFF);
10486 sv_catpvs(msg, "\"");
10488 sv_catpvs(msg, "end of string");
10490 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10493 /* output mangled stuff ... */
10499 /* ... right here, because formatting flags should not apply */
10500 SvGROW(sv, SvCUR(sv) + elen + 1);
10502 Copy(eptr, p, elen, char);
10505 SvCUR_set(sv, p - SvPVX_const(sv));
10507 continue; /* not "break" */
10510 if (is_utf8 != has_utf8) {
10513 sv_utf8_upgrade(sv);
10516 const STRLEN old_elen = elen;
10517 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10518 sv_utf8_upgrade(nsv);
10519 eptr = SvPVX_const(nsv);
10522 if (width) { /* fudge width (can't fudge elen) */
10523 width += elen - old_elen;
10529 have = esignlen + zeros + elen;
10531 Perl_croak_nocontext("%s", PL_memory_wrap);
10533 need = (have > width ? have : width);
10536 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10537 Perl_croak_nocontext("%s", PL_memory_wrap);
10538 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10540 if (esignlen && fill == '0') {
10542 for (i = 0; i < (int)esignlen; i++)
10543 *p++ = esignbuf[i];
10545 if (gap && !left) {
10546 memset(p, fill, gap);
10549 if (esignlen && fill != '0') {
10551 for (i = 0; i < (int)esignlen; i++)
10552 *p++ = esignbuf[i];
10556 for (i = zeros; i; i--)
10560 Copy(eptr, p, elen, char);
10564 memset(p, ' ', gap);
10569 Copy(dotstr, p, dotstrlen, char);
10573 vectorize = FALSE; /* done iterating over vecstr */
10580 SvCUR_set(sv, p - SvPVX_const(sv));
10589 /* =========================================================================
10591 =head1 Cloning an interpreter
10593 All the macros and functions in this section are for the private use of
10594 the main function, perl_clone().
10596 The foo_dup() functions make an exact copy of an existing foo thingy.
10597 During the course of a cloning, a hash table is used to map old addresses
10598 to new addresses. The table is created and manipulated with the
10599 ptr_table_* functions.
10603 * =========================================================================*/
10606 #if defined(USE_ITHREADS)
10608 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
10609 #ifndef GpREFCNT_inc
10610 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
10614 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
10615 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
10616 If this changes, please unmerge ss_dup.
10617 Likewise, sv_dup_inc_multiple() relies on this fact. */
10618 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
10619 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
10620 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
10621 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
10622 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
10623 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
10624 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
10625 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
10626 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
10627 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
10628 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
10629 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
10630 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
10632 /* clone a parser */
10635 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
10639 PERL_ARGS_ASSERT_PARSER_DUP;
10644 /* look for it in the table first */
10645 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
10649 /* create anew and remember what it is */
10650 Newxz(parser, 1, yy_parser);
10651 ptr_table_store(PL_ptr_table, proto, parser);
10653 parser->yyerrstatus = 0;
10654 parser->yychar = YYEMPTY; /* Cause a token to be read. */
10656 /* XXX these not yet duped */
10657 parser->old_parser = NULL;
10658 parser->stack = NULL;
10660 parser->stack_size = 0;
10661 /* XXX parser->stack->state = 0; */
10663 /* XXX eventually, just Copy() most of the parser struct ? */
10665 parser->lex_brackets = proto->lex_brackets;
10666 parser->lex_casemods = proto->lex_casemods;
10667 parser->lex_brackstack = savepvn(proto->lex_brackstack,
10668 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
10669 parser->lex_casestack = savepvn(proto->lex_casestack,
10670 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
10671 parser->lex_defer = proto->lex_defer;
10672 parser->lex_dojoin = proto->lex_dojoin;
10673 parser->lex_expect = proto->lex_expect;
10674 parser->lex_formbrack = proto->lex_formbrack;
10675 parser->lex_inpat = proto->lex_inpat;
10676 parser->lex_inwhat = proto->lex_inwhat;
10677 parser->lex_op = proto->lex_op;
10678 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
10679 parser->lex_starts = proto->lex_starts;
10680 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
10681 parser->multi_close = proto->multi_close;
10682 parser->multi_open = proto->multi_open;
10683 parser->multi_start = proto->multi_start;
10684 parser->multi_end = proto->multi_end;
10685 parser->pending_ident = proto->pending_ident;
10686 parser->preambled = proto->preambled;
10687 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
10688 parser->linestr = sv_dup_inc(proto->linestr, param);
10689 parser->expect = proto->expect;
10690 parser->copline = proto->copline;
10691 parser->last_lop_op = proto->last_lop_op;
10692 parser->lex_state = proto->lex_state;
10693 parser->rsfp = fp_dup(proto->rsfp, '<', param);
10694 /* rsfp_filters entries have fake IoDIRP() */
10695 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
10696 parser->in_my = proto->in_my;
10697 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
10698 parser->error_count = proto->error_count;
10701 parser->linestr = sv_dup_inc(proto->linestr, param);
10704 char * const ols = SvPVX(proto->linestr);
10705 char * const ls = SvPVX(parser->linestr);
10707 parser->bufptr = ls + (proto->bufptr >= ols ?
10708 proto->bufptr - ols : 0);
10709 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
10710 proto->oldbufptr - ols : 0);
10711 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
10712 proto->oldoldbufptr - ols : 0);
10713 parser->linestart = ls + (proto->linestart >= ols ?
10714 proto->linestart - ols : 0);
10715 parser->last_uni = ls + (proto->last_uni >= ols ?
10716 proto->last_uni - ols : 0);
10717 parser->last_lop = ls + (proto->last_lop >= ols ?
10718 proto->last_lop - ols : 0);
10720 parser->bufend = ls + SvCUR(parser->linestr);
10723 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
10727 parser->endwhite = proto->endwhite;
10728 parser->faketokens = proto->faketokens;
10729 parser->lasttoke = proto->lasttoke;
10730 parser->nextwhite = proto->nextwhite;
10731 parser->realtokenstart = proto->realtokenstart;
10732 parser->skipwhite = proto->skipwhite;
10733 parser->thisclose = proto->thisclose;
10734 parser->thismad = proto->thismad;
10735 parser->thisopen = proto->thisopen;
10736 parser->thisstuff = proto->thisstuff;
10737 parser->thistoken = proto->thistoken;
10738 parser->thiswhite = proto->thiswhite;
10740 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
10741 parser->curforce = proto->curforce;
10743 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
10744 Copy(proto->nexttype, parser->nexttype, 5, I32);
10745 parser->nexttoke = proto->nexttoke;
10748 /* XXX should clone saved_curcop here, but we aren't passed
10749 * proto_perl; so do it in perl_clone_using instead */
10755 /* duplicate a file handle */
10758 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
10762 PERL_ARGS_ASSERT_FP_DUP;
10763 PERL_UNUSED_ARG(type);
10766 return (PerlIO*)NULL;
10768 /* look for it in the table first */
10769 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
10773 /* create anew and remember what it is */
10774 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
10775 ptr_table_store(PL_ptr_table, fp, ret);
10779 /* duplicate a directory handle */
10782 Perl_dirp_dup(pTHX_ DIR *const dp)
10784 PERL_UNUSED_CONTEXT;
10791 /* duplicate a typeglob */
10794 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
10798 PERL_ARGS_ASSERT_GP_DUP;
10802 /* look for it in the table first */
10803 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
10807 /* create anew and remember what it is */
10809 ptr_table_store(PL_ptr_table, gp, ret);
10812 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
10813 on Newxz() to do this for us. */
10814 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
10815 ret->gp_io = io_dup_inc(gp->gp_io, param);
10816 ret->gp_form = cv_dup_inc(gp->gp_form, param);
10817 ret->gp_av = av_dup_inc(gp->gp_av, param);
10818 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
10819 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
10820 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
10821 ret->gp_cvgen = gp->gp_cvgen;
10822 ret->gp_line = gp->gp_line;
10823 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
10827 /* duplicate a chain of magic */
10830 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
10832 MAGIC *mgret = NULL;
10833 MAGIC **mgprev_p = &mgret;
10835 PERL_ARGS_ASSERT_MG_DUP;
10837 for (; mg; mg = mg->mg_moremagic) {
10840 if ((param->flags & CLONEf_JOIN_IN)
10841 && mg->mg_type == PERL_MAGIC_backref)
10842 /* when joining, we let the individual SVs add themselves to
10843 * backref as needed. */
10846 Newx(nmg, 1, MAGIC);
10848 mgprev_p = &(nmg->mg_moremagic);
10850 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
10851 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
10852 from the original commit adding Perl_mg_dup() - revision 4538.
