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_XIV(), del_XIV(),
151 new_XNV(), del_XNV(),
156 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
160 * ========================================================================= */
163 * "A time to plant, and a time to uproot what was planted..."
167 Perl_offer_nice_chunk(pTHX_ void *const chunk, const U32 chunk_size)
173 PERL_ARGS_ASSERT_OFFER_NICE_CHUNK;
175 new_chunk = (void *)(chunk);
176 new_chunk_size = (chunk_size);
177 if (new_chunk_size > PL_nice_chunk_size) {
178 Safefree(PL_nice_chunk);
179 PL_nice_chunk = (char *) new_chunk;
180 PL_nice_chunk_size = new_chunk_size;
187 # define MEM_LOG_NEW_SV(sv, file, line, func) \
188 Perl_mem_log_new_sv(sv, file, line, func)
189 # define MEM_LOG_DEL_SV(sv, file, line, func) \
190 Perl_mem_log_del_sv(sv, file, line, func)
192 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
193 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
196 #ifdef DEBUG_LEAKING_SCALARS
197 # define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
198 # define DEBUG_SV_SERIAL(sv) \
199 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
200 PTR2UV(sv), (long)(sv)->sv_debug_serial))
202 # define FREE_SV_DEBUG_FILE(sv)
203 # define DEBUG_SV_SERIAL(sv) NOOP
207 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
208 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
209 /* Whilst I'd love to do this, it seems that things like to check on
211 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
213 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
214 PoisonNew(&SvREFCNT(sv), 1, U32)
216 # define SvARENA_CHAIN(sv) SvANY(sv)
217 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
218 # define POSION_SV_HEAD(sv)
221 /* Mark an SV head as unused, and add to free list.
223 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
224 * its refcount artificially decremented during global destruction, so
225 * there may be dangling pointers to it. The last thing we want in that
226 * case is for it to be reused. */
228 #define plant_SV(p) \
230 const U32 old_flags = SvFLAGS(p); \
231 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
232 DEBUG_SV_SERIAL(p); \
233 FREE_SV_DEBUG_FILE(p); \
235 SvFLAGS(p) = SVTYPEMASK; \
236 if (!(old_flags & SVf_BREAK)) { \
237 SvARENA_CHAIN_SET(p, PL_sv_root); \
243 #define uproot_SV(p) \
246 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
251 /* make some more SVs by adding another arena */
260 sv_add_arena(PL_nice_chunk, PL_nice_chunk_size, 0);
261 PL_nice_chunk = NULL;
262 PL_nice_chunk_size = 0;
265 char *chunk; /* must use New here to match call to */
266 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
267 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
273 /* new_SV(): return a new, empty SV head */
275 #ifdef DEBUG_LEAKING_SCALARS
276 /* provide a real function for a debugger to play with */
278 S_new_SV(pTHX_ const char *file, int line, const char *func)
285 sv = S_more_sv(aTHX);
289 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
290 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
296 sv->sv_debug_inpad = 0;
297 sv->sv_debug_cloned = 0;
298 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
300 sv->sv_debug_serial = PL_sv_serial++;
302 MEM_LOG_NEW_SV(sv, file, line, func);
303 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
304 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
308 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
316 (p) = S_more_sv(aTHX); \
320 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
325 /* del_SV(): return an empty SV head to the free list */
338 S_del_sv(pTHX_ SV *p)
342 PERL_ARGS_ASSERT_DEL_SV;
347 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
348 const SV * const sv = sva + 1;
349 const SV * const svend = &sva[SvREFCNT(sva)];
350 if (p >= sv && p < svend) {
356 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
357 "Attempt to free non-arena SV: 0x%"UVxf
358 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
365 #else /* ! DEBUGGING */
367 #define del_SV(p) plant_SV(p)
369 #endif /* DEBUGGING */
373 =head1 SV Manipulation Functions
375 =for apidoc sv_add_arena
377 Given a chunk of memory, link it to the head of the list of arenas,
378 and split it into a list of free SVs.
384 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
387 SV *const sva = MUTABLE_SV(ptr);
391 PERL_ARGS_ASSERT_SV_ADD_ARENA;
393 /* The first SV in an arena isn't an SV. */
394 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
395 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
396 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
398 PL_sv_arenaroot = sva;
399 PL_sv_root = sva + 1;
401 svend = &sva[SvREFCNT(sva) - 1];
404 SvARENA_CHAIN_SET(sv, (sv + 1));
408 /* Must always set typemask because it's always checked in on cleanup
409 when the arenas are walked looking for objects. */
410 SvFLAGS(sv) = SVTYPEMASK;
413 SvARENA_CHAIN_SET(sv, 0);
417 SvFLAGS(sv) = SVTYPEMASK;
420 /* visit(): call the named function for each non-free SV in the arenas
421 * whose flags field matches the flags/mask args. */
424 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
430 PERL_ARGS_ASSERT_VISIT;
432 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
433 register const SV * const svend = &sva[SvREFCNT(sva)];
435 for (sv = sva + 1; sv < svend; ++sv) {
436 if (SvTYPE(sv) != SVTYPEMASK
437 && (sv->sv_flags & mask) == flags
450 /* called by sv_report_used() for each live SV */
453 do_report_used(pTHX_ SV *const sv)
455 if (SvTYPE(sv) != SVTYPEMASK) {
456 PerlIO_printf(Perl_debug_log, "****\n");
463 =for apidoc sv_report_used
465 Dump the contents of all SVs not yet freed. (Debugging aid).
471 Perl_sv_report_used(pTHX)
474 visit(do_report_used, 0, 0);
480 /* called by sv_clean_objs() for each live SV */
483 do_clean_objs(pTHX_ SV *const ref)
488 SV * const target = SvRV(ref);
489 if (SvOBJECT(target)) {
490 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
491 if (SvWEAKREF(ref)) {
492 sv_del_backref(target, ref);
498 SvREFCNT_dec(target);
503 /* XXX Might want to check arrays, etc. */
506 /* called by sv_clean_objs() for each live SV */
508 #ifndef DISABLE_DESTRUCTOR_KLUDGE
510 do_clean_named_objs(pTHX_ SV *const sv)
513 assert(SvTYPE(sv) == SVt_PVGV);
514 assert(isGV_with_GP(sv));
517 #ifdef PERL_DONT_CREATE_GVSV
520 SvOBJECT(GvSV(sv))) ||
521 (GvAV(sv) && SvOBJECT(GvAV(sv))) ||
522 (GvHV(sv) && SvOBJECT(GvHV(sv))) ||
523 /* In certain rare cases GvIOp(sv) can be NULL, which would make SvOBJECT(GvIO(sv)) dereference NULL. */
524 (GvIO(sv) ? (SvFLAGS(GvIOp(sv)) & SVs_OBJECT) : 0) ||
525 (GvCV(sv) && SvOBJECT(GvCV(sv))) )
527 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning named glob object:\n "), sv_dump(sv)));
528 SvFLAGS(sv) |= SVf_BREAK;
536 =for apidoc sv_clean_objs
538 Attempt to destroy all objects not yet freed
544 Perl_sv_clean_objs(pTHX)
547 PL_in_clean_objs = TRUE;
548 visit(do_clean_objs, SVf_ROK, SVf_ROK);
549 #ifndef DISABLE_DESTRUCTOR_KLUDGE
550 /* some barnacles may yet remain, clinging to typeglobs */
551 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
553 PL_in_clean_objs = FALSE;
556 /* called by sv_clean_all() for each live SV */
559 do_clean_all(pTHX_ SV *const sv)
562 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
563 /* don't clean pid table and strtab */
566 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
567 SvFLAGS(sv) |= SVf_BREAK;
572 =for apidoc sv_clean_all
574 Decrement the refcnt of each remaining SV, possibly triggering a
575 cleanup. This function may have to be called multiple times to free
576 SVs which are in complex self-referential hierarchies.
582 Perl_sv_clean_all(pTHX)
586 PL_in_clean_all = TRUE;
587 cleaned = visit(do_clean_all, 0,0);
588 PL_in_clean_all = FALSE;
593 ARENASETS: a meta-arena implementation which separates arena-info
594 into struct arena_set, which contains an array of struct
595 arena_descs, each holding info for a single arena. By separating
596 the meta-info from the arena, we recover the 1st slot, formerly
597 borrowed for list management. The arena_set is about the size of an
598 arena, avoiding the needless malloc overhead of a naive linked-list.
600 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
601 memory in the last arena-set (1/2 on average). In trade, we get
602 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
603 smaller types). The recovery of the wasted space allows use of
604 small arenas for large, rare body types, by changing array* fields
605 in body_details_by_type[] below.
608 char *arena; /* the raw storage, allocated aligned */
609 size_t size; /* its size ~4k typ */
610 svtype utype; /* bodytype stored in arena */
615 /* Get the maximum number of elements in set[] such that struct arena_set
616 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
617 therefore likely to be 1 aligned memory page. */
619 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
620 - 2 * sizeof(int)) / sizeof (struct arena_desc))
623 struct arena_set* next;
624 unsigned int set_size; /* ie ARENAS_PER_SET */
625 unsigned int curr; /* index of next available arena-desc */
626 struct arena_desc set[ARENAS_PER_SET];
630 =for apidoc sv_free_arenas
632 Deallocate the memory used by all arenas. Note that all the individual SV
633 heads and bodies within the arenas must already have been freed.
638 Perl_sv_free_arenas(pTHX)
645 /* Free arenas here, but be careful about fake ones. (We assume
646 contiguity of the fake ones with the corresponding real ones.) */
648 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
649 svanext = MUTABLE_SV(SvANY(sva));
650 while (svanext && SvFAKE(svanext))
651 svanext = MUTABLE_SV(SvANY(svanext));
658 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
661 struct arena_set *current = aroot;
664 assert(aroot->set[i].arena);
665 Safefree(aroot->set[i].arena);
673 i = PERL_ARENA_ROOTS_SIZE;
675 PL_body_roots[i] = 0;
677 Safefree(PL_nice_chunk);
678 PL_nice_chunk = NULL;
679 PL_nice_chunk_size = 0;
685 Here are mid-level routines that manage the allocation of bodies out
686 of the various arenas. There are 5 kinds of arenas:
688 1. SV-head arenas, which are discussed and handled above
689 2. regular body arenas
690 3. arenas for reduced-size bodies
693 Arena types 2 & 3 are chained by body-type off an array of
694 arena-root pointers, which is indexed by svtype. Some of the
695 larger/less used body types are malloced singly, since a large
696 unused block of them is wasteful. Also, several svtypes dont have
697 bodies; the data fits into the sv-head itself. The arena-root
698 pointer thus has a few unused root-pointers (which may be hijacked
699 later for arena types 4,5)
701 3 differs from 2 as an optimization; some body types have several
702 unused fields in the front of the structure (which are kept in-place
703 for consistency). These bodies can be allocated in smaller chunks,
704 because the leading fields arent accessed. Pointers to such bodies
705 are decremented to point at the unused 'ghost' memory, knowing that
706 the pointers are used with offsets to the real memory.
708 HE, HEK arenas are managed separately, with separate code, but may
709 be merge-able later..
712 /* get_arena(size): this creates custom-sized arenas
713 TBD: export properly for hv.c: S_more_he().
716 Perl_get_arena(pTHX_ const size_t arena_size, const svtype bodytype)
719 struct arena_desc* adesc;
720 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
723 /* shouldnt need this
724 if (!arena_size) arena_size = PERL_ARENA_SIZE;
727 /* may need new arena-set to hold new arena */
728 if (!aroot || aroot->curr >= aroot->set_size) {
729 struct arena_set *newroot;
730 Newxz(newroot, 1, struct arena_set);
731 newroot->set_size = ARENAS_PER_SET;
732 newroot->next = aroot;
734 PL_body_arenas = (void *) newroot;
735 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
738 /* ok, now have arena-set with at least 1 empty/available arena-desc */
739 curr = aroot->curr++;
740 adesc = &(aroot->set[curr]);
741 assert(!adesc->arena);
743 Newx(adesc->arena, arena_size, char);
744 adesc->size = arena_size;
745 adesc->utype = bodytype;
746 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
747 curr, (void*)adesc->arena, (UV)arena_size));
753 /* return a thing to the free list */
755 #define del_body(thing, root) \
757 void ** const thing_copy = (void **)thing;\
758 *thing_copy = *root; \
759 *root = (void*)thing_copy; \
764 =head1 SV-Body Allocation
766 Allocation of SV-bodies is similar to SV-heads, differing as follows;
767 the allocation mechanism is used for many body types, so is somewhat
768 more complicated, it uses arena-sets, and has no need for still-live
771 At the outermost level, (new|del)_X*V macros return bodies of the
772 appropriate type. These macros call either (new|del)_body_type or
773 (new|del)_body_allocated macro pairs, depending on specifics of the
774 type. Most body types use the former pair, the latter pair is used to
775 allocate body types with "ghost fields".
777 "ghost fields" are fields that are unused in certain types, and
778 consequently don't need to actually exist. They are declared because
779 they're part of a "base type", which allows use of functions as
780 methods. The simplest examples are AVs and HVs, 2 aggregate types
781 which don't use the fields which support SCALAR semantics.
783 For these types, the arenas are carved up into appropriately sized
784 chunks, we thus avoid wasted memory for those unaccessed members.
785 When bodies are allocated, we adjust the pointer back in memory by the
786 size of the part not allocated, so it's as if we allocated the full
787 structure. (But things will all go boom if you write to the part that
788 is "not there", because you'll be overwriting the last members of the
789 preceding structure in memory.)
791 We calculate the correction using the STRUCT_OFFSET macro on the first
792 member present. If the allocated structure is smaller (no initial NV
793 actually allocated) then the net effect is to subtract the size of the NV
794 from the pointer, to return a new pointer as if an initial NV were actually
795 allocated. (We were using structures named *_allocated for this, but
796 this turned out to be a subtle bug, because a structure without an NV
797 could have a lower alignment constraint, but the compiler is allowed to
798 optimised accesses based on the alignment constraint of the actual pointer
799 to the full structure, for example, using a single 64 bit load instruction
800 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
802 This is the same trick as was used for NV and IV bodies. Ironically it
803 doesn't need to be used for NV bodies any more, because NV is now at
804 the start of the structure. IV bodies don't need it either, because
805 they are no longer allocated.