10853 Similarly there is the annotation "XXX random ptr?" next to the
10854 assignment to nmg->mg_ptr. */
10857 /* FIXME for plugins
10858 if (nmg->mg_type == PERL_MAGIC_qr) {
10859 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
10863 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
10864 ? nmg->mg_type == PERL_MAGIC_backref
10865 /* The backref AV has its reference
10866 * count deliberately bumped by 1 */
10867 ? SvREFCNT_inc(av_dup_inc((const AV *)
10868 nmg->mg_obj, param))
10869 : sv_dup_inc(nmg->mg_obj, param)
10870 : sv_dup(nmg->mg_obj, param);
10872 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
10873 if (nmg->mg_len > 0) {
10874 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
10875 if (nmg->mg_type == PERL_MAGIC_overload_table &&
10876 AMT_AMAGIC((AMT*)nmg->mg_ptr))
10878 AMT * const namtp = (AMT*)nmg->mg_ptr;
10879 sv_dup_inc_multiple((SV**)(namtp->table),
10880 (SV**)(namtp->table), NofAMmeth, param);
10883 else if (nmg->mg_len == HEf_SVKEY)
10884 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
10886 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
10887 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
10893 #endif /* USE_ITHREADS */
10895 struct ptr_tbl_arena {
10896 struct ptr_tbl_arena *next;
10897 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
10900 /* create a new pointer-mapping table */
10903 Perl_ptr_table_new(pTHX)
10906 PERL_UNUSED_CONTEXT;
10908 Newx(tbl, 1, PTR_TBL_t);
10909 tbl->tbl_max = 511;
10910 tbl->tbl_items = 0;
10911 tbl->tbl_arena = NULL;
10912 tbl->tbl_arena_next = NULL;
10913 tbl->tbl_arena_end = NULL;
10914 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
10918 #define PTR_TABLE_HASH(ptr) \
10919 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
10921 /* map an existing pointer using a table */
10923 STATIC PTR_TBL_ENT_t *
10924 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
10926 PTR_TBL_ENT_t *tblent;
10927 const UV hash = PTR_TABLE_HASH(sv);
10929 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
10931 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
10932 for (; tblent; tblent = tblent->next) {
10933 if (tblent->oldval == sv)
10940 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
10942 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
10944 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
10945 PERL_UNUSED_CONTEXT;
10947 return tblent ? tblent->newval : NULL;
10950 /* add a new entry to a pointer-mapping table */
10953 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
10955 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
10957 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
10958 PERL_UNUSED_CONTEXT;
10961 tblent->newval = newsv;
10963 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
10965 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
10966 struct ptr_tbl_arena *new_arena;
10968 Newx(new_arena, 1, struct ptr_tbl_arena);
10969 new_arena->next = tbl->tbl_arena;
10970 tbl->tbl_arena = new_arena;
10971 tbl->tbl_arena_next = new_arena->array;
10972 tbl->tbl_arena_end = new_arena->array
10973 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
10976 tblent = tbl->tbl_arena_next++;
10978 tblent->oldval = oldsv;
10979 tblent->newval = newsv;
10980 tblent->next = tbl->tbl_ary[entry];
10981 tbl->tbl_ary[entry] = tblent;
10983 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
10984 ptr_table_split(tbl);
10988 /* double the hash bucket size of an existing ptr table */
10991 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
10993 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
10994 const UV oldsize = tbl->tbl_max + 1;
10995 UV newsize = oldsize * 2;
10998 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
10999 PERL_UNUSED_CONTEXT;
11001 Renew(ary, newsize, PTR_TBL_ENT_t*);
11002 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11003 tbl->tbl_max = --newsize;
11004 tbl->tbl_ary = ary;
11005 for (i=0; i < oldsize; i++, ary++) {
11006 PTR_TBL_ENT_t **entp = ary;
11007 PTR_TBL_ENT_t *ent = *ary;
11008 PTR_TBL_ENT_t **curentp;
11011 curentp = ary + oldsize;
11013 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11015 ent->next = *curentp;
11025 /* remove all the entries from a ptr table */
11026 /* Deprecated - will be removed post 5.14 */
11029 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11031 if (tbl && tbl->tbl_items) {
11032 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11034 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11037 struct ptr_tbl_arena *next = arena->next;
11043 tbl->tbl_items = 0;
11044 tbl->tbl_arena = NULL;
11045 tbl->tbl_arena_next = NULL;
11046 tbl->tbl_arena_end = NULL;
11050 /* clear and free a ptr table */
11053 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11055 struct ptr_tbl_arena *arena;
11061 arena = tbl->tbl_arena;
11064 struct ptr_tbl_arena *next = arena->next;
11070 Safefree(tbl->tbl_ary);
11074 #if defined(USE_ITHREADS)
11077 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11079 PERL_ARGS_ASSERT_RVPV_DUP;
11082 if (SvWEAKREF(sstr)) {
11083 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11084 if (param->flags & CLONEf_JOIN_IN) {
11085 /* if joining, we add any back references individually rather
11086 * than copying the whole backref array */
11087 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11091 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11093 else if (SvPVX_const(sstr)) {
11094 /* Has something there */
11096 /* Normal PV - clone whole allocated space */
11097 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11098 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11099 /* Not that normal - actually sstr is copy on write.
11100 But we are a true, independant SV, so: */
11101 SvREADONLY_off(dstr);
11106 /* Special case - not normally malloced for some reason */
11107 if (isGV_with_GP(sstr)) {
11108 /* Don't need to do anything here. */
11110 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11111 /* A "shared" PV - clone it as "shared" PV */
11113 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11117 /* Some other special case - random pointer */
11118 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11123 /* Copy the NULL */
11124 SvPV_set(dstr, NULL);
11128 /* duplicate a list of SVs. source and dest may point to the same memory. */
11130 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11131 SSize_t items, CLONE_PARAMS *const param)
11133 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11135 while (items-- > 0) {
11136 *dest++ = sv_dup_inc(*source++, param);
11142 /* duplicate an SV of any type (including AV, HV etc) */
11145 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11150 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11152 if (SvTYPE(sstr) == SVTYPEMASK) {
11153 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11158 /* look for it in the table first */
11159 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11163 if(param->flags & CLONEf_JOIN_IN) {
11164 /** We are joining here so we don't want do clone
11165 something that is bad **/
11166 if (SvTYPE(sstr) == SVt_PVHV) {
11167 const HEK * const hvname = HvNAME_HEK(sstr);
11169 /** don't clone stashes if they already exist **/
11170 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
11171 ptr_table_store(PL_ptr_table, sstr, dstr);
11177 /* create anew and remember what it is */
11180 #ifdef DEBUG_LEAKING_SCALARS
11181 dstr->sv_debug_optype = sstr->sv_debug_optype;
11182 dstr->sv_debug_line = sstr->sv_debug_line;
11183 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11184 dstr->sv_debug_parent = (SV*)sstr;
11185 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11188 ptr_table_store(PL_ptr_table, sstr, dstr);
11191 SvFLAGS(dstr) = SvFLAGS(sstr);
11192 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11193 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11196 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11197 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11198 (void*)PL_watch_pvx, SvPVX_const(sstr));
11201 /* don't clone objects whose class has asked us not to */
11202 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11207 switch (SvTYPE(sstr)) {
11209 SvANY(dstr) = NULL;
11212 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11214 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11216 SvIV_set(dstr, SvIVX(sstr));
11220 SvANY(dstr) = new_XNV();
11221 SvNV_set(dstr, SvNVX(sstr));
11223 /* case SVt_BIND: */
11226 /* These are all the types that need complex bodies allocating. */
11228 const svtype sv_type = SvTYPE(sstr);
11229 const struct body_details *const sv_type_details
11230 = bodies_by_type + sv_type;
11234 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11249 assert(sv_type_details->body_size);
11250 if (sv_type_details->arena) {
11251 new_body_inline(new_body, sv_type);
11253 = (void*)((char*)new_body - sv_type_details->offset);
11255 new_body = new_NOARENA(sv_type_details);
11259 SvANY(dstr) = new_body;
11262 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11263 ((char*)SvANY(dstr)) + sv_type_details->offset,
11264 sv_type_details->copy, char);
11266 Copy(((char*)SvANY(sstr)),
11267 ((char*)SvANY(dstr)),
11268 sv_type_details->body_size + sv_type_details->offset, char);
11271 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11272 && !isGV_with_GP(dstr)
11273 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11274 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11276 /* The Copy above means that all the source (unduplicated) pointers
11277 are now in the destination. We can check the flags and the
11278 pointers in either, but it's possible that there's less cache
11279 missing by always going for the destination.
11280 FIXME - instrument and check that assumption */
11281 if (sv_type >= SVt_PVMG) {
11282 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11283 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11284 } else if (SvMAGIC(dstr))
11285 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11287 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11290 /* The cast silences a GCC warning about unhandled types. */
11291 switch ((int)sv_type) {
11301 /* FIXME for plugins */
11302 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11305 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11306 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11307 LvTARG(dstr) = dstr;
11308 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11309 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11311 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11313 if(isGV_with_GP(sstr)) {
11314 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11315 /* Don't call sv_add_backref here as it's going to be
11316 created as part of the magic cloning of the symbol
11317 table--unless this is during a join and the stash
11318 is not actually being cloned. */
11319 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11320 at the point of this comment. */
11321 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11322 if (param->flags & CLONEf_JOIN_IN)
11323 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
11324 GvGP(dstr) = gp_dup(GvGP(sstr), param);
11325 (void)GpREFCNT_inc(GvGP(dstr));
11327 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11330 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11331 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11332 /* I have no idea why fake dirp (rsfps)
11333 should be treated differently but otherwise
11334 we end up with leaks -- sky*/
11335 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11336 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11337 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11339 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11340 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11341 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11342 if (IoDIRP(dstr)) {
11343 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
11346 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11348 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
11350 if (IoOFP(dstr) == IoIFP(sstr))
11351 IoOFP(dstr) = IoIFP(dstr);
11353 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11354 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11355 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11356 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11359 /* avoid cloning an empty array */
11360 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11361 SV **dst_ary, **src_ary;
11362 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11364 src_ary = AvARRAY((const AV *)sstr);
11365 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11366 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11367 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11368 AvALLOC((const AV *)dstr) = dst_ary;
11369 if (AvREAL((const AV *)sstr)) {
11370 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11374 while (items-- > 0)
11375 *dst_ary++ = sv_dup(*src_ary++, param);
11377 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11378 while (items-- > 0) {
11379 *dst_ary++ = &PL_sv_undef;
11383 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11384 AvALLOC((const AV *)dstr) = (SV**)NULL;
11385 AvMAX( (const AV *)dstr) = -1;
11386 AvFILLp((const AV *)dstr) = -1;
11390 if (HvARRAY((const HV *)sstr)) {
11392 const bool sharekeys = !!HvSHAREKEYS(sstr);
11393 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11394 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11396 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11397 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11399 HvARRAY(dstr) = (HE**)darray;
11400 while (i <= sxhv->xhv_max) {
11401 const HE * const source = HvARRAY(sstr)[i];
11402 HvARRAY(dstr)[i] = source
11403 ? he_dup(source, sharekeys, param) : 0;
11408 const struct xpvhv_aux * const saux = HvAUX(sstr);
11409 struct xpvhv_aux * const daux = HvAUX(dstr);
11410 /* This flag isn't copied. */
11411 /* SvOOK_on(hv) attacks the IV flags. */
11412 SvFLAGS(dstr) |= SVf_OOK;
11414 hvname = saux->xhv_name;
11415 daux->xhv_name = hek_dup(hvname, param);
11417 daux->xhv_riter = saux->xhv_riter;
11418 daux->xhv_eiter = saux->xhv_eiter
11419 ? he_dup(saux->xhv_eiter,
11420 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
11421 /* backref array needs refcnt=2; see sv_add_backref */
11422 daux->xhv_backreferences =
11423 (param->flags & CLONEf_JOIN_IN)
11424 /* when joining, we let the individual GVs and
11425 * CVs add themselves to backref as
11426 * needed. This avoids pulling in stuff
11427 * that isn't required, and simplifies the
11428 * case where stashes aren't cloned back
11429 * if they already exist in the parent
11432 : saux->xhv_backreferences
11433 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
11434 ? MUTABLE_AV(SvREFCNT_inc(
11435 sv_dup_inc((const SV *)
11436 saux->xhv_backreferences, param)))
11437 : MUTABLE_AV(sv_dup((const SV *)
11438 saux->xhv_backreferences, param))
11441 daux->xhv_mro_meta = saux->xhv_mro_meta
11442 ? mro_meta_dup(saux->xhv_mro_meta, param)
11445 /* Record stashes for possible cloning in Perl_clone(). */
11447 av_push(param->stashes, dstr);
11451 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11454 if (!(param->flags & CLONEf_COPY_STACKS)) {
11459 /* NOTE: not refcounted */
11460 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
11461 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
11462 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
11464 if (!CvISXSUB(dstr))
11465 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
11467 if (CvCONST(dstr) && CvISXSUB(dstr)) {
11468 CvXSUBANY(dstr).any_ptr =
11469 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
11471 /* don't dup if copying back - CvGV isn't refcounted, so the
11472 * duped GV may never be freed. A bit of a hack! DAPM */
11473 SvANY(MUTABLE_CV(dstr))->xcv_gv =
11475 ? gv_dup_inc(CvGV(sstr), param)
11476 : (param->flags & CLONEf_JOIN_IN)
11478 : gv_dup(CvGV(sstr), param);
11480 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
11482 CvWEAKOUTSIDE(sstr)
11483 ? cv_dup( CvOUTSIDE(dstr), param)
11484 : cv_dup_inc(CvOUTSIDE(dstr), param);
11485 if (!CvISXSUB(dstr))
11486 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
11492 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
11499 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11501 PERL_ARGS_ASSERT_SV_DUP_INC;
11502 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
11506 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11508 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
11509 PERL_ARGS_ASSERT_SV_DUP;
11511 /* Track every SV that (at least initially) had a reference count of 0.
11512 We need to do this by holding an actual reference to it in this array.
11513 If we attempt to cheat, turn AvREAL_off(), and store only pointers
11514 (akin to the stashes hash, and the perl stack), we come unstuck if
11515 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
11516 thread) is manipulated in a CLONE method, because CLONE runs before the
11517 unreferenced array is walked to find SVs still with SvREFCNT() == 0
11518 (and fix things up by giving each a reference via the temps stack).
11519 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
11520 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
11521 before the walk of unreferenced happens and a reference to that is SV
11522 added to the temps stack. At which point we have the same SV considered
11523 to be in use, and free to be re-used. Not good.
11525 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
11526 assert(param->unreferenced);
11527 av_push(param->unreferenced, SvREFCNT_inc(dstr));
11533 /* duplicate a context */
11536 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
11538 PERL_CONTEXT *ncxs;
11540 PERL_ARGS_ASSERT_CX_DUP;
11543 return (PERL_CONTEXT*)NULL;
11545 /* look for it in the table first */
11546 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
11550 /* create anew and remember what it is */
11551 Newx(ncxs, max + 1, PERL_CONTEXT);
11552 ptr_table_store(PL_ptr_table, cxs, ncxs);
11553 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
11556 PERL_CONTEXT * const ncx = &ncxs[ix];
11557 if (CxTYPE(ncx) == CXt_SUBST) {
11558 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
11561 switch (CxTYPE(ncx)) {
11563 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
11564 ? cv_dup_inc(ncx->blk_sub.cv, param)
11565 : cv_dup(ncx->blk_sub.cv,param));
11566 ncx->blk_sub.argarray = (CxHASARGS(ncx)
11567 ? av_dup_inc(ncx->blk_sub.argarray,
11570 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
11572 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
11573 ncx->blk_sub.oldcomppad);
11576 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
11578 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
11580 case CXt_LOOP_LAZYSV:
11581 ncx->blk_loop.state_u.lazysv.end
11582 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
11583 /* We are taking advantage of av_dup_inc and sv_dup_inc
11584 actually being the same function, and order equivalance of
11586 We can assert the later [but only at run time :-(] */
11587 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
11588 (void *) &ncx->blk_loop.state_u.lazysv.cur);
11590 ncx->blk_loop.state_u.ary.ary
11591 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
11592 case CXt_LOOP_LAZYIV:
11593 case CXt_LOOP_PLAIN:
11594 if (CxPADLOOP(ncx)) {
11595 ncx->blk_loop.oldcomppad
11596 = (PAD*)ptr_table_fetch(PL_ptr_table,
11597 ncx->blk_loop.oldcomppad);
11599 ncx->blk_loop.oldcomppad
11600 = (PAD*)gv_dup((const GV *)ncx->blk_loop.oldcomppad,
11605 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
11606 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
11607 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
11620 /* duplicate a stack info structure */
11623 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
11627 PERL_ARGS_ASSERT_SI_DUP;
11630 return (PERL_SI*)NULL;
11632 /* look for it in the table first */
11633 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
11637 /* create anew and remember what it is */
11638 Newxz(nsi, 1, PERL_SI);
11639 ptr_table_store(PL_ptr_table, si, nsi);
11641 nsi->si_stack = av_dup_inc(si->si_stack, param);
11642 nsi->si_cxix = si->si_cxix;
11643 nsi->si_cxmax = si->si_cxmax;
11644 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
11645 nsi->si_type = si->si_type;
11646 nsi->si_prev = si_dup(si->si_prev, param);
11647 nsi->si_next = si_dup(si->si_next, param);
11648 nsi->si_markoff = si->si_markoff;
11653 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