807 In turn, the new_body_* allocators call S_new_body(), which invokes
808 new_body_inline macro, which takes a lock, and takes a body off the
809 linked list at PL_body_roots[sv_type], calling S_more_bodies() if
810 necessary to refresh an empty list. Then the lock is released, and
811 the body is returned.
813 S_more_bodies calls get_arena(), and carves it up into an array of N
814 bodies, which it strings into a linked list. It looks up arena-size
815 and body-size from the body_details table described below, thus
816 supporting the multiple body-types.
818 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
819 the (new|del)_X*V macros are mapped directly to malloc/free.
825 For each sv-type, struct body_details bodies_by_type[] carries
826 parameters which control these aspects of SV handling:
828 Arena_size determines whether arenas are used for this body type, and if
829 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
830 zero, forcing individual mallocs and frees.
832 Body_size determines how big a body is, and therefore how many fit into
833 each arena. Offset carries the body-pointer adjustment needed for
834 "ghost fields", and is used in *_allocated macros.
836 But its main purpose is to parameterize info needed in
837 Perl_sv_upgrade(). The info here dramatically simplifies the function
838 vs the implementation in 5.8.8, making it table-driven. All fields
839 are used for this, except for arena_size.
841 For the sv-types that have no bodies, arenas are not used, so those
842 PL_body_roots[sv_type] are unused, and can be overloaded. In
843 something of a special case, SVt_NULL is borrowed for HE arenas;
844 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
845 bodies_by_type[SVt_NULL] slot is not used, as the table is not
850 struct body_details {
851 U8 body_size; /* Size to allocate */
852 U8 copy; /* Size of structure to copy (may be shorter) */
854 unsigned int type : 4; /* We have space for a sanity check. */
855 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
856 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
857 unsigned int arena : 1; /* Allocated from an arena */
858 size_t arena_size; /* Size of arena to allocate */
866 /* With -DPURFIY we allocate everything directly, and don't use arenas.
867 This seems a rather elegant way to simplify some of the code below. */
868 #define HASARENA FALSE
870 #define HASARENA TRUE
872 #define NOARENA FALSE
874 /* Size the arenas to exactly fit a given number of bodies. A count
875 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
876 simplifying the default. If count > 0, the arena is sized to fit
877 only that many bodies, allowing arenas to be used for large, rare
878 bodies (XPVFM, XPVIO) without undue waste. The arena size is
879 limited by PERL_ARENA_SIZE, so we can safely oversize the
882 #define FIT_ARENA0(body_size) \
883 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
884 #define FIT_ARENAn(count,body_size) \
885 ( count * body_size <= PERL_ARENA_SIZE) \
886 ? count * body_size \
887 : FIT_ARENA0 (body_size)
888 #define FIT_ARENA(count,body_size) \
890 ? FIT_ARENAn (count, body_size) \
891 : FIT_ARENA0 (body_size)
893 /* Calculate the length to copy. Specifically work out the length less any
894 final padding the compiler needed to add. See the comment in sv_upgrade
895 for why copying the padding proved to be a bug. */
897 #define copy_length(type, last_member) \
898 STRUCT_OFFSET(type, last_member) \
899 + sizeof (((type*)SvANY((const SV *)0))->last_member)
901 static const struct body_details bodies_by_type[] = {
902 { sizeof(HE), 0, 0, SVt_NULL,
903 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
905 /* The bind placeholder pretends to be an RV for now.
906 Also it's marked as "can't upgrade" to stop anyone using it before it's
908 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
910 /* IVs are in the head, so the allocation size is 0. */
912 sizeof(IV), /* This is used to copy out the IV body. */
913 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
914 NOARENA /* IVS don't need an arena */, 0
917 /* 8 bytes on most ILP32 with IEEE doubles */
918 { sizeof(NV), sizeof(NV),
919 STRUCT_OFFSET(XPVNV, xnv_u),
920 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
922 /* 8 bytes on most ILP32 with IEEE doubles */
924 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
925 + STRUCT_OFFSET(XPV, xpv_cur),
926 SVt_PV, FALSE, NONV, HASARENA,
927 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
929 #if 2 *PTRSIZE <= IVSIZE
932 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
933 + STRUCT_OFFSET(XPV, xpv_cur),
934 SVt_PVIV, FALSE, NONV, HASARENA,
935 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
939 copy_length(XPVIV, xiv_u),
941 SVt_PVIV, FALSE, NONV, HASARENA,
942 FIT_ARENA(0, sizeof(XPVIV)) },
945 #if (2 *PTRSIZE <= IVSIZE) && (2 *PTRSIZE <= NVSIZE)
948 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
949 + STRUCT_OFFSET(XPV, xpv_cur),
950 SVt_PVNV, FALSE, HADNV, HASARENA,
951 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
954 { sizeof(XPVNV), copy_length(XPVNV, xnv_u), 0, SVt_PVNV, FALSE, HADNV,
955 HASARENA, FIT_ARENA(0, sizeof(XPVNV)) },
959 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
960 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
966 SVt_REGEXP, FALSE, NONV, HASARENA,
967 FIT_ARENA(0, sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur))
971 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
972 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
975 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
976 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
979 copy_length(XPVAV, xav_alloc),
981 SVt_PVAV, TRUE, NONV, HASARENA,
982 FIT_ARENA(0, sizeof(XPVAV)) },
985 copy_length(XPVHV, xhv_max),
987 SVt_PVHV, TRUE, NONV, HASARENA,
988 FIT_ARENA(0, sizeof(XPVHV)) },
994 SVt_PVCV, TRUE, NONV, HASARENA,
995 FIT_ARENA(0, sizeof(XPVCV)) },
1000 SVt_PVFM, TRUE, NONV, NOARENA,
1001 FIT_ARENA(20, sizeof(XPVFM)) },
1003 /* XPVIO is 84 bytes, fits 48x */
1007 SVt_PVIO, TRUE, NONV, HASARENA,
1008 FIT_ARENA(24, sizeof(XPVIO)) },
1011 #define new_body_allocated(sv_type) \
1012 (void *)((char *)S_new_body(aTHX_ sv_type) \
1013 - bodies_by_type[sv_type].offset)
1015 #define del_body_allocated(p, sv_type) \
1016 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
1019 #define my_safemalloc(s) (void*)safemalloc(s)
1020 #define my_safecalloc(s) (void*)safecalloc(s, 1)
1021 #define my_safefree(p) safefree((char*)p)
1025 #define new_XNV() my_safemalloc(sizeof(XPVNV))
1026 #define del_XNV(p) my_safefree(p)
1028 #define new_XPVNV() my_safemalloc(sizeof(XPVNV))
1029 #define del_XPVNV(p) my_safefree(p)
1031 #define new_XPVAV() my_safemalloc(sizeof(XPVAV))
1032 #define del_XPVAV(p) my_safefree(p)
1034 #define new_XPVHV() my_safemalloc(sizeof(XPVHV))
1035 #define del_XPVHV(p) my_safefree(p)
1037 #define new_XPVMG() my_safemalloc(sizeof(XPVMG))
1038 #define del_XPVMG(p) my_safefree(p)
1040 #define new_XPVGV() my_safemalloc(sizeof(XPVGV))
1041 #define del_XPVGV(p) my_safefree(p)
1045 #define new_XNV() new_body_allocated(SVt_NV)
1046 #define del_XNV(p) del_body_allocated(p, SVt_NV)
1048 #define new_XPVNV() new_body_allocated(SVt_PVNV)
1049 #define del_XPVNV(p) del_body_allocated(p, SVt_PVNV)
1051 #define new_XPVAV() new_body_allocated(SVt_PVAV)
1052 #define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
1054 #define new_XPVHV() new_body_allocated(SVt_PVHV)
1055 #define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
1057 #define new_XPVMG() new_body_allocated(SVt_PVMG)
1058 #define del_XPVMG(p) del_body_allocated(p, SVt_PVMG)
1060 #define new_XPVGV() new_body_allocated(SVt_PVGV)
1061 #define del_XPVGV(p) del_body_allocated(p, SVt_PVGV)
1065 /* no arena for you! */
1067 #define new_NOARENA(details) \
1068 my_safemalloc((details)->body_size + (details)->offset)
1069 #define new_NOARENAZ(details) \
1070 my_safecalloc((details)->body_size + (details)->offset)
1073 S_more_bodies (pTHX_ const svtype sv_type)
1076 void ** const root = &PL_body_roots[sv_type];
1077 const struct body_details * const bdp = &bodies_by_type[sv_type];
1078 const size_t body_size = bdp->body_size;
1081 const size_t arena_size = Perl_malloc_good_size(bdp->arena_size);
1082 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1083 static bool done_sanity_check;
1085 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1086 * variables like done_sanity_check. */
1087 if (!done_sanity_check) {
1088 unsigned int i = SVt_LAST;
1090 done_sanity_check = TRUE;
1093 assert (bodies_by_type[i].type == i);
1097 assert(bdp->arena_size);
1099 start = (char*) Perl_get_arena(aTHX_ arena_size, sv_type);
1101 end = start + arena_size - 2 * body_size;
1103 /* computed count doesnt reflect the 1st slot reservation */
1104 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1105 DEBUG_m(PerlIO_printf(Perl_debug_log,
1106 "arena %p end %p arena-size %d (from %d) type %d "
1108 (void*)start, (void*)end, (int)arena_size,
1109 (int)bdp->arena_size, sv_type, (int)body_size,
1110 (int)arena_size / (int)body_size));
1112 DEBUG_m(PerlIO_printf(Perl_debug_log,
1113 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1114 (void*)start, (void*)end,
1115 (int)bdp->arena_size, sv_type, (int)body_size,
1116 (int)bdp->arena_size / (int)body_size));
1118 *root = (void *)start;
1120 while (start <= end) {
1121 char * const next = start + body_size;
1122 *(void**) start = (void *)next;
1125 *(void **)start = 0;
1130 /* grab a new thing from the free list, allocating more if necessary.
1131 The inline version is used for speed in hot routines, and the
1132 function using it serves the rest (unless PURIFY).
1134 #define new_body_inline(xpv, sv_type) \
1136 void ** const r3wt = &PL_body_roots[sv_type]; \
1137 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1138 ? *((void **)(r3wt)) : more_bodies(sv_type)); \
1139 *(r3wt) = *(void**)(xpv); \
1145 S_new_body(pTHX_ const svtype sv_type)
1149 new_body_inline(xpv, sv_type);
1155 static const struct body_details fake_rv =
1156 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1159 =for apidoc sv_upgrade
1161 Upgrade an SV to a more complex form. Generally adds a new body type to the
1162 SV, then copies across as much information as possible from the old body.
1163 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1169 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1174 const svtype old_type = SvTYPE(sv);
1175 const struct body_details *new_type_details;
1176 const struct body_details *old_type_details
1177 = bodies_by_type + old_type;
1178 SV *referant = NULL;
1180 PERL_ARGS_ASSERT_SV_UPGRADE;
1182 if (old_type == new_type)
1185 /* This clause was purposefully added ahead of the early return above to
1186 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1187 inference by Nick I-S that it would fix other troublesome cases. See
1188 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1190 Given that shared hash key scalars are no longer PVIV, but PV, there is
1191 no longer need to unshare so as to free up the IVX slot for its proper
1192 purpose. So it's safe to move the early return earlier. */
1194 if (new_type != SVt_PV && SvIsCOW(sv)) {
1195 sv_force_normal_flags(sv, 0);
1198 old_body = SvANY(sv);
1200 /* Copying structures onto other structures that have been neatly zeroed
1201 has a subtle gotcha. Consider XPVMG
1203 +------+------+------+------+------+-------+-------+
1204 | NV | CUR | LEN | IV | MAGIC | STASH |
1205 +------+------+------+------+------+-------+-------+
1206 0 4 8 12 16 20 24 28
1208 where NVs are aligned to 8 bytes, so that sizeof that structure is
1209 actually 32 bytes long, with 4 bytes of padding at the end:
1211 +------+------+------+------+------+-------+-------+------+
1212 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1213 +------+------+------+------+------+-------+-------+------+
1214 0 4 8 12 16 20 24 28 32
1216 so what happens if you allocate memory for this structure:
1218 +------+------+------+------+------+-------+-------+------+------+...
1219 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1220 +------+------+------+------+------+-------+-------+------+------+...
1221 0 4 8 12 16 20 24 28 32 36
1223 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1224 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1225 started out as zero once, but it's quite possible that it isn't. So now,
1226 rather than a nicely zeroed GP, you have it pointing somewhere random.
1229 (In fact, GP ends up pointing at a previous GP structure, because the
1230 principle cause of the padding in XPVMG getting garbage is a copy of
1231 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1232 this happens to be moot because XPVGV has been re-ordered, with GP
1233 no longer after STASH)
1235 So we are careful and work out the size of used parts of all the
1243 referant = SvRV(sv);
1244 old_type_details = &fake_rv;
1245 if (new_type == SVt_NV)
1246 new_type = SVt_PVNV;
1248 if (new_type < SVt_PVIV) {
1249 new_type = (new_type == SVt_NV)
1250 ? SVt_PVNV : SVt_PVIV;
1255 if (new_type < SVt_PVNV) {
1256 new_type = SVt_PVNV;
1260 assert(new_type > SVt_PV);
1261 assert(SVt_IV < SVt_PV);
1262 assert(SVt_NV < SVt_PV);
1269 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1270 there's no way that it can be safely upgraded, because perl.c
1271 expects to Safefree(SvANY(PL_mess_sv)) */
1272 assert(sv != PL_mess_sv);
1273 /* This flag bit is used to mean other things in other scalar types.