11654 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
11655 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
11656 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
11657 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
11658 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
11659 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
11660 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
11661 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
11662 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
11663 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
11664 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
11665 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
11666 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
11667 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
11668 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
11671 #define pv_dup_inc(p) SAVEPV(p)
11672 #define pv_dup(p) SAVEPV(p)
11673 #define svp_dup_inc(p,pp) any_dup(p,pp)
11675 /* map any object to the new equivent - either something in the
11676 * ptr table, or something in the interpreter structure
11680 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
11684 PERL_ARGS_ASSERT_ANY_DUP;
11687 return (void*)NULL;
11689 /* look for it in the table first */
11690 ret = ptr_table_fetch(PL_ptr_table, v);
11694 /* see if it is part of the interpreter structure */
11695 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
11696 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
11704 /* duplicate the save stack */
11707 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
11710 ANY * const ss = proto_perl->Isavestack;
11711 const I32 max = proto_perl->Isavestack_max;
11712 I32 ix = proto_perl->Isavestack_ix;
11725 void (*dptr) (void*);
11726 void (*dxptr) (pTHX_ void*);
11728 PERL_ARGS_ASSERT_SS_DUP;
11730 Newxz(nss, max, ANY);
11733 const UV uv = POPUV(ss,ix);
11734 const U8 type = (U8)uv & SAVE_MASK;
11736 TOPUV(nss,ix) = uv;
11738 case SAVEt_CLEARSV:
11740 case SAVEt_HELEM: /* hash element */
11741 sv = (const SV *)POPPTR(ss,ix);
11742 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11744 case SAVEt_ITEM: /* normal string */
11745 case SAVEt_SV: /* scalar reference */
11746 sv = (const SV *)POPPTR(ss,ix);
11747 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11750 case SAVEt_MORTALIZESV:
11751 sv = (const SV *)POPPTR(ss,ix);
11752 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11754 case SAVEt_SHARED_PVREF: /* char* in shared space */
11755 c = (char*)POPPTR(ss,ix);
11756 TOPPTR(nss,ix) = savesharedpv(c);
11757 ptr = POPPTR(ss,ix);
11758 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11760 case SAVEt_GENERIC_SVREF: /* generic sv */
11761 case SAVEt_SVREF: /* scalar reference */
11762 sv = (const SV *)POPPTR(ss,ix);
11763 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11764 ptr = POPPTR(ss,ix);
11765 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
11767 case SAVEt_HV: /* hash reference */
11768 case SAVEt_AV: /* array reference */
11769 sv = (const SV *) POPPTR(ss,ix);
11770 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11772 case SAVEt_COMPPAD:
11774 sv = (const SV *) POPPTR(ss,ix);
11775 TOPPTR(nss,ix) = sv_dup(sv, param);
11777 case SAVEt_INT: /* int reference */
11778 ptr = POPPTR(ss,ix);
11779 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11780 intval = (int)POPINT(ss,ix);
11781 TOPINT(nss,ix) = intval;
11783 case SAVEt_LONG: /* long reference */
11784 ptr = POPPTR(ss,ix);
11785 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11786 longval = (long)POPLONG(ss,ix);
11787 TOPLONG(nss,ix) = longval;
11789 case SAVEt_I32: /* I32 reference */
11790 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
11791 ptr = POPPTR(ss,ix);
11792 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11794 TOPINT(nss,ix) = i;
11796 case SAVEt_IV: /* IV reference */
11797 ptr = POPPTR(ss,ix);
11798 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11800 TOPIV(nss,ix) = iv;
11802 case SAVEt_HPTR: /* HV* reference */
11803 case SAVEt_APTR: /* AV* reference */
11804 case SAVEt_SPTR: /* SV* reference */
11805 ptr = POPPTR(ss,ix);
11806 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11807 sv = (const SV *)POPPTR(ss,ix);
11808 TOPPTR(nss,ix) = sv_dup(sv, param);
11810 case SAVEt_VPTR: /* random* reference */
11811 ptr = POPPTR(ss,ix);
11812 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11814 case SAVEt_INT_SMALL:
11815 case SAVEt_I32_SMALL:
11816 case SAVEt_I16: /* I16 reference */
11817 case SAVEt_I8: /* I8 reference */
11819 ptr = POPPTR(ss,ix);
11820 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11822 case SAVEt_GENERIC_PVREF: /* generic char* */
11823 case SAVEt_PPTR: /* char* reference */
11824 ptr = POPPTR(ss,ix);
11825 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11826 c = (char*)POPPTR(ss,ix);
11827 TOPPTR(nss,ix) = pv_dup(c);
11829 case SAVEt_GP: /* scalar reference */
11830 gv = (const GV *)POPPTR(ss,ix);
11831 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
11832 gp = (GP*)POPPTR(ss,ix);
11833 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
11834 (void)GpREFCNT_inc(gp);
11836 TOPINT(nss,ix) = i;
11839 ptr = POPPTR(ss,ix);
11840 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
11841 /* these are assumed to be refcounted properly */
11843 switch (((OP*)ptr)->op_type) {
11845 case OP_LEAVESUBLV:
11849 case OP_LEAVEWRITE:
11850 TOPPTR(nss,ix) = ptr;
11853 (void) OpREFCNT_inc(o);
11857 TOPPTR(nss,ix) = NULL;
11862 TOPPTR(nss,ix) = NULL;
11865 hv = (const HV *)POPPTR(ss,ix);
11866 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11868 TOPINT(nss,ix) = i;
11871 c = (char*)POPPTR(ss,ix);
11872 TOPPTR(nss,ix) = pv_dup_inc(c);
11874 case SAVEt_STACK_POS: /* Position on Perl stack */
11876 TOPINT(nss,ix) = i;
11878 case SAVEt_DESTRUCTOR:
11879 ptr = POPPTR(ss,ix);
11880 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11881 dptr = POPDPTR(ss,ix);
11882 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
11883 any_dup(FPTR2DPTR(void *, dptr),
11886 case SAVEt_DESTRUCTOR_X:
11887 ptr = POPPTR(ss,ix);
11888 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11889 dxptr = POPDXPTR(ss,ix);
11890 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
11891 any_dup(FPTR2DPTR(void *, dxptr),
11894 case SAVEt_REGCONTEXT:
11896 ix -= uv >> SAVE_TIGHT_SHIFT;
11898 case SAVEt_AELEM: /* array element */
11899 sv = (const SV *)POPPTR(ss,ix);
11900 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11902 TOPINT(nss,ix) = i;
11903 av = (const AV *)POPPTR(ss,ix);
11904 TOPPTR(nss,ix) = av_dup_inc(av, param);
11907 ptr = POPPTR(ss,ix);
11908 TOPPTR(nss,ix) = ptr;
11911 ptr = POPPTR(ss,ix);
11914 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
11915 HINTS_REFCNT_UNLOCK;
11917 TOPPTR(nss,ix) = ptr;
11919 TOPINT(nss,ix) = i;
11920 if (i & HINT_LOCALIZE_HH) {
11921 hv = (const HV *)POPPTR(ss,ix);
11922 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11925 case SAVEt_PADSV_AND_MORTALIZE:
11926 longval = (long)POPLONG(ss,ix);
11927 TOPLONG(nss,ix) = longval;
11928 ptr = POPPTR(ss,ix);
11929 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11930 sv = (const SV *)POPPTR(ss,ix);
11931 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11933 case SAVEt_SET_SVFLAGS:
11935 TOPINT(nss,ix) = i;
11937 TOPINT(nss,ix) = i;
11938 sv = (const SV *)POPPTR(ss,ix);
11939 TOPPTR(nss,ix) = sv_dup(sv, param);
11941 case SAVEt_RE_STATE:
11943 const struct re_save_state *const old_state
11944 = (struct re_save_state *)
11945 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11946 struct re_save_state *const new_state
11947 = (struct re_save_state *)
11948 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11950 Copy(old_state, new_state, 1, struct re_save_state);
11951 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
11953 new_state->re_state_bostr
11954 = pv_dup(old_state->re_state_bostr);
11955 new_state->re_state_reginput
11956 = pv_dup(old_state->re_state_reginput);
11957 new_state->re_state_regeol
11958 = pv_dup(old_state->re_state_regeol);
11959 new_state->re_state_regoffs
11960 = (regexp_paren_pair*)
11961 any_dup(old_state->re_state_regoffs, proto_perl);
11962 new_state->re_state_reglastparen
11963 = (U32*) any_dup(old_state->re_state_reglastparen,
11965 new_state->re_state_reglastcloseparen
11966 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
11968 /* XXX This just has to be broken. The old save_re_context
11969 code did SAVEGENERICPV(PL_reg_start_tmp);
11970 PL_reg_start_tmp is char **.
11971 Look above to what the dup code does for
11972 SAVEt_GENERIC_PVREF
11973 It can never have worked.
11974 So this is merely a faithful copy of the exiting bug: */
11975 new_state->re_state_reg_start_tmp
11976 = (char **) pv_dup((char *)
11977 old_state->re_state_reg_start_tmp);
11978 /* I assume that it only ever "worked" because no-one called
11979 (pseudo)fork while the regexp engine had re-entered itself.
11981 #ifdef PERL_OLD_COPY_ON_WRITE
11982 new_state->re_state_nrs
11983 = sv_dup(old_state->re_state_nrs, param);
11985 new_state->re_state_reg_magic
11986 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
11988 new_state->re_state_reg_oldcurpm
11989 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
11991 new_state->re_state_reg_curpm
11992 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
11994 new_state->re_state_reg_oldsaved
11995 = pv_dup(old_state->re_state_reg_oldsaved);
11996 new_state->re_state_reg_poscache
11997 = pv_dup(old_state->re_state_reg_poscache);
11998 new_state->re_state_reg_starttry
11999 = pv_dup(old_state->re_state_reg_starttry);
12002 case SAVEt_COMPILE_WARNINGS:
12003 ptr = POPPTR(ss,ix);
12004 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12007 ptr = POPPTR(ss,ix);
12008 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12012 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12020 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12021 * flag to the result. This is done for each stash before cloning starts,
12022 * so we know which stashes want their objects cloned */
12025 do_mark_cloneable_stash(pTHX_ SV *const sv)
12027 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12029 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12030 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12031 if (cloner && GvCV(cloner)) {
12038 mXPUSHs(newSVhek(hvname));
12040 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12047 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12055 =for apidoc perl_clone
12057 Create and return a new interpreter by cloning the current one.
12059 perl_clone takes these flags as parameters:
12061 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12062 without it we only clone the data and zero the stacks,
12063 with it we copy the stacks and the new perl interpreter is
12064 ready to run at the exact same point as the previous one.
12065 The pseudo-fork code uses COPY_STACKS while the
12066 threads->create doesn't.
12068 CLONEf_KEEP_PTR_TABLE
12069 perl_clone keeps a ptr_table with the pointer of the old
12070 variable as a key and the new variable as a value,
12071 this allows it to check if something has been cloned and not
12072 clone it again but rather just use the value and increase the
12073 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12074 the ptr_table using the function
12075 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12076 reason to keep it around is if you want to dup some of your own
12077 variable who are outside the graph perl scans, example of this
12078 code is in threads.xs create
12081 This is a win32 thing, it is ignored on unix, it tells perls
12082 win32host code (which is c++) to clone itself, this is needed on
12083 win32 if you want to run two threads at the same time,
12084 if you just want to do some stuff in a separate perl interpreter
12085 and then throw it away and return to the original one,
12086 you don't need to do anything.
12091 /* XXX the above needs expanding by someone who actually understands it ! */
12092 EXTERN_C PerlInterpreter *
12093 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12096 perl_clone(PerlInterpreter *proto_perl, UV flags)
12099 #ifdef PERL_IMPLICIT_SYS
12101 PERL_ARGS_ASSERT_PERL_CLONE;
12103 /* perlhost.h so we need to call into it
12104 to clone the host, CPerlHost should have a c interface, sky */
12106 if (flags & CLONEf_CLONE_HOST) {
12107 return perl_clone_host(proto_perl,flags);
12109 return perl_clone_using(proto_perl, flags,
12111 proto_perl->IMemShared,
12112 proto_perl->IMemParse,
12114 proto_perl->IStdIO,
12118 proto_perl->IProc);
12122 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12123 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12124 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12125 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12126 struct IPerlDir* ipD, struct IPerlSock* ipS,
12127 struct IPerlProc* ipP)
12129 /* XXX many of the string copies here can be optimized if they're
12130 * constants; they need to be allocated as common memory and just
12131 * their pointers copied. */
12134 CLONE_PARAMS clone_params;
12135 CLONE_PARAMS* const param = &clone_params;
12137 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12139 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12140 #else /* !PERL_IMPLICIT_SYS */
12142 CLONE_PARAMS clone_params;
12143 CLONE_PARAMS* param = &clone_params;
12144 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12146 PERL_ARGS_ASSERT_PERL_CLONE;
12147 #endif /* PERL_IMPLICIT_SYS */
12149 /* for each stash, determine whether its objects should be cloned */
12150 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12151 PERL_SET_THX(my_perl);
12154 PoisonNew(my_perl, 1, PerlInterpreter);
12159 PL_scopestack_name = 0;
12161 PL_savestack_ix = 0;
12162 PL_savestack_max = -1;
12163 PL_sig_pending = 0;
12165 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12166 # ifdef DEBUG_LEAKING_SCALARS
12167 PL_sv_serial = (((U32)my_perl >> 2) & 0xfff) * 1000000;
12169 #else /* !DEBUGGING */
12170 Zero(my_perl, 1, PerlInterpreter);
12171 #endif /* DEBUGGING */
12173 #ifdef PERL_IMPLICIT_SYS
12174 /* host pointers */
12176 PL_MemShared = ipMS;
12177 PL_MemParse = ipMP;
12184 #endif /* PERL_IMPLICIT_SYS */
12186 param->flags = flags;
12187 /* Nothing in the core code uses this, but we make it available to
12188 extensions (using mg_dup). */
12189 param->proto_perl = proto_perl;
12190 /* Likely nothing will use this, but it is initialised to be consistent
12191 with Perl_clone_params_new(). */
12192 param->proto_perl = my_perl;
12193 param->unreferenced = NULL;
12195 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12197 PL_body_arenas = NULL;
12198 Zero(&PL_body_roots, 1, PL_body_roots);
12200 PL_nice_chunk = NULL;
12201 PL_nice_chunk_size = 0;
12203 PL_sv_objcount = 0;
12205 PL_sv_arenaroot = NULL;
12207 PL_debug = proto_perl->Idebug;
12209 PL_hash_seed = proto_perl->Ihash_seed;
12210 PL_rehash_seed = proto_perl->Irehash_seed;
12212 #ifdef USE_REENTRANT_API
12213 /* XXX: things like -Dm will segfault here in perlio, but doing
12214 * PERL_SET_CONTEXT(proto_perl);
12215 * breaks too many other things
12217 Perl_reentrant_init(aTHX);
12220 /* create SV map for pointer relocation */
12221 PL_ptr_table = ptr_table_new();
12223 /* initialize these special pointers as early as possible */
12224 SvANY(&PL_sv_undef) = NULL;
12225 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12226 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12227 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
12229 SvANY(&PL_sv_no) = new_XPVNV();
12230 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12231 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12232 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12233 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
12234 SvCUR_set(&PL_sv_no, 0);
12235 SvLEN_set(&PL_sv_no, 1);
12236 SvIV_set(&PL_sv_no, 0);
12237 SvNV_set(&PL_sv_no, 0);
12238 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
12240 SvANY(&PL_sv_yes) = new_XPVNV();
12241 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12242 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12243 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12244 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
12245 SvCUR_set(&PL_sv_yes, 1);
12246 SvLEN_set(&PL_sv_yes, 2);
12247 SvIV_set(&PL_sv_yes, 1);
12248 SvNV_set(&PL_sv_yes, 1);
12249 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
12251 /* dbargs array probably holds garbage */
12254 /* create (a non-shared!) shared string table */
12255 PL_strtab = newHV();
12256 HvSHAREKEYS_off(PL_strtab);
12257 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
12258 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
12260 PL_compiling = proto_perl->Icompiling;
12262 /* These two PVs will be free'd special way so must set them same way op.c does */
12263 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
12264 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
12266 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
12267 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
12269 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
12270 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
12271 if (PL_compiling.cop_hints_hash) {
12273 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
12274 HINTS_REFCNT_UNLOCK;
12276 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
12277 #ifdef PERL_DEBUG_READONLY_OPS
12282 /* pseudo environmental stuff */
12283 PL_origargc = proto_perl->Iorigargc;
12284 PL_origargv = proto_perl->Iorigargv;
12286 param->stashes = newAV(); /* Setup array of objects to call clone on */
12287 /* This makes no difference to the implementation, as it always pushes
12288 and shifts pointers to other SVs without changing their reference
12289 count, with the array becoming empty before it is freed. However, it
12290 makes it conceptually clear what is going on, and will avoid some
12291 work inside av.c, filling slots between AvFILL() and AvMAX() with
12292 &PL_sv_undef, and SvREFCNT_dec()ing those. */
12293 AvREAL_off(param->stashes);
12295 if (!(flags & CLONEf_COPY_STACKS)) {
12296 param->unreferenced = newAV();
12299 /* Set tainting stuff before PerlIO_debug can possibly get called */
12300 PL_tainting = proto_perl->Itainting;
12301 PL_taint_warn = proto_perl->Itaint_warn;
12303 #ifdef PERLIO_LAYERS
12304 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
12305 PerlIO_clone(aTHX_ proto_perl, param);
12308 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
12309 PL_incgv = gv_dup(proto_perl->Iincgv, param);
12310 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
12311 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
12312 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
12313 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
12316 PL_minus_c = proto_perl->Iminus_c;
12317 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
12318 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
12319 PL_localpatches = proto_perl->Ilocalpatches;
12320 PL_splitstr = proto_perl->Isplitstr;
12321 PL_minus_n = proto_perl->Iminus_n;
12322 PL_minus_p = proto_perl->Iminus_p;
12323 PL_minus_l = proto_perl->Iminus_l;
12324 PL_minus_a = proto_perl->Iminus_a;
12325 PL_minus_E = proto_perl->Iminus_E;
12326 PL_minus_F = proto_perl->Iminus_F;
12327 PL_doswitches = proto_perl->Idoswitches;
12328 PL_dowarn = proto_perl->Idowarn;
12329 PL_doextract = proto_perl->Idoextract;
12330 PL_sawampersand = proto_perl->Isawampersand;
12331 PL_unsafe = proto_perl->Iunsafe;
12332 PL_inplace = SAVEPV(proto_perl->Iinplace);
12333 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
12334 PL_perldb = proto_perl->Iperldb;
12335 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12336 PL_exit_flags = proto_perl->Iexit_flags;
12338 /* magical thingies */
12339 /* XXX time(&PL_basetime) when asked for? */
12340 PL_basetime = proto_perl->Ibasetime;
12341 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
12343 PL_maxsysfd = proto_perl->Imaxsysfd;
12344 PL_statusvalue = proto_perl->Istatusvalue;
12346 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12348 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12350 PL_encoding = sv_dup(proto_perl->Iencoding, param);
12352 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
12353 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
12354 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
12357 /* RE engine related */
12358 Zero(&PL_reg_state, 1, struct re_save_state);
12359 PL_reginterp_cnt = 0;
12360 PL_regmatch_slab = NULL;
12362 /* Clone the regex array */
12363 /* ORANGE FIXME for plugins, probably in the SV dup code.