1274 Given that it only has meaning inside the pad, it shouldn't be set
1275 on anything that can get upgraded. */
1276 assert(!SvPAD_TYPED(sv));
1279 if (old_type_details->cant_upgrade)
1280 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1281 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1284 if (old_type > new_type)
1285 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1286 (int)old_type, (int)new_type);
1288 new_type_details = bodies_by_type + new_type;
1290 SvFLAGS(sv) &= ~SVTYPEMASK;
1291 SvFLAGS(sv) |= new_type;
1293 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1294 the return statements above will have triggered. */
1295 assert (new_type != SVt_NULL);
1298 assert(old_type == SVt_NULL);
1299 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1303 assert(old_type == SVt_NULL);
1304 SvANY(sv) = new_XNV();
1309 assert(new_type_details->body_size);
1312 assert(new_type_details->arena);
1313 assert(new_type_details->arena_size);
1314 /* This points to the start of the allocated area. */
1315 new_body_inline(new_body, new_type);
1316 Zero(new_body, new_type_details->body_size, char);
1317 new_body = ((char *)new_body) - new_type_details->offset;
1319 /* We always allocated the full length item with PURIFY. To do this
1320 we fake things so that arena is false for all 16 types.. */
1321 new_body = new_NOARENAZ(new_type_details);
1323 SvANY(sv) = new_body;
1324 if (new_type == SVt_PVAV) {
1328 if (old_type_details->body_size) {
1331 /* It will have been zeroed when the new body was allocated.
1332 Lets not write to it, in case it confuses a write-back
1338 #ifndef NODEFAULT_SHAREKEYS
1339 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1341 HvMAX(sv) = 7; /* (start with 8 buckets) */
1344 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1345 The target created by newSVrv also is, and it can have magic.
1346 However, it never has SvPVX set.
1348 if (old_type == SVt_IV) {
1350 } else if (old_type >= SVt_PV) {
1351 assert(SvPVX_const(sv) == 0);
1354 if (old_type >= SVt_PVMG) {
1355 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1356 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1358 sv->sv_u.svu_array = NULL; /* or svu_hash */
1364 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1365 sv_force_normal_flags(sv) is called. */
1368 /* XXX Is this still needed? Was it ever needed? Surely as there is
1369 no route from NV to PVIV, NOK can never be true */
1370 assert(!SvNOKp(sv));
1381 assert(new_type_details->body_size);
1382 /* We always allocated the full length item with PURIFY. To do this
1383 we fake things so that arena is false for all 16 types.. */
1384 if(new_type_details->arena) {
1385 /* This points to the start of the allocated area. */
1386 new_body_inline(new_body, new_type);
1387 Zero(new_body, new_type_details->body_size, char);
1388 new_body = ((char *)new_body) - new_type_details->offset;
1390 new_body = new_NOARENAZ(new_type_details);
1392 SvANY(sv) = new_body;
1394 if (old_type_details->copy) {
1395 /* There is now the potential for an upgrade from something without
1396 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1397 int offset = old_type_details->offset;
1398 int length = old_type_details->copy;
1400 if (new_type_details->offset > old_type_details->offset) {
1401 const int difference
1402 = new_type_details->offset - old_type_details->offset;
1403 offset += difference;
1404 length -= difference;
1406 assert (length >= 0);
1408 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1412 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1413 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1414 * correct 0.0 for us. Otherwise, if the old body didn't have an
1415 * NV slot, but the new one does, then we need to initialise the
1416 * freshly created NV slot with whatever the correct bit pattern is
1418 if (old_type_details->zero_nv && !new_type_details->zero_nv
1419 && !isGV_with_GP(sv))
1423 if (new_type == SVt_PVIO) {
1424 IO * const io = MUTABLE_IO(sv);
1425 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1428 /* Clear the stashcache because a new IO could overrule a package
1430 hv_clear(PL_stashcache);
1432 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1433 IoPAGE_LEN(sv) = 60;
1435 if (old_type < SVt_PV) {
1436 /* referant will be NULL unless the old type was SVt_IV emulating
1438 sv->sv_u.svu_rv = referant;
1442 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1443 (unsigned long)new_type);
1446 if (old_type > SVt_IV) {
1448 my_safefree(old_body);
1450 /* Note that there is an assumption that all bodies of types that
1451 can be upgraded came from arenas. Only the more complex non-
1452 upgradable types are allowed to be directly malloc()ed. */
1453 assert(old_type_details->arena);
1454 del_body((void*)((char*)old_body + old_type_details->offset),
1455 &PL_body_roots[old_type]);
1461 =for apidoc sv_backoff
1463 Remove any string offset. You should normally use the C<SvOOK_off> macro
1470 Perl_sv_backoff(pTHX_ register SV *const sv)
1473 const char * const s = SvPVX_const(sv);
1475 PERL_ARGS_ASSERT_SV_BACKOFF;
1476 PERL_UNUSED_CONTEXT;
1479 assert(SvTYPE(sv) != SVt_PVHV);
1480 assert(SvTYPE(sv) != SVt_PVAV);
1482 SvOOK_offset(sv, delta);
1484 SvLEN_set(sv, SvLEN(sv) + delta);
1485 SvPV_set(sv, SvPVX(sv) - delta);
1486 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1487 SvFLAGS(sv) &= ~SVf_OOK;
1494 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1495 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1496 Use the C<SvGROW> wrapper instead.
1502 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1506 PERL_ARGS_ASSERT_SV_GROW;
1508 if (PL_madskills && newlen >= 0x100000) {
1509 PerlIO_printf(Perl_debug_log,
1510 "Allocation too large: %"UVxf"\n", (UV)newlen);
1512 #ifdef HAS_64K_LIMIT
1513 if (newlen >= 0x10000) {
1514 PerlIO_printf(Perl_debug_log,
1515 "Allocation too large: %"UVxf"\n", (UV)newlen);
1518 #endif /* HAS_64K_LIMIT */
1521 if (SvTYPE(sv) < SVt_PV) {
1522 sv_upgrade(sv, SVt_PV);
1523 s = SvPVX_mutable(sv);
1525 else if (SvOOK(sv)) { /* pv is offset? */
1527 s = SvPVX_mutable(sv);
1528 if (newlen > SvLEN(sv))
1529 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1530 #ifdef HAS_64K_LIMIT
1531 if (newlen >= 0x10000)
1536 s = SvPVX_mutable(sv);
1538 if (newlen > SvLEN(sv)) { /* need more room? */
1539 #ifndef Perl_safesysmalloc_size
1540 newlen = PERL_STRLEN_ROUNDUP(newlen);
1542 if (SvLEN(sv) && s) {
1543 s = (char*)saferealloc(s, newlen);
1546 s = (char*)safemalloc(newlen);
1547 if (SvPVX_const(sv) && SvCUR(sv)) {
1548 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1552 #ifdef Perl_safesysmalloc_size
1553 /* Do this here, do it once, do it right, and then we will never get
1554 called back into sv_grow() unless there really is some growing
1556 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1558 SvLEN_set(sv, newlen);
1565 =for apidoc sv_setiv
1567 Copies an integer into the given SV, upgrading first if necessary.
1568 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1574 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1578 PERL_ARGS_ASSERT_SV_SETIV;
1580 SV_CHECK_THINKFIRST_COW_DROP(sv);
1581 switch (SvTYPE(sv)) {
1584 sv_upgrade(sv, SVt_IV);
1587 sv_upgrade(sv, SVt_PVIV);
1591 if (!isGV_with_GP(sv))
1598 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1602 (void)SvIOK_only(sv); /* validate number */
1608 =for apidoc sv_setiv_mg
1610 Like C<sv_setiv>, but also handles 'set' magic.
1616 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1618 PERL_ARGS_ASSERT_SV_SETIV_MG;
1625 =for apidoc sv_setuv
1627 Copies an unsigned integer into the given SV, upgrading first if necessary.
1628 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1634 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1636 PERL_ARGS_ASSERT_SV_SETUV;
1638 /* With these two if statements:
1639 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1642 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1644 If you wish to remove them, please benchmark to see what the effect is
1646 if (u <= (UV)IV_MAX) {
1647 sv_setiv(sv, (IV)u);
1656 =for apidoc sv_setuv_mg
1658 Like C<sv_setuv>, but also handles 'set' magic.
1664 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1666 PERL_ARGS_ASSERT_SV_SETUV_MG;
1673 =for apidoc sv_setnv
1675 Copies a double into the given SV, upgrading first if necessary.
1676 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1682 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1686 PERL_ARGS_ASSERT_SV_SETNV;
1688 SV_CHECK_THINKFIRST_COW_DROP(sv);
1689 switch (SvTYPE(sv)) {
1692 sv_upgrade(sv, SVt_NV);
1696 sv_upgrade(sv, SVt_PVNV);
1700 if (!isGV_with_GP(sv))
1707 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1712 (void)SvNOK_only(sv); /* validate number */
1717 =for apidoc sv_setnv_mg
1719 Like C<sv_setnv>, but also handles 'set' magic.
1725 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1727 PERL_ARGS_ASSERT_SV_SETNV_MG;
1733 /* Print an "isn't numeric" warning, using a cleaned-up,
1734 * printable version of the offending string
1738 S_not_a_number(pTHX_ SV *const sv)
1745 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1748 dsv = newSVpvs_flags("", SVs_TEMP);
1749 pv = sv_uni_display(dsv, sv, 10, 0);
1752 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1753 /* each *s can expand to 4 chars + "...\0",
1754 i.e. need room for 8 chars */
1756 const char *s = SvPVX_const(sv);
1757 const char * const end = s + SvCUR(sv);
1758 for ( ; s < end && d < limit; s++ ) {
1760 if (ch & 128 && !isPRINT_LC(ch)) {
1769 else if (ch == '\r') {
1773 else if (ch == '\f') {
1777 else if (ch == '\\') {
1781 else if (ch == '\0') {
1785 else if (isPRINT_LC(ch))
1802 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1803 "Argument \"%s\" isn't numeric in %s", pv,
1806 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1807 "Argument \"%s\" isn't numeric", pv);
1811 =for apidoc looks_like_number
1813 Test if the content of an SV looks like a number (or is a number).
1814 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1815 non-numeric warning), even if your atof() doesn't grok them.
1821 Perl_looks_like_number(pTHX_ SV *const sv)
1823 register const char *sbegin;
1826 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1829 sbegin = SvPVX_const(sv);
1832 else if (SvPOKp(sv))
1833 sbegin = SvPV_const(sv, len);
1835 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1836 return grok_number(sbegin, len, NULL);
1840 S_glob_2number(pTHX_ GV * const gv)
1842 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1843 SV *const buffer = sv_newmortal();
1845 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1847 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1850 gv_efullname3(buffer, gv, "*");
1851 SvFLAGS(gv) |= wasfake;
1853 /* We know that all GVs stringify to something that is not-a-number,
1854 so no need to test that. */
1855 if (ckWARN(WARN_NUMERIC))
1856 not_a_number(buffer);
1857 /* We just want something true to return, so that S_sv_2iuv_common
1858 can tail call us and return true. */
1862 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1863 until proven guilty, assume that things are not that bad... */
1868 As 64 bit platforms often have an NV that doesn't preserve all bits of
1869 an IV (an assumption perl has been based on to date) it becomes necessary
1870 to remove the assumption that the NV always carries enough precision to
1871 recreate the IV whenever needed, and that the NV is the canonical form.
1872 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1873 precision as a side effect of conversion (which would lead to insanity
1874 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1875 1) to distinguish between IV/UV/NV slots that have cached a valid
1876 conversion where precision was lost and IV/UV/NV slots that have a
1877 valid conversion which has lost no precision
1878 2) to ensure that if a numeric conversion to one form is requested that
1879 would lose precision, the precise conversion (or differently
1880 imprecise conversion) is also performed and cached, to prevent
1881 requests for different numeric formats on the same SV causing
1882 lossy conversion chains. (lossless conversion chains are perfectly
1887 SvIOKp is true if the IV slot contains a valid value
1888 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1889 SvNOKp is true if the NV slot contains a valid value
1890 SvNOK is true only if the NV value is accurate
1893 while converting from PV to NV, check to see if converting that NV to an
1894 IV(or UV) would lose accuracy over a direct conversion from PV to
1895 IV(or UV). If it would, cache both conversions, return NV, but mark
1896 SV as IOK NOKp (ie not NOK).
1898 While converting from PV to IV, check to see if converting that IV to an
1899 NV would lose accuracy over a direct conversion from PV to NV. If it
1900 would, cache both conversions, flag similarly.
1902 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1903 correctly because if IV & NV were set NV *always* overruled.
1904 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1905 changes - now IV and NV together means that the two are interchangeable:
1906 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1908 The benefit of this is that operations such as pp_add know that if
1909 SvIOK is true for both left and right operands, then integer addition
1910 can be used instead of floating point (for cases where the result won't
1911 overflow). Before, floating point was always used, which could lead to
1912 loss of precision compared with integer addition.
1914 * making IV and NV equal status should make maths accurate on 64 bit
1916 * may speed up maths somewhat if pp_add and friends start to use
1917 integers when possible instead of fp. (Hopefully the overhead in
1918 looking for SvIOK and checking for overflow will not outweigh the
1919 fp to integer speedup)
1920 * will slow down integer operations (callers of SvIV) on "inaccurate"
1921 values, as the change from SvIOK to SvIOKp will cause a call into
1922 sv_2iv each time rather than a macro access direct to the IV slot
1923 * should speed up number->string conversion on integers as IV is
1924 favoured when IV and NV are equally accurate
1926 ####################################################################
1927 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1928 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1929 On the other hand, SvUOK is true iff UV.
1930 ####################################################################
1932 Your mileage will vary depending your CPU's relative fp to integer
1936 #ifndef NV_PRESERVES_UV
1937 # define IS_NUMBER_UNDERFLOW_IV 1
1938 # define IS_NUMBER_UNDERFLOW_UV 2
1939 # define IS_NUMBER_IV_AND_UV 2
1940 # define IS_NUMBER_OVERFLOW_IV 4
1941 # define IS_NUMBER_OVERFLOW_UV 5
1943 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1945 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1947 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1955 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1957 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));
1958 if (SvNVX(sv) < (NV)IV_MIN) {
1959 (void)SvIOKp_on(sv);
1961 SvIV_set(sv, IV_MIN);
1962 return IS_NUMBER_UNDERFLOW_IV;
1964 if (SvNVX(sv) > (NV)UV_MAX) {
1965 (void)SvIOKp_on(sv);
1968 SvUV_set(sv, UV_MAX);
1969 return IS_NUMBER_OVERFLOW_UV;
1971 (void)SvIOKp_on(sv);
1973 /* Can't use strtol etc to convert this string. (See truth table in
1975 if (SvNVX(sv) <= (UV)IV_MAX) {
1976 SvIV_set(sv, I_V(SvNVX(sv)));
1977 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1978 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1980 /* Integer is imprecise. NOK, IOKp */
1982 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1985 SvUV_set(sv, U_V(SvNVX(sv)));
1986 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1987 if (SvUVX(sv) == UV_MAX) {
1988 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1989 possibly be preserved by NV. Hence, it must be overflow.