12364 newSViv(PTR2IV(CALLREGDUPE(
12365 INT2PTR(REGEXP *, SvIVX(regex)), param))))
12367 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
12368 PL_regex_pad = AvARRAY(PL_regex_padav);
12370 /* shortcuts to various I/O objects */
12371 PL_ofsgv = gv_dup(proto_perl->Iofsgv, param);
12372 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
12373 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
12374 PL_defgv = gv_dup(proto_perl->Idefgv, param);
12375 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
12376 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
12377 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
12379 /* shortcuts to regexp stuff */
12380 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
12382 /* shortcuts to misc objects */
12383 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
12385 /* shortcuts to debugging objects */
12386 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
12387 PL_DBline = gv_dup(proto_perl->IDBline, param);
12388 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
12389 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
12390 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
12391 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
12393 /* symbol tables */
12394 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
12395 PL_curstash = hv_dup(proto_perl->Icurstash, param);
12396 PL_debstash = hv_dup(proto_perl->Idebstash, param);
12397 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
12398 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
12400 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
12401 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
12402 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
12403 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
12404 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
12405 PL_endav = av_dup_inc(proto_perl->Iendav, param);
12406 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
12407 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
12409 PL_sub_generation = proto_perl->Isub_generation;
12410 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
12412 /* funky return mechanisms */
12413 PL_forkprocess = proto_perl->Iforkprocess;
12415 /* subprocess state */
12416 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
12418 /* internal state */
12419 PL_maxo = proto_perl->Imaxo;
12420 if (proto_perl->Iop_mask)
12421 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
12424 /* PL_asserting = proto_perl->Iasserting; */
12426 /* current interpreter roots */
12427 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
12429 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
12431 PL_main_start = proto_perl->Imain_start;
12432 PL_eval_root = proto_perl->Ieval_root;
12433 PL_eval_start = proto_perl->Ieval_start;
12435 /* runtime control stuff */
12436 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
12438 PL_filemode = proto_perl->Ifilemode;
12439 PL_lastfd = proto_perl->Ilastfd;
12440 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12443 PL_gensym = proto_perl->Igensym;
12444 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
12445 PL_laststatval = proto_perl->Ilaststatval;
12446 PL_laststype = proto_perl->Ilaststype;
12449 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
12451 /* interpreter atexit processing */
12452 PL_exitlistlen = proto_perl->Iexitlistlen;
12453 if (PL_exitlistlen) {
12454 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12455 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12458 PL_exitlist = (PerlExitListEntry*)NULL;
12460 PL_my_cxt_size = proto_perl->Imy_cxt_size;
12461 if (PL_my_cxt_size) {
12462 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
12463 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
12464 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12465 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
12466 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
12470 PL_my_cxt_list = (void**)NULL;
12471 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12472 PL_my_cxt_keys = (const char**)NULL;
12475 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
12476 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
12477 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
12479 PL_profiledata = NULL;
12481 PL_compcv = cv_dup(proto_perl->Icompcv, param);
12483 PAD_CLONE_VARS(proto_perl, param);
12485 #ifdef HAVE_INTERP_INTERN
12486 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
12489 /* more statics moved here */
12490 PL_generation = proto_perl->Igeneration;
12491 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
12493 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12494 PL_in_clean_all = proto_perl->Iin_clean_all;
12496 PL_uid = proto_perl->Iuid;
12497 PL_euid = proto_perl->Ieuid;
12498 PL_gid = proto_perl->Igid;
12499 PL_egid = proto_perl->Iegid;
12500 PL_nomemok = proto_perl->Inomemok;
12501 PL_an = proto_perl->Ian;
12502 PL_evalseq = proto_perl->Ievalseq;
12503 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12504 PL_origalen = proto_perl->Iorigalen;
12505 #ifdef PERL_USES_PL_PIDSTATUS
12506 PL_pidstatus = newHV(); /* XXX flag for cloning? */
12508 PL_osname = SAVEPV(proto_perl->Iosname);
12509 PL_sighandlerp = proto_perl->Isighandlerp;
12511 PL_runops = proto_perl->Irunops;
12513 PL_parser = parser_dup(proto_perl->Iparser, param);
12515 /* XXX this only works if the saved cop has already been cloned */
12516 if (proto_perl->Iparser) {
12517 PL_parser->saved_curcop = (COP*)any_dup(
12518 proto_perl->Iparser->saved_curcop,
12522 PL_subline = proto_perl->Isubline;
12523 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
12526 PL_cryptseen = proto_perl->Icryptseen;
12529 PL_hints = proto_perl->Ihints;
12531 PL_amagic_generation = proto_perl->Iamagic_generation;
12533 #ifdef USE_LOCALE_COLLATE
12534 PL_collation_ix = proto_perl->Icollation_ix;
12535 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
12536 PL_collation_standard = proto_perl->Icollation_standard;
12537 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12538 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12539 #endif /* USE_LOCALE_COLLATE */
12541 #ifdef USE_LOCALE_NUMERIC
12542 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
12543 PL_numeric_standard = proto_perl->Inumeric_standard;
12544 PL_numeric_local = proto_perl->Inumeric_local;
12545 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
12546 #endif /* !USE_LOCALE_NUMERIC */
12548 /* utf8 character classes */
12549 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
12550 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
12551 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
12552 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
12553 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
12554 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
12555 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
12556 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
12557 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
12558 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
12559 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
12560 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
12561 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
12562 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
12563 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
12564 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
12565 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
12566 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
12567 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
12568 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
12569 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
12570 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
12571 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
12572 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
12573 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
12574 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
12575 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
12576 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
12577 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
12579 /* Did the locale setup indicate UTF-8? */
12580 PL_utf8locale = proto_perl->Iutf8locale;
12581 /* Unicode features (see perlrun/-C) */
12582 PL_unicode = proto_perl->Iunicode;
12584 /* Pre-5.8 signals control */
12585 PL_signals = proto_perl->Isignals;
12587 /* times() ticks per second */
12588 PL_clocktick = proto_perl->Iclocktick;
12590 /* Recursion stopper for PerlIO_find_layer */
12591 PL_in_load_module = proto_perl->Iin_load_module;
12593 /* sort() routine */
12594 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12596 /* Not really needed/useful since the reenrant_retint is "volatile",
12597 * but do it for consistency's sake. */
12598 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12600 /* Hooks to shared SVs and locks. */
12601 PL_sharehook = proto_perl->Isharehook;
12602 PL_lockhook = proto_perl->Ilockhook;
12603 PL_unlockhook = proto_perl->Iunlockhook;
12604 PL_threadhook = proto_perl->Ithreadhook;
12605 PL_destroyhook = proto_perl->Idestroyhook;
12606 PL_signalhook = proto_perl->Isignalhook;
12608 #ifdef THREADS_HAVE_PIDS
12609 PL_ppid = proto_perl->Ippid;
12613 PL_last_swash_hv = NULL; /* reinits on demand */
12614 PL_last_swash_klen = 0;
12615 PL_last_swash_key[0]= '\0';
12616 PL_last_swash_tmps = (U8*)NULL;
12617 PL_last_swash_slen = 0;
12619 PL_glob_index = proto_perl->Iglob_index;
12620 PL_srand_called = proto_perl->Isrand_called;
12622 if (proto_perl->Ipsig_pend) {
12623 Newxz(PL_psig_pend, SIG_SIZE, int);
12626 PL_psig_pend = (int*)NULL;
12629 if (proto_perl->Ipsig_name) {
12630 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
12631 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
12633 PL_psig_ptr = PL_psig_name + SIG_SIZE;
12636 PL_psig_ptr = (SV**)NULL;
12637 PL_psig_name = (SV**)NULL;
12640 /* intrpvar.h stuff */
12642 if (flags & CLONEf_COPY_STACKS) {
12643 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12644 PL_tmps_ix = proto_perl->Itmps_ix;
12645 PL_tmps_max = proto_perl->Itmps_max;
12646 PL_tmps_floor = proto_perl->Itmps_floor;
12647 Newx(PL_tmps_stack, PL_tmps_max, SV*);
12648 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
12649 PL_tmps_ix+1, param);
12651 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
12652 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
12653 Newxz(PL_markstack, i, I32);
12654 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
12655 - proto_perl->Imarkstack);
12656 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
12657 - proto_perl->Imarkstack);
12658 Copy(proto_perl->Imarkstack, PL_markstack,
12659 PL_markstack_ptr - PL_markstack + 1, I32);
12661 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12662 * NOTE: unlike the others! */
12663 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12664 PL_scopestack_max = proto_perl->Iscopestack_max;
12665 Newxz(PL_scopestack, PL_scopestack_max, I32);
12666 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
12669 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
12670 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
12672 /* NOTE: si_dup() looks at PL_markstack */
12673 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
12675 /* PL_curstack = PL_curstackinfo->si_stack; */
12676 PL_curstack = av_dup(proto_perl->Icurstack, param);
12677 PL_mainstack = av_dup(proto_perl->Imainstack, param);
12679 /* next PUSHs() etc. set *(PL_stack_sp+1) */
12680 PL_stack_base = AvARRAY(PL_curstack);
12681 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
12682 - proto_perl->Istack_base);
12683 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
12685 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12686 * NOTE: unlike the others! */
12687 PL_savestack_ix = proto_perl->Isavestack_ix;
12688 PL_savestack_max = proto_perl->Isavestack_max;
12689 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
12690 PL_savestack = ss_dup(proto_perl, param);
12694 ENTER; /* perl_destruct() wants to LEAVE; */
12697 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12698 PL_top_env = &PL_start_env;
12700 PL_op = proto_perl->Iop;
12703 PL_Xpv = (XPV*)NULL;
12704 my_perl->Ina = proto_perl->Ina;
12706 PL_statbuf = proto_perl->Istatbuf;
12707 PL_statcache = proto_perl->Istatcache;
12708 PL_statgv = gv_dup(proto_perl->Istatgv, param);
12709 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
12711 PL_timesbuf = proto_perl->Itimesbuf;
12714 PL_tainted = proto_perl->Itainted;
12715 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12716 PL_rs = sv_dup_inc(proto_perl->Irs, param);
12717 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
12718 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
12719 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12720 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
12721 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
12722 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
12724 PL_restartjmpenv = proto_perl->Irestartjmpenv;
12725 PL_restartop = proto_perl->Irestartop;
12726 PL_in_eval = proto_perl->Iin_eval;
12727 PL_delaymagic = proto_perl->Idelaymagic;
12728 PL_dirty = proto_perl->Idirty;
12729 PL_localizing = proto_perl->Ilocalizing;
12731 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
12732 PL_hv_fetch_ent_mh = NULL;
12733 PL_modcount = proto_perl->Imodcount;
12734 PL_lastgotoprobe = NULL;
12735 PL_dumpindent = proto_perl->Idumpindent;
12737 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
12738 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
12739 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
12740 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
12741 PL_efloatbuf = NULL; /* reinits on demand */
12742 PL_efloatsize = 0; /* reinits on demand */
12746 PL_screamfirst = NULL;
12747 PL_screamnext = NULL;
12748 PL_maxscream = -1; /* reinits on demand */
12749 PL_lastscream = NULL;
12752 PL_regdummy = proto_perl->Iregdummy;
12753 PL_colorset = 0; /* reinits PL_colors[] */
12754 /*PL_colors[6] = {0,0,0,0,0,0};*/
12758 /* Pluggable optimizer */
12759 PL_peepp = proto_perl->Ipeepp;
12760 /* op_free() hook */
12761 PL_opfreehook = proto_perl->Iopfreehook;
12763 PL_stashcache = newHV();
12765 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
12766 proto_perl->Iwatchaddr);
12767 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
12768 if (PL_debug && PL_watchaddr) {
12769 PerlIO_printf(Perl_debug_log,
12770 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
12771 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
12772 PTR2UV(PL_watchok));
12775 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
12776 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
12778 /* Call the ->CLONE method, if it exists, for each of the stashes
12779 identified by sv_dup() above.
12781 while(av_len(param->stashes) != -1) {
12782 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
12783 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
12784 if (cloner && GvCV(cloner)) {
12789 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
12791 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
12797 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
12798 ptr_table_free(PL_ptr_table);
12799 PL_ptr_table = NULL;
12802 if (!(flags & CLONEf_COPY_STACKS)) {
12803 unreferenced_to_tmp_stack(param->unreferenced);
12806 SvREFCNT_dec(param->stashes);
12808 /* orphaned? eg threads->new inside BEGIN or use */
12809 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
12810 SvREFCNT_inc_simple_void(PL_compcv);
12811 SAVEFREESV(PL_compcv);
12818 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
12820 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
12822 if (AvFILLp(unreferenced) > -1) {
12823 SV **svp = AvARRAY(unreferenced);
12824 SV **const last = svp + AvFILLp(unreferenced);
12828 if (SvREFCNT(*svp) == 1)
12830 } while (++svp <= last);
12832 EXTEND_MORTAL(count);
12833 svp = AvARRAY(unreferenced);
12836 if (SvREFCNT(*svp) == 1) {
12837 /* Our reference is the only one to this SV. This means that
12838 in this thread, the scalar effectively has a 0 reference.
12839 That doesn't work (cleanup never happens), so donate our
12840 reference to it onto the save stack. */
12841 PL_tmps_stack[++PL_tmps_ix] = *svp;
12843 /* As an optimisation, because we are already walking the
12844 entire array, instead of above doing either
12845 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
12846 release our reference to the scalar, so that at the end of
12847 the array owns zero references to the scalars it happens to
12848 point to. We are effectively converting the array from
12849 AvREAL() on to AvREAL() off. This saves the av_clear()
12850 (triggered by the SvREFCNT_dec(unreferenced) below) from
12851 walking the array a second time. */
12852 SvREFCNT_dec(*svp);
12855 } while (++svp <= last);
12856 AvREAL_off(unreferenced);
12858 SvREFCNT_dec(unreferenced);
12862 Perl_clone_params_del(CLONE_PARAMS *param)
12864 PerlInterpreter *const was = PERL_GET_THX;
12865 PerlInterpreter *const to = param->new_perl;
12868 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
12874 SvREFCNT_dec(param->stashes);
12875 if (param->unreferenced)
12876 unreferenced_to_tmp_stack(param->unreferenced);
12886 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
12888 /* Need to play this game, as newAV() can call safesysmalloc(), and that
12889 does a dTHX; to get the context from thread local storage.
12890 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
12891 a version that passes in my_perl. */
12892 PerlInterpreter *const was = PERL_GET_THX;
12893 CLONE_PARAMS *param;
12895 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
12901 /* Given that we've set the context, we can do this unshared. */
12902 Newx(param, 1, CLONE_PARAMS);
12905 param->proto_perl = from;
12906 param->new_perl = to;
12907 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
12908 AvREAL_off(param->stashes);
12909 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
12917 #endif /* USE_ITHREADS */
12920 =head1 Unicode Support
12922 =for apidoc sv_recode_to_utf8
12924 The encoding is assumed to be an Encode object, on entry the PV
12925 of the sv is assumed to be octets in that encoding, and the sv
12926 will be converted into Unicode (and UTF-8).
12928 If the sv already is UTF-8 (or if it is not POK), or if the encoding
12929 is not a reference, nothing is done to the sv. If the encoding is not
12930 an C<Encode::XS> Encoding object, bad things will happen.
12931 (See F<lib/encoding.pm> and L<Encode>).
12933 The PV of the sv is returned.
12938 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
12942 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
12944 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
12958 Passing sv_yes is wrong - it needs to be or'ed set of constants
12959 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
12960 remove converted chars from source.
12962 Both will default the value - let them.
12964 XPUSHs(&PL_sv_yes);
12967 call_method("decode", G_SCALAR);
12971 s = SvPV_const(uni, len);
12972 if (s != SvPVX_const(sv)) {
12973 SvGROW(sv, len + 1);
12974 Move(s, SvPVX(sv), len + 1, char);
12975 SvCUR_set(sv, len);
12982 return SvPOKp(sv) ? SvPVX(sv) : NULL;
12986 =for apidoc sv_cat_decode
12988 The encoding is assumed to be an Encode object, the PV of the ssv is
12989 assumed to be octets in that encoding and decoding the input starts
12990 from the position which (PV + *offset) pointed to. The dsv will be
12991 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
12992 when the string tstr appears in decoding output or the input ends on
12993 the PV of the ssv. The value which the offset points will be modified
12994 to the last input position on the ssv.
12996 Returns TRUE if the terminator was found, else returns FALSE.
13001 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13002 SV *ssv, int *offset, char *tstr, int tlen)
13007 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13009 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13020 offsv = newSViv(*offset);
13022 mXPUSHp(tstr, tlen);
13024 call_method("cat_decode", G_SCALAR);
13026 ret = SvTRUE(TOPs);
13027 *offset = SvIV(offsv);
13033 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13038 /* ---------------------------------------------------------------------
13040 * support functions for report_uninit()
13043 /* the maxiumum size of array or hash where we will scan looking
13044 * for the undefined element that triggered the warning */
13046 #define FUV_MAX_SEARCH_SIZE 1000
13048 /* Look for an entry in the hash whose value has the same SV as val;
13049 * If so, return a mortal copy of the key. */
13052 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13055 register HE **array;
13058 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13060 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13061 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13064 array = HvARRAY(hv);
13066 for (i=HvMAX(hv); i>0; i--) {
13067 register HE *entry;
13068 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13069 if (HeVAL(entry) != val)
13071 if ( HeVAL(entry) == &PL_sv_undef ||
13072 HeVAL(entry) == &PL_sv_placeholder)
13076 if (HeKLEN(entry) == HEf_SVKEY)
13077 return sv_mortalcopy(HeKEY_sv(entry));
13078 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13084 /* Look for an entry in the array whose value has the same SV as val;
13085 * If so, return the index, otherwise return -1. */
13088 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13092 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13094 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13095 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13098 if (val != &PL_sv_undef) {
13099 SV ** const svp = AvARRAY(av);
13102 for (i=AvFILLp(av); i>=0; i--)
13109 /* S_varname(): return the name of a variable, optionally with a subscript.