1991 return IS_NUMBER_OVERFLOW_UV;
1993 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1995 /* Integer is imprecise. NOK, IOKp */
1997 return IS_NUMBER_OVERFLOW_IV;
1999 #endif /* !NV_PRESERVES_UV*/
2002 S_sv_2iuv_common(pTHX_ SV *const sv)
2006 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
2009 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
2010 * without also getting a cached IV/UV from it at the same time
2011 * (ie PV->NV conversion should detect loss of accuracy and cache
2012 * IV or UV at same time to avoid this. */
2013 /* IV-over-UV optimisation - choose to cache IV if possible */
2015 if (SvTYPE(sv) == SVt_NV)
2016 sv_upgrade(sv, SVt_PVNV);
2018 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
2019 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
2020 certainly cast into the IV range at IV_MAX, whereas the correct
2021 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
2023 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2024 if (Perl_isnan(SvNVX(sv))) {
2030 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2031 SvIV_set(sv, I_V(SvNVX(sv)));
2032 if (SvNVX(sv) == (NV) SvIVX(sv)
2033 #ifndef NV_PRESERVES_UV
2034 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2035 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2036 /* Don't flag it as "accurately an integer" if the number
2037 came from a (by definition imprecise) NV operation, and
2038 we're outside the range of NV integer precision */
2042 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2044 /* scalar has trailing garbage, eg "42a" */
2046 DEBUG_c(PerlIO_printf(Perl_debug_log,
2047 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2053 /* IV not precise. No need to convert from PV, as NV
2054 conversion would already have cached IV if it detected
2055 that PV->IV would be better than PV->NV->IV
2056 flags already correct - don't set public IOK. */
2057 DEBUG_c(PerlIO_printf(Perl_debug_log,
2058 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2063 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2064 but the cast (NV)IV_MIN rounds to a the value less (more
2065 negative) than IV_MIN which happens to be equal to SvNVX ??
2066 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2067 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2068 (NV)UVX == NVX are both true, but the values differ. :-(
2069 Hopefully for 2s complement IV_MIN is something like
2070 0x8000000000000000 which will be exact. NWC */
2073 SvUV_set(sv, U_V(SvNVX(sv)));
2075 (SvNVX(sv) == (NV) SvUVX(sv))
2076 #ifndef NV_PRESERVES_UV
2077 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2078 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2079 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2080 /* Don't flag it as "accurately an integer" if the number
2081 came from a (by definition imprecise) NV operation, and
2082 we're outside the range of NV integer precision */
2088 DEBUG_c(PerlIO_printf(Perl_debug_log,
2089 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2095 else if (SvPOKp(sv) && SvLEN(sv)) {
2097 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2098 /* We want to avoid a possible problem when we cache an IV/ a UV which
2099 may be later translated to an NV, and the resulting NV is not
2100 the same as the direct translation of the initial string
2101 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2102 be careful to ensure that the value with the .456 is around if the
2103 NV value is requested in the future).
2105 This means that if we cache such an IV/a UV, we need to cache the
2106 NV as well. Moreover, we trade speed for space, and do not
2107 cache the NV if we are sure it's not needed.
2110 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2111 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2112 == IS_NUMBER_IN_UV) {
2113 /* It's definitely an integer, only upgrade to PVIV */
2114 if (SvTYPE(sv) < SVt_PVIV)
2115 sv_upgrade(sv, SVt_PVIV);
2117 } else if (SvTYPE(sv) < SVt_PVNV)
2118 sv_upgrade(sv, SVt_PVNV);
2120 /* If NVs preserve UVs then we only use the UV value if we know that
2121 we aren't going to call atof() below. If NVs don't preserve UVs
2122 then the value returned may have more precision than atof() will
2123 return, even though value isn't perfectly accurate. */
2124 if ((numtype & (IS_NUMBER_IN_UV
2125 #ifdef NV_PRESERVES_UV
2128 )) == IS_NUMBER_IN_UV) {
2129 /* This won't turn off the public IOK flag if it was set above */
2130 (void)SvIOKp_on(sv);
2132 if (!(numtype & IS_NUMBER_NEG)) {
2134 if (value <= (UV)IV_MAX) {
2135 SvIV_set(sv, (IV)value);
2137 /* it didn't overflow, and it was positive. */
2138 SvUV_set(sv, value);
2142 /* 2s complement assumption */
2143 if (value <= (UV)IV_MIN) {
2144 SvIV_set(sv, -(IV)value);
2146 /* Too negative for an IV. This is a double upgrade, but
2147 I'm assuming it will be rare. */
2148 if (SvTYPE(sv) < SVt_PVNV)
2149 sv_upgrade(sv, SVt_PVNV);
2153 SvNV_set(sv, -(NV)value);
2154 SvIV_set(sv, IV_MIN);
2158 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2159 will be in the previous block to set the IV slot, and the next
2160 block to set the NV slot. So no else here. */
2162 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2163 != IS_NUMBER_IN_UV) {
2164 /* It wasn't an (integer that doesn't overflow the UV). */
2165 SvNV_set(sv, Atof(SvPVX_const(sv)));
2167 if (! numtype && ckWARN(WARN_NUMERIC))
2170 #if defined(USE_LONG_DOUBLE)
2171 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2172 PTR2UV(sv), SvNVX(sv)));
2174 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2175 PTR2UV(sv), SvNVX(sv)));
2178 #ifdef NV_PRESERVES_UV
2179 (void)SvIOKp_on(sv);
2181 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2182 SvIV_set(sv, I_V(SvNVX(sv)));
2183 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2186 NOOP; /* Integer is imprecise. NOK, IOKp */
2188 /* UV will not work better than IV */
2190 if (SvNVX(sv) > (NV)UV_MAX) {
2192 /* Integer is inaccurate. NOK, IOKp, is UV */
2193 SvUV_set(sv, UV_MAX);
2195 SvUV_set(sv, U_V(SvNVX(sv)));
2196 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2197 NV preservse UV so can do correct comparison. */
2198 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2201 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2206 #else /* NV_PRESERVES_UV */
2207 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2208 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2209 /* The IV/UV slot will have been set from value returned by
2210 grok_number above. The NV slot has just been set using
2213 assert (SvIOKp(sv));
2215 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2216 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2217 /* Small enough to preserve all bits. */
2218 (void)SvIOKp_on(sv);
2220 SvIV_set(sv, I_V(SvNVX(sv)));
2221 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2223 /* Assumption: first non-preserved integer is < IV_MAX,
2224 this NV is in the preserved range, therefore: */
2225 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2227 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);
2231 0 0 already failed to read UV.
2232 0 1 already failed to read UV.
2233 1 0 you won't get here in this case. IV/UV
2234 slot set, public IOK, Atof() unneeded.
2235 1 1 already read UV.
2236 so there's no point in sv_2iuv_non_preserve() attempting
2237 to use atol, strtol, strtoul etc. */
2239 sv_2iuv_non_preserve (sv, numtype);
2241 sv_2iuv_non_preserve (sv);
2245 #endif /* NV_PRESERVES_UV */
2246 /* It might be more code efficient to go through the entire logic above
2247 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2248 gets complex and potentially buggy, so more programmer efficient
2249 to do it this way, by turning off the public flags: */
2251 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2255 if (isGV_with_GP(sv))
2256 return glob_2number(MUTABLE_GV(sv));
2258 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2259 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2262 if (SvTYPE(sv) < SVt_IV)
2263 /* Typically the caller expects that sv_any is not NULL now. */
2264 sv_upgrade(sv, SVt_IV);
2265 /* Return 0 from the caller. */
2272 =for apidoc sv_2iv_flags
2274 Return the integer value of an SV, doing any necessary string
2275 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2276 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2282 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2287 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2288 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2289 cache IVs just in case. In practice it seems that they never
2290 actually anywhere accessible by user Perl code, let alone get used
2291 in anything other than a string context. */
2292 if (flags & SV_GMAGIC)
2297 return I_V(SvNVX(sv));
2299 if (SvPOKp(sv) && SvLEN(sv)) {
2302 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2304 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2305 == IS_NUMBER_IN_UV) {
2306 /* It's definitely an integer */
2307 if (numtype & IS_NUMBER_NEG) {
2308 if (value < (UV)IV_MIN)
2311 if (value < (UV)IV_MAX)
2316 if (ckWARN(WARN_NUMERIC))
2319 return I_V(Atof(SvPVX_const(sv)));
2324 assert(SvTYPE(sv) >= SVt_PVMG);
2325 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2326 } else if (SvTHINKFIRST(sv)) {
2331 if (flags & SV_SKIP_OVERLOAD)
2333 tmpstr=AMG_CALLun(sv,numer);
2334 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2335 return SvIV(tmpstr);
2338 return PTR2IV(SvRV(sv));
2341 sv_force_normal_flags(sv, 0);
2343 if (SvREADONLY(sv) && !SvOK(sv)) {
2344 if (ckWARN(WARN_UNINITIALIZED))
2350 if (S_sv_2iuv_common(aTHX_ sv))
2353 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2354 PTR2UV(sv),SvIVX(sv)));
2355 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2359 =for apidoc sv_2uv_flags
2361 Return the unsigned integer value of an SV, doing any necessary string
2362 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2363 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2369 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2374 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2375 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2376 cache IVs just in case. */
2377 if (flags & SV_GMAGIC)
2382 return U_V(SvNVX(sv));
2383 if (SvPOKp(sv) && SvLEN(sv)) {
2386 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2388 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2389 == IS_NUMBER_IN_UV) {
2390 /* It's definitely an integer */
2391 if (!(numtype & IS_NUMBER_NEG))
2395 if (ckWARN(WARN_NUMERIC))
2398 return U_V(Atof(SvPVX_const(sv)));
2403 assert(SvTYPE(sv) >= SVt_PVMG);
2404 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2405 } else if (SvTHINKFIRST(sv)) {
2410 if (flags & SV_SKIP_OVERLOAD)
2412 tmpstr = AMG_CALLun(sv,numer);
2413 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2414 return SvUV(tmpstr);
2417 return PTR2UV(SvRV(sv));
2420 sv_force_normal_flags(sv, 0);
2422 if (SvREADONLY(sv) && !SvOK(sv)) {
2423 if (ckWARN(WARN_UNINITIALIZED))
2429 if (S_sv_2iuv_common(aTHX_ sv))
2433 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2434 PTR2UV(sv),SvUVX(sv)));
2435 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2439 =for apidoc sv_2nv_flags
2441 Return the num value of an SV, doing any necessary string or integer
2442 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2443 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2449 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2454 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2455 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2456 cache IVs just in case. */
2457 if (flags & SV_GMAGIC)
2461 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2462 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2463 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2465 return Atof(SvPVX_const(sv));
2469 return (NV)SvUVX(sv);
2471 return (NV)SvIVX(sv);
2476 assert(SvTYPE(sv) >= SVt_PVMG);
2477 /* This falls through to the report_uninit near the end of the
2479 } else if (SvTHINKFIRST(sv)) {
2484 if (flags & SV_SKIP_OVERLOAD)
2486 tmpstr = AMG_CALLun(sv,numer);
2487 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2488 return SvNV(tmpstr);
2491 return PTR2NV(SvRV(sv));
2494 sv_force_normal_flags(sv, 0);
2496 if (SvREADONLY(sv) && !SvOK(sv)) {
2497 if (ckWARN(WARN_UNINITIALIZED))
2502 if (SvTYPE(sv) < SVt_NV) {
2503 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2504 sv_upgrade(sv, SVt_NV);
2505 #ifdef USE_LONG_DOUBLE
2507 STORE_NUMERIC_LOCAL_SET_STANDARD();
2508 PerlIO_printf(Perl_debug_log,
2509 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2510 PTR2UV(sv), SvNVX(sv));
2511 RESTORE_NUMERIC_LOCAL();
2515 STORE_NUMERIC_LOCAL_SET_STANDARD();
2516 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2517 PTR2UV(sv), SvNVX(sv));
2518 RESTORE_NUMERIC_LOCAL();
2522 else if (SvTYPE(sv) < SVt_PVNV)
2523 sv_upgrade(sv, SVt_PVNV);
2528 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2529 #ifdef NV_PRESERVES_UV
2535 /* Only set the public NV OK flag if this NV preserves the IV */
2536 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2538 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2539 : (SvIVX(sv) == I_V(SvNVX(sv))))
2545 else if (SvPOKp(sv) && SvLEN(sv)) {
2547 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2548 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2550 #ifdef NV_PRESERVES_UV
2551 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2552 == IS_NUMBER_IN_UV) {
2553 /* It's definitely an integer */
2554 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2556 SvNV_set(sv, Atof(SvPVX_const(sv)));
2562 SvNV_set(sv, Atof(SvPVX_const(sv)));
2563 /* Only set the public NV OK flag if this NV preserves the value in
2564 the PV at least as well as an IV/UV would.
2565 Not sure how to do this 100% reliably. */
2566 /* if that shift count is out of range then Configure's test is
2567 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2569 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2570 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2571 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2572 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2573 /* Can't use strtol etc to convert this string, so don't try.
2574 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2577 /* value has been set. It may not be precise. */
2578 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2579 /* 2s complement assumption for (UV)IV_MIN */
2580 SvNOK_on(sv); /* Integer is too negative. */
2585 if (numtype & IS_NUMBER_NEG) {
2586 SvIV_set(sv, -(IV)value);
2587 } else if (value <= (UV)IV_MAX) {
2588 SvIV_set(sv, (IV)value);
2590 SvUV_set(sv, value);
2594 if (numtype & IS_NUMBER_NOT_INT) {
2595 /* I believe that even if the original PV had decimals,
2596 they are lost beyond the limit of the FP precision.
2597 However, neither is canonical, so both only get p
2598 flags. NWC, 2000/11/25 */
2599 /* Both already have p flags, so do nothing */
2601 const NV nv = SvNVX(sv);
2602 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2603 if (SvIVX(sv) == I_V(nv)) {
2606 /* It had no "." so it must be integer. */
2610 /* between IV_MAX and NV(UV_MAX).
2611 Could be slightly > UV_MAX */
2613 if (numtype & IS_NUMBER_NOT_INT) {
2614 /* UV and NV both imprecise. */
2616 const UV nv_as_uv = U_V(nv);
2618 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2627 /* It might be more code efficient to go through the entire logic above
2628 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2629 gets complex and potentially buggy, so more programmer efficient
2630 to do it this way, by turning off the public flags: */
2632 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2633 #endif /* NV_PRESERVES_UV */
2636 if (isGV_with_GP(sv)) {
2637 glob_2number(MUTABLE_GV(sv));
2641 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2643 assert (SvTYPE(sv) >= SVt_NV);
2644 /* Typically the caller expects that sv_any is not NULL now. */
2645 /* XXX Ilya implies that this is a bug in callers that assume this
2646 and ideally should be fixed. */
2649 #if defined(USE_LONG_DOUBLE)
2651 STORE_NUMERIC_LOCAL_SET_STANDARD();
2652 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2653 PTR2UV(sv), SvNVX(sv));
2654 RESTORE_NUMERIC_LOCAL();
2658 STORE_NUMERIC_LOCAL_SET_STANDARD();
2659 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2660 PTR2UV(sv), SvNVX(sv));
2661 RESTORE_NUMERIC_LOCAL();
2670 Return an SV with the numeric value of the source SV, doing any necessary
2671 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2672 access this function.
2678 Perl_sv_2num(pTHX_ register SV *const sv)
2680 PERL_ARGS_ASSERT_SV_2NUM;
2685 SV * const tmpsv = AMG_CALLun(sv,numer);
2686 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2687 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2688 return sv_2num(tmpsv);
2690 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2693 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2694 * UV as a string towards the end of buf, and return pointers to start and
2697 * We assume that buf is at least TYPE_CHARS(UV) long.
2701 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2703 char *ptr = buf + TYPE_CHARS(UV);
2704 char * const ebuf = ptr;
2707 PERL_ARGS_ASSERT_UIV_2BUF;
2719 *--ptr = '0' + (char)(uv % 10);
2728 =for apidoc sv_2pv_flags
2730 Returns a pointer to the string value of an SV, and sets *lp to its length.
2731 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2733 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2734 usually end up here too.
2740 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2750 if (SvGMAGICAL(sv)) {
2751 if (flags & SV_GMAGIC)
2756 if (flags & SV_MUTABLE_RETURN)
2757 return SvPVX_mutable(sv);
2758 if (flags & SV_CONST_RETURN)
2759 return (char *)SvPVX_const(sv);
2762 if (SvIOKp(sv) || SvNOKp(sv)) {
2763 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2768 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2769 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2771 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2778 #ifdef FIXNEGATIVEZERO
2779 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2785 SvUPGRADE(sv, SVt_PV);
2788 s = SvGROW_mutable(sv, len + 1);
2791 return (char*)memcpy(s, tbuf, len + 1);
2797 assert(SvTYPE(sv) >= SVt_PVMG);
2798 /* This falls through to the report_uninit near the end of the
2800 } else if (SvTHINKFIRST(sv)) {
2805 if (flags & SV_SKIP_OVERLOAD)
2807 tmpstr = AMG_CALLun(sv,string);
2808 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2809 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2811 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2815 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2816 if (flags & SV_CONST_RETURN) {
2817 pv = (char *) SvPVX_const(tmpstr);
2819 pv = (flags & SV_MUTABLE_RETURN)
2820 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2823 *lp = SvCUR(tmpstr);
2825 pv = sv_2pv_flags(tmpstr, lp, flags);
2838 SV *const referent = SvRV(sv);
2842 retval = buffer = savepvn("NULLREF", len);
2843 } else if (SvTYPE(referent) == SVt_REGEXP) {
2844 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2849 /* If the regex is UTF-8 we want the containing scalar to
2850 have an UTF-8 flag too */
2856 if ((seen_evals = RX_SEEN_EVALS(re)))
2857 PL_reginterp_cnt += seen_evals;
2860 *lp = RX_WRAPLEN(re);
2862 return RX_WRAPPED(re);
2864 const char *const typestr = sv_reftype(referent, 0);
2865 const STRLEN typelen = strlen(typestr);
2866 UV addr = PTR2UV(referent);
2867 const char *stashname = NULL;
2868 STRLEN stashnamelen = 0; /* hush, gcc */
2869 const char *buffer_end;
2871 if (SvOBJECT(referent)) {
2872 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2875 stashname = HEK_KEY(name);
2876 stashnamelen = HEK_LEN(name);
2878 if (HEK_UTF8(name)) {
2884 stashname = "__ANON__";
2887 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2888 + 2 * sizeof(UV) + 2 /* )\0 */;
2890 len = typelen + 3 /* (0x */
2891 + 2 * sizeof(UV) + 2 /* )\0 */;
2894 Newx(buffer, len, char);
2895 buffer_end = retval = buffer + len;
2897 /* Working backwards */
2901 *--retval = PL_hexdigit[addr & 15];
2902 } while (addr >>= 4);
2908 memcpy(retval, typestr, typelen);
2912 retval -= stashnamelen;
2913 memcpy(retval, stashname, stashnamelen);
2915 /* retval may not neccesarily have reached the start of the
2917 assert (retval >= buffer);
2919 len = buffer_end - retval - 1; /* -1 for that \0 */
2927 if (SvREADONLY(sv) && !SvOK(sv)) {
2930 if (flags & SV_UNDEF_RETURNS_NULL)
2932 if (ckWARN(WARN_UNINITIALIZED))
2937 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2938 /* I'm assuming that if both IV and NV are equally valid then
2939 converting the IV is going to be more efficient */
2940 const U32 isUIOK = SvIsUV(sv);
2941 char buf[TYPE_CHARS(UV)];
2945 if (SvTYPE(sv) < SVt_PVIV)
2946 sv_upgrade(sv, SVt_PVIV);
2947 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2949 /* inlined from sv_setpvn */
2950 s = SvGROW_mutable(sv, len + 1);
2951 Move(ptr, s, len, char);
2955 else if (SvNOKp(sv)) {
2957 if (SvTYPE(sv) < SVt_PVNV)
2958 sv_upgrade(sv, SVt_PVNV);
2959 /* The +20 is pure guesswork. Configure test needed. --jhi */
2960 s = SvGROW_mutable(sv, NV_DIG + 20);
2961 /* some Xenix systems wipe out errno here */
2963 if (SvNVX(sv) == 0.0)
2964 my_strlcpy(s, "0", SvLEN(sv));
2968 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2971 #ifdef FIXNEGATIVEZERO
2972 if (*s == '-' && s[1] == '0' && !s[2]) {
2984 if (isGV_with_GP(sv)) {
2985 GV *const gv = MUTABLE_GV(sv);
2986 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2987 SV *const buffer = sv_newmortal();
2989 /* FAKE globs can get coerced, so need to turn this off temporarily
2992 gv_efullname3(buffer, gv, "*");
2993 SvFLAGS(gv) |= wasfake;
2995 if (SvPOK(buffer)) {
2997 *lp = SvCUR(buffer);
2999 return SvPVX(buffer);
3010 if (flags & SV_UNDEF_RETURNS_NULL)
3012 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
3014 if (SvTYPE(sv) < SVt_PV)
3015 /* Typically the caller expects that sv_any is not NULL now. */
3016 sv_upgrade(sv, SVt_PV);
3020 const STRLEN len = s - SvPVX_const(sv);
3026 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3027 PTR2UV(sv),SvPVX_const(sv)));
3028 if (flags & SV_CONST_RETURN)
3029 return (char *)SvPVX_const(sv);
3030 if (flags & SV_MUTABLE_RETURN)
3031 return SvPVX_mutable(sv);
3036 =for apidoc sv_copypv
3038 Copies a stringified representation of the source SV into the
3039 destination SV. Automatically performs any necessary mg_get and
3040 coercion of numeric values into strings. Guaranteed to preserve
3041 UTF8 flag even from overloaded objects. Similar in nature to
3042 sv_2pv[_flags] but operates directly on an SV instead of just the
3043 string. Mostly uses sv_2pv_flags to do its work, except when that
3044 would lose the UTF-8'ness of the PV.
3050 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3053 const char * const s = SvPV_const(ssv,len);
3055 PERL_ARGS_ASSERT_SV_COPYPV;
3057 sv_setpvn(dsv,s,len);
3065 =for apidoc sv_2pvbyte
3067 Return a pointer to the byte-encoded representation of the SV, and set *lp
3068 to its length. May cause the SV to be downgraded from UTF-8 as a
3071 Usually accessed via the C<SvPVbyte> macro.
3077 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3079 PERL_ARGS_ASSERT_SV_2PVBYTE;
3081 sv_utf8_downgrade(sv,0);
3082 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3086 =for apidoc sv_2pvutf8
3088 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3089 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3091 Usually accessed via the C<SvPVutf8> macro.
3097 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3099 PERL_ARGS_ASSERT_SV_2PVUTF8;
3101 sv_utf8_upgrade(sv);
3102 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3107 =for apidoc sv_2bool
3109 This function is only called on magical items, and is only used by
3110 sv_true() or its macro equivalent.
3116 Perl_sv_2bool(pTHX_ register SV *const sv)
3120 PERL_ARGS_ASSERT_SV_2BOOL;
3128 SV * const tmpsv = AMG_CALLun(sv,bool_);
3129 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3130 return cBOOL(SvTRUE(tmpsv));
3132 return SvRV(sv) != 0;
3135 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3137 (*sv->sv_u.svu_pv > '0' ||
3138 Xpvtmp->xpv_cur > 1 ||
3139 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3146 return SvIVX(sv) != 0;
3149 return SvNVX(sv) != 0.0;
3151 if (isGV_with_GP(sv))
3161 =for apidoc sv_utf8_upgrade
3163 Converts the PV of an SV to its UTF-8-encoded form.
3164 Forces the SV to string form if it is not already.
3165 Will C<mg_get> on C<sv> if appropriate.
3166 Always sets the SvUTF8 flag to avoid future validity checks even
3167 if the whole string is the same in UTF-8 as not.
3168 Returns the number of bytes in the converted string
3170 This is not as a general purpose byte encoding to Unicode interface:
3171 use the Encode extension for that.
3173 =for apidoc sv_utf8_upgrade_nomg
3175 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3177 =for apidoc sv_utf8_upgrade_flags
3179 Converts the PV of an SV to its UTF-8-encoded form.
3180 Forces the SV to string form if it is not already.
3181 Always sets the SvUTF8 flag to avoid future validity checks even
3182 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3183 will C<mg_get> on C<sv> if appropriate, else not.
3184 Returns the number of bytes in the converted string
3185 C<sv_utf8_upgrade> and
3186 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3188 This is not as a general purpose byte encoding to Unicode interface:
3189 use the Encode extension for that.
3193 The grow version is currently not externally documented. It adds a parameter,
3194 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3195 have free after it upon return. This allows the caller to reserve extra space
3196 that it intends to fill, to avoid extra grows.
3198 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3199 which can be used to tell this function to not first check to see if there are
3200 any characters that are different in UTF-8 (variant characters) which would
3201 force it to allocate a new string to sv, but to assume there are. Typically
3202 this flag is used by a routine that has already parsed the string to find that
3203 there are such characters, and passes this information on so that the work
3204 doesn't have to be repeated.
3206 (One might think that the calling routine could pass in the position of the
3207 first such variant, so it wouldn't have to be found again. But that is not the
3208 case, because typically when the caller is likely to use this flag, it won't be
3209 calling this routine unless it finds something that won't fit into a byte.
3210 Otherwise it tries to not upgrade and just use bytes. But some things that
3211 do fit into a byte are variants in utf8, and the caller may not have been
3212 keeping track of these.)
3214 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3215 isn't guaranteed due to having other routines do the work in some input cases,
3216 or if the input is already flagged as being in utf8.
3218 The speed of this could perhaps be improved for many cases if someone wanted to
3219 write a fast function that counts the number of variant characters in a string,
3220 especially if it could return the position of the first one.
3225 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3229 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3231 if (sv == &PL_sv_undef)
3235 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3236 (void) sv_2pv_flags(sv,&len, flags);
3238 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3242 (void) SvPV_force(sv,len);
3247 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3252 sv_force_normal_flags(sv, 0);
3255 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3256 sv_recode_to_utf8(sv, PL_encoding);
3257 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3261 if (SvCUR(sv) == 0) {
3262 if (extra) SvGROW(sv, extra);
3263 } else { /* Assume Latin-1/EBCDIC */
3264 /* This function could be much more efficient if we
3265 * had a FLAG in SVs to signal if there are any variant
3266 * chars in the PV. Given that there isn't such a flag
3267 * make the loop as fast as possible (although there are certainly ways
3268 * to speed this up, eg. through vectorization) */
3269 U8 * s = (U8 *) SvPVX_const(sv);
3270 U8 * e = (U8 *) SvEND(sv);
3272 STRLEN two_byte_count = 0;
3274 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3276 /* See if really will need to convert to utf8. We mustn't rely on our
3277 * incoming SV being well formed and having a trailing '\0', as certain
3278 * code in pp_formline can send us partially built SVs. */
3282 if (NATIVE_IS_INVARIANT(ch)) continue;
3284 t--; /* t already incremented; re-point to first variant */
3289 /* utf8 conversion not needed because all are invariants. Mark as
3290 * UTF-8 even if no variant - saves scanning loop */
3296 /* Here, the string should be converted to utf8, either because of an
3297 * input flag (two_byte_count = 0), or because a character that
3298 * requires 2 bytes was found (two_byte_count = 1). t points either to
3299 * the beginning of the string (if we didn't examine anything), or to
3300 * the first variant. In either case, everything from s to t - 1 will
3301 * occupy only 1 byte each on output.
3303 * There are two main ways to convert. One is to create a new string
3304 * and go through the input starting from the beginning, appending each
3305 * converted value onto the new string as we go along. It's probably
3306 * best to allocate enough space in the string for the worst possible
3307 * case rather than possibly running out of space and having to
3308 * reallocate and then copy what we've done so far. Since everything
3309 * from s to t - 1 is invariant, the destination can be initialized
3310 * with these using a fast memory copy
3312 * The other way is to figure out exactly how big the string should be
3313 * by parsing the entire input. Then you don't have to make it big
3314 * enough to handle the worst possible case, and more importantly, if
3315 * the string you already have is large enough, you don't have to
3316 * allocate a new string, you can copy the last character in the input
3317 * string to the final position(s) that will be occupied by the
3318 * converted string and go backwards, stopping at t, since everything
3319 * before that is invariant.
3321 * There are advantages and disadvantages to each method.
3323 * In the first method, we can allocate a new string, do the memory
3324 * copy from the s to t - 1, and then proceed through the rest of the
3325 * string byte-by-byte.
3327 * In the second method, we proceed through the rest of the input
3328 * string just calculating how big the converted string will be. Then
3329 * there are two cases:
3330 * 1) if the string has enough extra space to handle the converted
3331 * value. We go backwards through the string, converting until we
3332 * get to the position we are at now, and then stop. If this
3333 * position is far enough along in the string, this method is
3334 * faster than the other method. If the memory copy were the same
3335 * speed as the byte-by-byte loop, that position would be about
3336 * half-way, as at the half-way mark, parsing to the end and back
3337 * is one complete string's parse, the same amount as starting
3338 * over and going all the way through. Actually, it would be
3339 * somewhat less than half-way, as it's faster to just count bytes
3340 * than to also copy, and we don't have the overhead of allocating
3341 * a new string, changing the scalar to use it, and freeing the
3342 * existing one. But if the memory copy is fast, the break-even
3343 * point is somewhere after half way. The counting loop could be
3344 * sped up by vectorization, etc, to move the break-even point
3345 * further towards the beginning.
3346 * 2) if the string doesn't have enough space to handle the converted
3347 * value. A new string will have to be allocated, and one might
3348 * as well, given that, start from the beginning doing the first
3349 * method. We've spent extra time parsing the string and in
3350 * exchange all we've gotten is that we know precisely how big to
3351 * make the new one. Perl is more optimized for time than space,
3352 * so this case is a loser.
3353 * So what I've decided to do is not use the 2nd method unless it is
3354 * guaranteed that a new string won't have to be allocated, assuming
3355 * the worst case. I also decided not to put any more conditions on it
3356 * than this, for now. It seems likely that, since the worst case is
3357 * twice as big as the unknown portion of the string (plus 1), we won't
3358 * be guaranteed enough space, causing us to go to the first method,
3359 * unless the string is short, or the first variant character is near
3360 * the end of it. In either of these cases, it seems best to use the
3361 * 2nd method. The only circumstance I can think of where this would
3362 * be really slower is if the string had once had much more data in it
3363 * than it does now, but there is still a substantial amount in it */
3366 STRLEN invariant_head = t - s;
3367 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3368 if (SvLEN(sv) < size) {
3370 /* Here, have decided to allocate a new string */
3375 Newx(dst, size, U8);
3377 /* If no known invariants at the beginning of the input string,
3378 * set so starts from there. Otherwise, can use memory copy to
3379 * get up to where we are now, and then start from here */
3381 if (invariant_head <= 0) {
3384 Copy(s, dst, invariant_head, char);
3385 d = dst + invariant_head;
3389 const UV uv = NATIVE8_TO_UNI(*t++);
3390 if (UNI_IS_INVARIANT(uv))
3391 *d++ = (U8)UNI_TO_NATIVE(uv);
3393 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3394 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3398 SvPV_free(sv); /* No longer using pre-existing string */
3399 SvPV_set(sv, (char*)dst);
3400 SvCUR_set(sv, d - dst);
3401 SvLEN_set(sv, size);
3404 /* Here, have decided to get the exact size of the string.
3405 * Currently this happens only when we know that there is
3406 * guaranteed enough space to fit the converted string, so
3407 * don't have to worry about growing. If two_byte_count is 0,
3408 * then t points to the first byte of the string which hasn't
3409 * been examined yet. Otherwise two_byte_count is 1, and t
3410 * points to the first byte in the string that will expand to
3411 * two. Depending on this, start examining at t or 1 after t.
3414 U8 *d = t + two_byte_count;
3417 /* Count up the remaining bytes that expand to two */
3420 const U8 chr = *d++;
3421 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3424 /* The string will expand by just the number of bytes that
3425 * occupy two positions. But we are one afterwards because of
3426 * the increment just above. This is the place to put the
3427 * trailing NUL, and to set the length before we decrement */
3429 d += two_byte_count;
3430 SvCUR_set(sv, d - s);
3434 /* Having decremented d, it points to the position to put the
3435 * very last byte of the expanded string. Go backwards through
3436 * the string, copying and expanding as we go, stopping when we
3437 * get to the part that is invariant the rest of the way down */
3441 const U8 ch = NATIVE8_TO_UNI(*e--);
3442 if (UNI_IS_INVARIANT(ch)) {
3443 *d-- = UNI_TO_NATIVE(ch);
3445 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3446 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3453 /* Mark as UTF-8 even if no variant - saves scanning loop */
3459 =for apidoc sv_utf8_downgrade
3461 Attempts to convert the PV of an SV from characters to bytes.
3462 If the PV contains a character that cannot fit
3463 in a byte, this conversion will fail;
3464 in this case, either returns false or, if C<fail_ok> is not
3467 This is not as a general purpose Unicode to byte encoding interface:
3468 use the Encode extension for that.
3474 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3478 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3480 if (SvPOKp(sv) && SvUTF8(sv)) {
3486 sv_force_normal_flags(sv, 0);
3488 s = (U8 *) SvPV(sv, len);
3489 if (!utf8_to_bytes(s, &len)) {
3494 Perl_croak(aTHX_ "Wide character in %s",
3497 Perl_croak(aTHX_ "Wide character");
3508 =for apidoc sv_utf8_encode
3510 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3511 flag off so that it looks like octets again.
3517 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3519 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3522 sv_force_normal_flags(sv, 0);
3524 if (SvREADONLY(sv)) {
3525 Perl_croak_no_modify(aTHX);
3527 (void) sv_utf8_upgrade(sv);
3532 =for apidoc sv_utf8_decode
3534 If the PV of the SV is an octet sequence in UTF-8
3535 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3536 so that it looks like a character. If the PV contains only single-byte
3537 characters, the C<SvUTF8> flag stays being off.
3538 Scans PV for validity and returns false if the PV is invalid UTF-8.
3544 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3546 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3552 /* The octets may have got themselves encoded - get them back as
3555 if (!sv_utf8_downgrade(sv, TRUE))
3558 /* it is actually just a matter of turning the utf8 flag on, but
3559 * we want to make sure everything inside is valid utf8 first.
3561 c = (const U8 *) SvPVX_const(sv);
3562 if (!is_utf8_string(c, SvCUR(sv)+1))
3564 e = (const U8 *) SvEND(sv);
3567 if (!UTF8_IS_INVARIANT(ch)) {
3577 =for apidoc sv_setsv
3579 Copies the contents of the source SV C<ssv> into the destination SV
3580 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3581 function if the source SV needs to be reused. Does not handle 'set' magic.
3582 Loosely speaking, it performs a copy-by-value, obliterating any previous
3583 content of the destination.
3585 You probably want to use one of the assortment of wrappers, such as
3586 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3587 C<SvSetMagicSV_nosteal>.
3589 =for apidoc sv_setsv_flags
3591 Copies the contents of the source SV C<ssv> into the destination SV
3592 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3593 function if the source SV needs to be reused. Does not handle 'set' magic.
3594 Loosely speaking, it performs a copy-by-value, obliterating any previous
3595 content of the destination.
3596 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3597 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3598 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3599 and C<sv_setsv_nomg> are implemented in terms of this function.
3601 You probably want to use one of the assortment of wrappers, such as
3602 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3603 C<SvSetMagicSV_nosteal>.
3605 This is the primary function for copying scalars, and most other
3606 copy-ish functions and macros use this underneath.
3612 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3614 I32 mro_changes = 0; /* 1 = method, 2 = isa */
3616 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3618 if (dtype != SVt_PVGV) {
3619 const char * const name = GvNAME(sstr);
3620 const STRLEN len = GvNAMELEN(sstr);
3622 if (dtype >= SVt_PV) {
3628 SvUPGRADE(dstr, SVt_PVGV);
3629 (void)SvOK_off(dstr);
3630 /* FIXME - why are we doing this, then turning it off and on again
3632 isGV_with_GP_on(dstr);
3634 GvSTASH(dstr) = GvSTASH(sstr);
3636 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3637 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3638 SvFAKE_on(dstr); /* can coerce to non-glob */
3641 if(GvGP(MUTABLE_GV(sstr))) {
3642 /* If source has method cache entry, clear it */
3644 SvREFCNT_dec(GvCV(sstr));
3648 /* If source has a real method, then a method is
3650 else if(GvCV((const GV *)sstr)) {
3655 /* If dest already had a real method, that's a change as well */
3656 if(!mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)) {
3660 if(strEQ(GvNAME((const GV *)dstr),"ISA"))
3663 gp_free(MUTABLE_GV(dstr));
3664 isGV_with_GP_off(dstr);
3665 (void)SvOK_off(dstr);
3666 isGV_with_GP_on(dstr);
3667 GvINTRO_off(dstr); /* one-shot flag */
3668 GvGP(dstr) = gp_ref(GvGP(sstr));
3669 if (SvTAINTED(sstr))
3671 if (GvIMPORTED(dstr) != GVf_IMPORTED
3672 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3674 GvIMPORTED_on(dstr);
3677 if(mro_changes == 2) mro_isa_changed_in(GvSTASH(dstr));
3678 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3683 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3685 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3687 const int intro = GvINTRO(dstr);
3690 const U32 stype = SvTYPE(sref);
3692 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3695 GvINTRO_off(dstr); /* one-shot flag */
3696 GvLINE(dstr) = CopLINE(PL_curcop);
3697 GvEGV(dstr) = MUTABLE_GV(dstr);
3702 location = (SV **) &GvCV(dstr);
3703 import_flag = GVf_IMPORTED_CV;
3706 location = (SV **) &GvHV(dstr);
3707 import_flag = GVf_IMPORTED_HV;
3710 location = (SV **) &GvAV(dstr);
3711 import_flag = GVf_IMPORTED_AV;
3714 location = (SV **) &GvIOp(dstr);
3717 location = (SV **) &GvFORM(dstr);
3720 location = &GvSV(dstr);
3721 import_flag = GVf_IMPORTED_SV;
3724 if (stype == SVt_PVCV) {
3725 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3726 if (GvCVGEN(dstr)) {
3727 SvREFCNT_dec(GvCV(dstr));
3729 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3732 SAVEGENERICSV(*location);
3736 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3737 CV* const cv = MUTABLE_CV(*location);
3739 if (!GvCVGEN((const GV *)dstr) &&
3740 (CvROOT(cv) || CvXSUB(cv)))
3742 /* Redefining a sub - warning is mandatory if
3743 it was a const and its value changed. */
3744 if (CvCONST(cv) && CvCONST((const CV *)sref)
3746 == cv_const_sv((const CV *)sref)) {
3748 /* They are 2 constant subroutines generated from
3749 the same constant. This probably means that
3750 they are really the "same" proxy subroutine
3751 instantiated in 2 places. Most likely this is
3752 when a constant is exported twice. Don't warn.
3755 else if (ckWARN(WARN_REDEFINE)
3757 && (!CvCONST((const CV *)sref)
3758 || sv_cmp(cv_const_sv(cv),
3759 cv_const_sv((const CV *)
3761 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3764 ? "Constant subroutine %s::%s redefined"
3765 : "Subroutine %s::%s redefined"),
3766 HvNAME_get(GvSTASH((const GV *)dstr)),
3767 GvENAME(MUTABLE_GV(dstr)));
3771 cv_ckproto_len(cv, (const GV *)dstr,
3772 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3773 SvPOK(sref) ? SvCUR(sref) : 0);
3775 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3776 GvASSUMECV_on(dstr);
3777 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3780 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3781 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3782 GvFLAGS(dstr) |= import_flag;
3784 if (stype == SVt_PVAV && strEQ(GvNAME((GV*)dstr), "ISA")) {
3785 sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3786 mro_isa_changed_in(GvSTASH(dstr));
3791 if (SvTAINTED(sstr))
3797 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3800 register U32 sflags;
3802 register svtype stype;
3804 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3809 if (SvIS_FREED(dstr)) {
3810 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3811 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3813 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3815 sstr = &PL_sv_undef;
3816 if (SvIS_FREED(sstr)) {
3817 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3818 (void*)sstr, (void*)dstr);
3820 stype = SvTYPE(sstr);
3821 dtype = SvTYPE(dstr);
3823 (void)SvAMAGIC_off(dstr);
3826 /* need to nuke the magic */
3830 /* There's a lot of redundancy below but we're going for speed here */
3835 if (dtype != SVt_PVGV) {
3836 (void)SvOK_off(dstr);
3844 sv_upgrade(dstr, SVt_IV);
3848 sv_upgrade(dstr, SVt_PVIV);
3851 goto end_of_first_switch;
3853 (void)SvIOK_only(dstr);
3854 SvIV_set(dstr, SvIVX(sstr));
3857 /* SvTAINTED can only be true if the SV has taint magic, which in
3858 turn means that the SV type is PVMG (or greater). This is the
3859 case statement for SVt_IV, so this cannot be true (whatever gcov
3861 assert(!SvTAINTED(sstr));
3866 if (dtype < SVt_PV && dtype != SVt_IV)
3867 sv_upgrade(dstr, SVt_IV);
3875 sv_upgrade(dstr, SVt_NV);
3879 sv_upgrade(dstr, SVt_PVNV);
3882 goto end_of_first_switch;
3884 SvNV_set(dstr, SvNVX(sstr));
3885 (void)SvNOK_only(dstr);
3886 /* SvTAINTED can only be true if the SV has taint magic, which in
3887 turn means that the SV type is PVMG (or greater). This is the
3888 case statement for SVt_NV, so this cannot be true (whatever gcov
3890 assert(!SvTAINTED(sstr));
3896 #ifdef PERL_OLD_COPY_ON_WRITE
3897 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3898 if (dtype < SVt_PVIV)
3899 sv_upgrade(dstr, SVt_PVIV);
3906 sv_upgrade(dstr, SVt_PV);
3909 if (dtype < SVt_PVIV)
3910 sv_upgrade(dstr, SVt_PVIV);
3913 if (dtype < SVt_PVNV)
3914 sv_upgrade(dstr, SVt_PVNV);
3918 const char * const type = sv_reftype(sstr,0);
3920 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
3922 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3927 if (dtype < SVt_REGEXP)
3928 sv_upgrade(dstr, SVt_REGEXP);
3931 /* case SVt_BIND: */
3934 if (isGV_with_GP(sstr) && dtype <= SVt_PVGV) {
3935 glob_assign_glob(dstr, sstr, dtype);
3938 /* SvVALID means that this PVGV is playing at being an FBM. */
3942 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
3944 if (SvTYPE(sstr) != stype) {
3945 stype = SvTYPE(sstr);
3946 if (isGV_with_GP(sstr) && stype == SVt_PVGV && dtype <= SVt_PVGV) {
3947 glob_assign_glob(dstr, sstr, dtype);
3952 if (stype == SVt_PVLV)
3953 SvUPGRADE(dstr, SVt_PVNV);
3955 SvUPGRADE(dstr, (svtype)stype);
3957 end_of_first_switch:
3959 /* dstr may have been upgraded. */
3960 dtype = SvTYPE(dstr);
3961 sflags = SvFLAGS(sstr);
3963 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
3964 /* Assigning to a subroutine sets the prototype. */
3967 const char *const ptr = SvPV_const(sstr, len);
3969 SvGROW(dstr, len + 1);
3970 Copy(ptr, SvPVX(dstr), len + 1, char);
3971 SvCUR_set(dstr, len);
3973 SvFLAGS(dstr) |= sflags & SVf_UTF8;
3977 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3978 const char * const type = sv_reftype(dstr,0);
3980 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
3982 Perl_croak(aTHX_ "Cannot copy to %s", type);
3983 } else if (sflags & SVf_ROK) {
3984 if (isGV_with_GP(dstr) && dtype == SVt_PVGV
3985 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
3988 if (GvIMPORTED(dstr) != GVf_IMPORTED
3989 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3991 GvIMPORTED_on(dstr);
3996 glob_assign_glob(dstr, sstr, dtype);
4000 if (dtype >= SVt_PV) {
4001 if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
4002 glob_assign_ref(dstr, sstr);
4005 if (SvPVX_const(dstr)) {
4011 (void)SvOK_off(dstr);
4012 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4013 SvFLAGS(dstr) |= sflags & SVf_ROK;
4014 assert(!(sflags & SVp_NOK));
4015 assert(!(sflags & SVp_IOK));
4016 assert(!(sflags & SVf_NOK));
4017 assert(!(sflags & SVf_IOK));
4019 else if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
4020 if (!(sflags & SVf_OK)) {
4021 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4022 "Undefined value assigned to typeglob");
4025 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4026 if (dstr != (const SV *)gv) {
4028 gp_free(MUTABLE_GV(dstr));
4029 GvGP(dstr) = gp_ref(GvGP(gv));
4033 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4034 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4036 else if (sflags & SVp_POK) {
4040 * Check to see if we can just swipe the string. If so, it's a
4041 * possible small lose on short strings, but a big win on long ones.
4042 * It might even be a win on short strings if SvPVX_const(dstr)
4043 * has to be allocated and SvPVX_const(sstr) has to be freed.
4044 * Likewise if we can set up COW rather than doing an actual copy, we
4045 * drop to the else clause, as the swipe code and the COW setup code
4046 * have much in common.
4049 /* Whichever path we take through the next code, we want this true,
4050 and doing it now facilitates the COW check. */
4051 (void)SvPOK_only(dstr);
4054 /* If we're already COW then this clause is not true, and if COW
4055 is allowed then we drop down to the else and make dest COW
4056 with us. If caller hasn't said that we're allowed to COW
4057 shared hash keys then we don't do the COW setup, even if the
4058 source scalar is a shared hash key scalar. */
4059 (((flags & SV_COW_SHARED_HASH_KEYS)
4060 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4061 : 1 /* If making a COW copy is forbidden then the behaviour we
4062 desire is as if the source SV isn't actually already
4063 COW, even if it is. So we act as if the source flags
4064 are not COW, rather than actually testing them. */
4066 #ifndef PERL_OLD_COPY_ON_WRITE
4067 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4068 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4069 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4070 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4071 but in turn, it's somewhat dead code, never expected to go
4072 live, but more kept as a placeholder on how to do it better
4073 in a newer implementation. */
4074 /* If we are COW and dstr is a suitable target then we drop down
4075 into the else and make dest a COW of us. */
4076 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4081 (sflags & SVs_TEMP) && /* slated for free anyway? */
4082 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4083 (!(flags & SV_NOSTEAL)) &&
4084 /* and we're allowed to steal temps */
4085 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4086 SvLEN(sstr)) /* and really is a string */
4087 #ifdef PERL_OLD_COPY_ON_WRITE
4088 && ((flags & SV_COW_SHARED_HASH_KEYS)
4089 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4090 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4091 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4095 /* Failed the swipe test, and it's not a shared hash key either.
4096 Have to copy the string. */
4097 STRLEN len = SvCUR(sstr);
4098 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4099 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4100 SvCUR_set(dstr, len);
4101 *SvEND(dstr) = '\0';
4103 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4105 /* Either it's a shared hash key, or it's suitable for
4106 copy-on-write or we can swipe the string. */
4108 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4112 #ifdef PERL_OLD_COPY_ON_WRITE
4114 if ((sflags & (SVf_FAKE | SVf_READONLY))
4115 != (SVf_FAKE | SVf_READONLY)) {
4116 SvREADONLY_on(sstr);
4118 /* Make the source SV into a loop of 1.
4119 (about to become 2) */
4120 SV_COW_NEXT_SV_SET(sstr, sstr);
4124 /* Initial code is common. */
4125 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4130 /* making another shared SV. */
4131 STRLEN cur = SvCUR(sstr);
4132 STRLEN len = SvLEN(sstr);
4133 #ifdef PERL_OLD_COPY_ON_WRITE
4135 assert (SvTYPE(dstr) >= SVt_PVIV);
4136 /* SvIsCOW_normal */
4137 /* splice us in between source and next-after-source. */
4138 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4139 SV_COW_NEXT_SV_SET(sstr, dstr);
4140 SvPV_set(dstr, SvPVX_mutable(sstr));
4144 /* SvIsCOW_shared_hash */
4145 DEBUG_C(PerlIO_printf(Perl_debug_log,
4146 "Copy on write: Sharing hash\n"));
4148 assert (SvTYPE(dstr) >= SVt_PV);
4150 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4152 SvLEN_set(dstr, len);
4153 SvCUR_set(dstr, cur);
4154 SvREADONLY_on(dstr);
4158 { /* Passes the swipe test. */
4159 SvPV_set(dstr, SvPVX_mutable(sstr));
4160 SvLEN_set(dstr, SvLEN(sstr));
4161 SvCUR_set(dstr, SvCUR(sstr));
4164 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4165 SvPV_set(sstr, NULL);
4171 if (sflags & SVp_NOK) {
4172 SvNV_set(dstr, SvNVX(sstr));
4174 if (sflags & SVp_IOK) {
4175 SvIV_set(dstr, SvIVX(sstr));
4176 /* Must do this otherwise some other overloaded use of 0x80000000
4177 gets confused. I guess SVpbm_VALID */
4178 if (sflags & SVf_IVisUV)
4181 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4183 const MAGIC * const smg = SvVSTRING_mg(sstr);
4185 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4186 smg->mg_ptr, smg->mg_len);
4187 SvRMAGICAL_on(dstr);
4191 else if (sflags & (SVp_IOK|SVp_NOK)) {
4192 (void)SvOK_off(dstr);
4193 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4194 if (sflags & SVp_IOK) {
4195 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4196 SvIV_set(dstr, SvIVX(sstr));
4198 if (sflags & SVp_NOK) {
4199 SvNV_set(dstr, SvNVX(sstr));
4203 if (isGV_with_GP(sstr)) {
4204 /* This stringification rule for globs is spread in 3 places.
4205 This feels bad. FIXME. */
4206 const U32 wasfake = sflags & SVf_FAKE;
4208 /* FAKE globs can get coerced, so need to turn this off
4209 temporarily if it is on. */
4211 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4212 SvFLAGS(sstr) |= wasfake;
4215 (void)SvOK_off(dstr);
4217 if (SvTAINTED(sstr))
4222 =for apidoc sv_setsv_mg
4224 Like C<sv_setsv>, but also handles 'set' magic.
4230 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4232 PERL_ARGS_ASSERT_SV_SETSV_MG;
4234 sv_setsv(dstr,sstr);
4238 #ifdef PERL_OLD_COPY_ON_WRITE
4240 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4242 STRLEN cur = SvCUR(sstr);
4243 STRLEN len = SvLEN(sstr);
4244 register char *new_pv;
4246 PERL_ARGS_ASSERT_SV_SETSV_COW;
4249 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4250 (void*)sstr, (void*)dstr);
4257 if (SvTHINKFIRST(dstr))
4258 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4259 else if (SvPVX_const(dstr))
4260 Safefree(SvPVX_const(dstr));
4264 SvUPGRADE(dstr, SVt_PVIV);
4266 assert (SvPOK(sstr));
4267 assert (SvPOKp(sstr));
4268 assert (!SvIOK(sstr));
4269 assert (!SvIOKp(sstr));
4270 assert (!SvNOK(sstr));
4271 assert (!SvNOKp(sstr));
4273 if (SvIsCOW(sstr)) {
4275 if (SvLEN(sstr) == 0) {
4276 /* source is a COW shared hash key. */
4277 DEBUG_C(PerlIO_printf(Perl_debug_log,
4278 "Fast copy on write: Sharing hash\n"));
4279 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4282 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4284 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4285 SvUPGRADE(sstr, SVt_PVIV);
4286 SvREADONLY_on(sstr);
4288 DEBUG_C(PerlIO_printf(Perl_debug_log,
4289 "Fast copy on write: Converting sstr to COW\n"));
4290 SV_COW_NEXT_SV_SET(dstr, sstr);
4292 SV_COW_NEXT_SV_SET(sstr, dstr);
4293 new_pv = SvPVX_mutable(sstr);
4296 SvPV_set(dstr, new_pv);
4297 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4300 SvLEN_set(dstr, len);
4301 SvCUR_set(dstr, cur);
4310 =for apidoc sv_setpvn
4312 Copies a string into an SV. The C<len> parameter indicates the number of
4313 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4314 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4320 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4323 register char *dptr;
4325 PERL_ARGS_ASSERT_SV_SETPVN;
4327 SV_CHECK_THINKFIRST_COW_DROP(sv);
4333 /* len is STRLEN which is unsigned, need to copy to signed */
4336 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4338 SvUPGRADE(sv, SVt_PV);
4340 dptr = SvGROW(sv, len + 1);
4341 Move(ptr,dptr,len,char);
4344 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4349 =for apidoc sv_setpvn_mg
4351 Like C<sv_setpvn>, but also handles 'set' magic.
4357 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4359 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4361 sv_setpvn(sv,ptr,len);
4366 =for apidoc sv_setpv
4368 Copies a string into an SV. The string must be null-terminated. Does not
4369 handle 'set' magic. See C<sv_setpv_mg>.
4375 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4378 register STRLEN len;
4380 PERL_ARGS_ASSERT_SV_SETPV;
4382 SV_CHECK_THINKFIRST_COW_DROP(sv);
4388 SvUPGRADE(sv, SVt_PV);
4390 SvGROW(sv, len + 1);
4391 Move(ptr,SvPVX(sv),len+1,char);
4393 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4398 =for apidoc sv_setpv_mg
4400 Like C<sv_setpv>, but also handles 'set' magic.
4406 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4408 PERL_ARGS_ASSERT_SV_SETPV_MG;
4415 =for apidoc sv_usepvn_flags
4417 Tells an SV to use C<ptr> to find its string value. Normally the
4418 string is stored inside the SV but sv_usepvn allows the SV to use an
4419 outside string. The C<ptr> should point to memory that was allocated
4420 by C<malloc>. The string length, C<len>, must be supplied. By default
4421 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4422 so that pointer should not be freed or used by the programmer after
4423 giving it to sv_usepvn, and neither should any pointers from "behind"
4424 that pointer (e.g. ptr + 1) be used.
4426 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4427 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4428 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4429 C<len>, and already meets the requirements for storing in C<SvPVX>)
4435 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4440 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4442 SV_CHECK_THINKFIRST_COW_DROP(sv);
4443 SvUPGRADE(sv, SVt_PV);
4446 if (flags & SV_SMAGIC)
4450 if (SvPVX_const(sv))
4454 if (flags & SV_HAS_TRAILING_NUL)
4455 assert(ptr[len] == '\0');
4458 allocate = (flags & SV_HAS_TRAILING_NUL)
4460 #ifdef Perl_safesysmalloc_size
4463 PERL_STRLEN_ROUNDUP(len + 1);
4465 if (flags & SV_HAS_TRAILING_NUL) {
4466 /* It's long enough - do nothing.
4467 Specfically Perl_newCONSTSUB is relying on this. */
4470 /* Force a move to shake out bugs in callers. */
4471 char *new_ptr = (char*)safemalloc(allocate);
4472 Copy(ptr, new_ptr, len, char);
4473 PoisonFree(ptr,len,char);
4477 ptr = (char*) saferealloc (ptr, allocate);
4480 #ifdef Perl_safesysmalloc_size
4481 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4483 SvLEN_set(sv, allocate);
4487 if (!(flags & SV_HAS_TRAILING_NUL)) {
4490 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4492 if (flags & SV_SMAGIC)
4496 #ifdef PERL_OLD_COPY_ON_WRITE
4497 /* Need to do this *after* making the SV normal, as we need the buffer
4498 pointer to remain valid until after we've copied it. If we let go too early,
4499 another thread could invalidate it by unsharing last of the same hash key
4500 (which it can do by means other than releasing copy-on-write Svs)
4501 or by changing the other copy-on-write SVs in the loop. */
4503 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4505 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4507 { /* this SV was SvIsCOW_normal(sv) */
4508 /* we need to find the SV pointing to us. */
4509 SV *current = SV_COW_NEXT_SV(after);
4511 if (current == sv) {
4512 /* The SV we point to points back to us (there were only two of us
4514 Hence other SV is no longer copy on write either. */
4516 SvREADONLY_off(after);
4518 /* We need to follow the pointers around the loop. */
4520 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4523 /* don't loop forever if the structure is bust, and we have
4524 a pointer into a closed loop. */
4525 assert (current != after);
4526 assert (SvPVX_const(current) == pvx);
4528 /* Make the SV before us point to the SV after us. */
4529 SV_COW_NEXT_SV_SET(current, after);
4535 =for apidoc sv_force_normal_flags
4537 Undo various types of fakery on an SV: if the PV is a shared string, make
4538 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4539 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4540 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4541 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4542 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4543 set to some other value.) In addition, the C<flags> parameter gets passed to
4544 C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4545 with flags set to 0.
4551 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4555 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4557 #ifdef PERL_OLD_COPY_ON_WRITE
4558 if (SvREADONLY(sv)) {
4560 const char * const pvx = SvPVX_const(sv);
4561 const STRLEN len = SvLEN(sv);
4562 const STRLEN cur = SvCUR(sv);
4563 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4564 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4565 we'll fail an assertion. */
4566 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4569 PerlIO_printf(Perl_debug_log,
4570 "Copy on write: Force normal %ld\n",
4576 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4579 if (flags & SV_COW_DROP_PV) {
4580 /* OK, so we don't need to copy our buffer. */
4583 SvGROW(sv, cur + 1);
4584 Move(pvx,SvPVX(sv),cur,char);
4589 sv_release_COW(sv, pvx, next);
4591 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4597 else if (IN_PERL_RUNTIME)
4598 Perl_croak_no_modify(aTHX);
4601 if (SvREADONLY(sv)) {
4603 const char * const pvx = SvPVX_const(sv);
4604 const STRLEN len = SvCUR(sv);
4609 SvGROW(sv, len + 1);
4610 Move(pvx,SvPVX(sv),len,char);
4612 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4614 else if (IN_PERL_RUNTIME)
4615 Perl_croak_no_modify(aTHX);
4619 sv_unref_flags(sv, flags);
4620 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4622 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4623 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analagous
4624 to sv_unglob. We only need it here, so inline it. */
4625 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4626 SV *const temp = newSV_type(new_type);
4627 void *const temp_p = SvANY(sv);
4629 if (new_type == SVt_PVMG) {
4630 SvMAGIC_set(temp, SvMAGIC(sv));
4631 SvMAGIC_set(sv, NULL);
4632 SvSTASH_set(temp, SvSTASH(sv));
4633 SvSTASH_set(sv, NULL);
4635 SvCUR_set(temp, SvCUR(sv));
4636 /* Remember that SvPVX is in the head, not the body. */
4638 SvLEN_set(temp, SvLEN(sv));
4639 /* This signals "buffer is owned by someone else" in sv_clear,
4640 which is the least effort way to stop it freeing the buffer.
4642 SvLEN_set(sv, SvLEN(sv)+1);
4644 /* Their buffer is already owned by someone else. */
4645 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4646 SvLEN_set(temp, SvCUR(sv)+1);
4649 /* Now swap the rest of the bodies. */
4651 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4652 SvFLAGS(sv) |= new_type;
4653 SvANY(sv) = SvANY(temp);
4655 SvFLAGS(temp) &= ~(SVTYPEMASK);
4656 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4657 SvANY(temp) = temp_p;
4666 Efficient removal of characters from the beginning of the string buffer.
4667 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4668 the string buffer. The C<ptr> becomes the first character of the adjusted
4669 string. Uses the "OOK hack".
4670 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4671 refer to the same chunk of data.
4677 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4683 const U8 *real_start;
4687 PERL_ARGS_ASSERT_SV_CHOP;
4689 if (!ptr || !SvPOKp(sv))
4691 delta = ptr - SvPVX_const(sv);
4693 /* Nothing to do. */
4696 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4697 nothing uses the value of ptr any more. */
4698 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4699 if (ptr <= SvPVX_const(sv))
4700 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4701 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4702 SV_CHECK_THINKFIRST(sv);
4703 if (delta > max_delta)
4704 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4705 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4706 SvPVX_const(sv) + max_delta);
4709 if (!SvLEN(sv)) { /* make copy of shared string */
4710 const char *pvx = SvPVX_const(sv);
4711 const STRLEN len = SvCUR(sv);
4712 SvGROW(sv, len + 1);
4713 Move(pvx,SvPVX(sv),len,char);
4716 SvFLAGS(sv) |= SVf_OOK;
4719 SvOOK_offset(sv, old_delta);
4721 SvLEN_set(sv, SvLEN(sv) - delta);
4722 SvCUR_set(sv, SvCUR(sv) - delta);
4723 SvPV_set(sv, SvPVX(sv) + delta);
4725 p = (U8 *)SvPVX_const(sv);
4730 real_start = p - delta;
4734 if (delta < 0x100) {
4738 p -= sizeof(STRLEN);
4739 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4743 /* Fill the preceding buffer with sentinals to verify that no-one is
4745 while (p > real_start) {
4753 =for apidoc sv_catpvn
4755 Concatenates the string onto the end of the string which is in the SV. The
4756 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4757 status set, then the bytes appended should be valid UTF-8.
4758 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4760 =for apidoc sv_catpvn_flags
4762 Concatenates the string onto the end of the string which is in the SV. The
4763 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4764 status set, then the bytes appended should be valid UTF-8.
4765 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4766 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4767 in terms of this function.
4773 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4777 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4779 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4781 SvGROW(dsv, dlen + slen + 1);
4783 sstr = SvPVX_const(dsv);
4784 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4785 SvCUR_set(dsv, SvCUR(dsv) + slen);
4787 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4789 if (flags & SV_SMAGIC)
4794 =for apidoc sv_catsv
4796 Concatenates the string from SV C<ssv> onto the end of the string in
4797 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4798 not 'set' magic. See C<sv_catsv_mg>.
4800 =for apidoc sv_catsv_flags
4802 Concatenates the string from SV C<ssv> onto the end of the string in
4803 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4804 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4805 and C<sv_catsv_nomg> are implemented in terms of this function.
4810 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4814 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4818 const char *spv = SvPV_const(ssv, slen);
4820 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4821 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4822 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4823 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4824 dsv->sv_flags doesn't have that bit set.
4825 Andy Dougherty 12 Oct 2001
4827 const I32 sutf8 = DO_UTF8(ssv);
4830 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4832 dutf8 = DO_UTF8(dsv);
4834 if (dutf8 != sutf8) {
4836 /* Not modifying source SV, so taking a temporary copy. */
4837 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
4839 sv_utf8_upgrade(csv);
4840 spv = SvPV_const(csv, slen);
4843 /* Leave enough space for the cat that's about to happen */
4844 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
4846 sv_catpvn_nomg(dsv, spv, slen);
4849 if (flags & SV_SMAGIC)
4854 =for apidoc sv_catpv
4856 Concatenates the string onto the end of the string which is in the SV.
4857 If the SV has the UTF-8 status set, then the bytes appended should be
4858 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
4863 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
4866 register STRLEN len;
4870 PERL_ARGS_ASSERT_SV_CATPV;
4874 junk = SvPV_force(sv, tlen);
4876 SvGROW(sv, tlen + len + 1);
4878 ptr = SvPVX_const(sv);
4879 Move(ptr,SvPVX(sv)+tlen,len+1,char);
4880 SvCUR_set(sv, SvCUR(sv) + len);
4881 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4886 =for apidoc sv_catpv_mg
4888 Like C<sv_catpv>, but also handles 'set' magic.
4894 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4896 PERL_ARGS_ASSERT_SV_CATPV_MG;
4905 Creates a new SV. A non-zero C<len> parameter indicates the number of
4906 bytes of preallocated string space the SV should have. An extra byte for a
4907 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4908 space is allocated.) The reference count for the new SV is set to 1.
4910 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4911 parameter, I<x>, a debug aid which allowed callers to identify themselves.
4912 This aid has been superseded by a new build option, PERL_MEM_LOG (see
4913 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4914 modules supporting older perls.
4920 Perl_newSV(pTHX_ const STRLEN len)
4927 sv_upgrade(sv, SVt_PV);
4928 SvGROW(sv, len + 1);
4933 =for apidoc sv_magicext
4935 Adds magic to an SV, upgrading it if necessary. Applies the
4936 supplied vtable and returns a pointer to the magic added.
4938 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4939 In particular, you can add magic to SvREADONLY SVs, and add more than
4940 one instance of the same 'how'.
4942 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4943 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4944 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4945 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
4947 (This is now used as a subroutine by C<sv_magic>.)
4952 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
4953 const MGVTBL *const vtable, const char *const name, const I32 namlen)
4958 PERL_ARGS_ASSERT_SV_MAGICEXT;
4960 SvUPGRADE(sv, SVt_PVMG);
4961 Newxz(mg, 1, MAGIC);
4962 mg->mg_moremagic = SvMAGIC(sv);
4963 SvMAGIC_set(sv, mg);
4965 /* Sometimes a magic contains a reference loop, where the sv and
4966 object refer to each other. To prevent a reference loop that
4967 would prevent such objects being freed, we look for such loops
4968 and if we find one we avoid incrementing the object refcount.
4970 Note we cannot do this to avoid self-tie loops as intervening RV must
4971 have its REFCNT incremented to keep it in existence.
4974 if (!obj || obj == sv ||
4975 how == PERL_MAGIC_arylen ||
4976 how == PERL_MAGIC_symtab ||
4977 (SvTYPE(obj) == SVt_PVGV &&
4978 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
4979 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
4980 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
4985 mg->mg_obj = SvREFCNT_inc_simple(obj);
4986 mg->mg_flags |= MGf_REFCOUNTED;
4989 /* Normal self-ties simply pass a null object, and instead of
4990 using mg_obj directly, use the SvTIED_obj macro to produce a
4991 new RV as needed. For glob "self-ties", we are tieing the PVIO
4992 with an RV obj pointing to the glob containing the PVIO. In
4993 this case, to avoid a reference loop, we need to weaken the
4997 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4998 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5004 mg->mg_len = namlen;
5007 mg->mg_ptr = savepvn(name, namlen);
5008 else if (namlen == HEf_SVKEY) {
5009 /* Yes, this is casting away const. This is only for the case of
5010 HEf_SVKEY. I think we need to document this abberation of the
5011 constness of the API, rather than making name non-const, as
5012 that change propagating outwards a long way. */
5013 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5015 mg->mg_ptr = (char *) name;
5017 mg->mg_virtual = (MGVTBL *) vtable;
5021 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5026 =for apidoc sv_magic
5028 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
5029 then adds a new magic item of type C<how> to the head of the magic list.
5031 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5032 handling of the C<name> and C<namlen> arguments.
5034 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5035 to add more than one instance of the same 'how'.
5041 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5042 const char *const name, const I32 namlen)
5045 const MGVTBL *vtable;
5048 PERL_ARGS_ASSERT_SV_MAGIC;
5050 #ifdef PERL_OLD_COPY_ON_WRITE
5052 sv_force_normal_flags(sv, 0);
5054 if (SvREADONLY(sv)) {
5056 /* its okay to attach magic to shared strings; the subsequent
5057 * upgrade to PVMG will unshare the string */
5058 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5061 && how != PERL_MAGIC_regex_global
5062 && how != PERL_MAGIC_bm
5063 && how != PERL_MAGIC_fm
5064 && how != PERL_MAGIC_sv
5065 && how != PERL_MAGIC_backref
5068 Perl_croak_no_modify(aTHX);
5071 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5072 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5073 /* sv_magic() refuses to add a magic of the same 'how' as an
5076 if (how == PERL_MAGIC_taint) {
5078 /* Any scalar which already had taint magic on which someone
5079 (erroneously?) did SvIOK_on() or similar will now be
5080 incorrectly sporting public "OK" flags. */
5081 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5089 vtable = &PL_vtbl_sv;
5091 case PERL_MAGIC_overload:
5092 vtable = &PL_vtbl_amagic;
5094 case PERL_MAGIC_overload_elem:
5095 vtable = &PL_vtbl_amagicelem;
5097 case PERL_MAGIC_overload_table:
5098 vtable = &PL_vtbl_ovrld;
5101 vtable = &PL_vtbl_bm;
5103 case PERL_MAGIC_regdata:
5104 vtable = &PL_vtbl_regdata;
5106 case PERL_MAGIC_regdatum:
5107 vtable = &PL_vtbl_regdatum;
5109 case PERL_MAGIC_env:
5110 vtable = &PL_vtbl_env;
5113 vtable = &PL_vtbl_fm;
5115 case PERL_MAGIC_envelem:
5116 vtable = &PL_vtbl_envelem;
5118 case PERL_MAGIC_regex_global:
5119 vtable = &PL_vtbl_mglob;
5121 case PERL_MAGIC_isa:
5122 vtable = &PL_vtbl_isa;
5124 case PERL_MAGIC_isaelem:
5125 vtable = &PL_vtbl_isaelem;
5127 case PERL_MAGIC_nkeys:
5128 vtable = &PL_vtbl_nkeys;
5130 case PERL_MAGIC_dbfile:
5133 case PERL_MAGIC_dbline:
5134 vtable = &PL_vtbl_dbline;
5136 #ifdef USE_LOCALE_COLLATE
5137 case PERL_MAGIC_collxfrm:
5138 vtable = &PL_vtbl_collxfrm;
5140 #endif /* USE_LOCALE_COLLATE */
5141 case PERL_MAGIC_tied:
5142 vtable = &PL_vtbl_pack;
5144 case PERL_MAGIC_tiedelem:
5145 case PERL_MAGIC_tiedscalar:
5146 vtable = &PL_vtbl_packelem;
5149 vtable = &PL_vtbl_regexp;
5151 case PERL_MAGIC_sig:
5152 vtable = &PL_vtbl_sig;
5154 case PERL_MAGIC_sigelem:
5155 vtable = &PL_vtbl_sigelem;
5157 case PERL_MAGIC_taint:
5158 vtable = &PL_vtbl_taint;
5160 case PERL_MAGIC_uvar:
5161 vtable = &PL_vtbl_uvar;
5163 case PERL_MAGIC_