13110 * If gv is non-zero, use the name of that global, along with gvtype (one
13111 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13112 * targ. Depending on the value of the subscript_type flag, return:
13115 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13116 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13117 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13118 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13121 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13122 const SV *const keyname, I32 aindex, int subscript_type)
13125 SV * const name = sv_newmortal();
13128 buffer[0] = gvtype;
13131 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13133 gv_fullname4(name, gv, buffer, 0);
13135 if ((unsigned int)SvPVX(name)[1] <= 26) {
13137 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13139 /* Swap the 1 unprintable control character for the 2 byte pretty
13140 version - ie substr($name, 1, 1) = $buffer; */
13141 sv_insert(name, 1, 1, buffer, 2);
13145 CV * const cv = find_runcv(NULL);
13149 if (!cv || !CvPADLIST(cv))
13151 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13152 sv = *av_fetch(av, targ, FALSE);
13153 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
13156 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13157 SV * const sv = newSV(0);
13158 *SvPVX(name) = '$';
13159 Perl_sv_catpvf(aTHX_ name, "{%s}",
13160 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
13163 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13164 *SvPVX(name) = '$';
13165 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13167 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13168 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13169 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13177 =for apidoc find_uninit_var
13179 Find the name of the undefined variable (if any) that caused the operator o
13180 to issue a "Use of uninitialized value" warning.
13181 If match is true, only return a name if it's value matches uninit_sv.
13182 So roughly speaking, if a unary operator (such as OP_COS) generates a
13183 warning, then following the direct child of the op may yield an
13184 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13185 other hand, with OP_ADD there are two branches to follow, so we only print
13186 the variable name if we get an exact match.
13188 The name is returned as a mortal SV.
13190 Assumes that PL_op is the op that originally triggered the error, and that
13191 PL_comppad/PL_curpad points to the currently executing pad.
13197 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13203 const OP *o, *o2, *kid;
13205 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13206 uninit_sv == &PL_sv_placeholder)))
13209 switch (obase->op_type) {
13216 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13217 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13220 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13222 if (pad) { /* @lex, %lex */
13223 sv = PAD_SVl(obase->op_targ);
13227 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13228 /* @global, %global */
13229 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13232 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13234 else /* @{expr}, %{expr} */
13235 return find_uninit_var(cUNOPx(obase)->op_first,
13239 /* attempt to find a match within the aggregate */
13241 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13243 subscript_type = FUV_SUBSCRIPT_HASH;
13246 index = find_array_subscript((const AV *)sv, uninit_sv);
13248 subscript_type = FUV_SUBSCRIPT_ARRAY;
13251 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
13254 return varname(gv, hash ? '%' : '@', obase->op_targ,
13255 keysv, index, subscript_type);
13259 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
13261 return varname(NULL, '$', obase->op_targ,
13262 NULL, 0, FUV_SUBSCRIPT_NONE);
13265 gv = cGVOPx_gv(obase);
13266 if (!gv || (match && GvSV(gv) != uninit_sv))
13268 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13271 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
13274 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
13275 if (!av || SvRMAGICAL(av))
13277 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13278 if (!svp || *svp != uninit_sv)
13281 return varname(NULL, '$', obase->op_targ,
13282 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13285 gv = cGVOPx_gv(obase);
13290 AV *const av = GvAV(gv);
13291 if (!av || SvRMAGICAL(av))
13293 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13294 if (!svp || *svp != uninit_sv)
13297 return varname(gv, '$', 0,
13298 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13303 o = cUNOPx(obase)->op_first;
13304 if (!o || o->op_type != OP_NULL ||
13305 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
13307 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
13311 if (PL_op == obase)
13312 /* $a[uninit_expr] or $h{uninit_expr} */
13313 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
13316 o = cBINOPx(obase)->op_first;
13317 kid = cBINOPx(obase)->op_last;
13319 /* get the av or hv, and optionally the gv */
13321 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
13322 sv = PAD_SV(o->op_targ);
13324 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
13325 && cUNOPo->op_first->op_type == OP_GV)
13327 gv = cGVOPx_gv(cUNOPo->op_first);
13331 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
13336 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
13337 /* index is constant */
13341 if (obase->op_type == OP_HELEM) {
13342 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
13343 if (!he || HeVAL(he) != uninit_sv)
13347 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
13348 if (!svp || *svp != uninit_sv)
13352 if (obase->op_type == OP_HELEM)
13353 return varname(gv, '%', o->op_targ,
13354 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
13356 return varname(gv, '@', o->op_targ, NULL,
13357 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
13360 /* index is an expression;
13361 * attempt to find a match within the aggregate */
13362 if (obase->op_type == OP_HELEM) {
13363 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13365 return varname(gv, '%', o->op_targ,
13366 keysv, 0, FUV_SUBSCRIPT_HASH);
13370 = find_array_subscript((const AV *)sv, uninit_sv);
13372 return varname(gv, '@', o->op_targ,
13373 NULL, index, FUV_SUBSCRIPT_ARRAY);
13378 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
13380 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
13385 /* only examine RHS */
13386 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
13389 o = cUNOPx(obase)->op_first;
13390 if (o->op_type == OP_PUSHMARK)
13393 if (!o->op_sibling) {
13394 /* one-arg version of open is highly magical */
13396 if (o->op_type == OP_GV) { /* open FOO; */
13398 if (match && GvSV(gv) != uninit_sv)
13400 return varname(gv, '$', 0,
13401 NULL, 0, FUV_SUBSCRIPT_NONE);
13403 /* other possibilities not handled are:
13404 * open $x; or open my $x; should return '${*$x}'
13405 * open expr; should return '$'.expr ideally
13411 /* ops where $_ may be an implicit arg */
13415 if ( !(obase->op_flags & OPf_STACKED)) {
13416 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
13417 ? PAD_SVl(obase->op_targ)
13420 sv = sv_newmortal();
13421 sv_setpvs(sv, "$_");
13430 match = 1; /* print etc can return undef on defined args */
13431 /* skip filehandle as it can't produce 'undef' warning */
13432 o = cUNOPx(obase)->op_first;
13433 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
13434 o = o->op_sibling->op_sibling;
13438 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
13440 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
13442 /* the following ops are capable of returning PL_sv_undef even for
13443 * defined arg(s) */
13462 case OP_GETPEERNAME:
13510 case OP_SMARTMATCH:
13519 /* XXX tmp hack: these two may call an XS sub, and currently
13520 XS subs don't have a SUB entry on the context stack, so CV and
13521 pad determination goes wrong, and BAD things happen. So, just
13522 don't try to determine the value under those circumstances.
13523 Need a better fix at dome point. DAPM 11/2007 */
13529 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
13530 if (gv && GvSV(gv) == uninit_sv)
13531 return newSVpvs_flags("$.", SVs_TEMP);
13536 /* def-ness of rval pos() is independent of the def-ness of its arg */
13537 if ( !(obase->op_flags & OPf_MOD))
13542 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
13543 return newSVpvs_flags("${$/}", SVs_TEMP);
13548 if (!(obase->op_flags & OPf_KIDS))
13550 o = cUNOPx(obase)->op_first;
13556 /* if all except one arg are constant, or have no side-effects,
13557 * or are optimized away, then it's unambiguous */
13559 for (kid=o; kid; kid = kid->op_sibling) {
13561 const OPCODE type = kid->op_type;
13562 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
13563 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
13564 || (type == OP_PUSHMARK)
13568 if (o2) { /* more than one found */
13575 return find_uninit_var(o2, uninit_sv, match);
13577 /* scan all args */
13579 sv = find_uninit_var(o, uninit_sv, 1);
13591 =for apidoc report_uninit
13593 Print appropriate "Use of uninitialized variable" warning
13599 Perl_report_uninit(pTHX_ const SV *uninit_sv)
13603 SV* varname = NULL;
13605 varname = find_uninit_var(PL_op, uninit_sv,0);
13607 sv_insert(varname, 0, 0, " ", 1);
13609 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13610 varname ? SvPV_nolen_const(varname) : "",
13611 " in ", OP_DESC(PL_op));
13614 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13620 * c-indentation-style: bsd
13621 * c-basic-offset: 4
13622 * indent-tabs-mode: t
13625 * ex: set ts=8 sts=4 sw=4 noet: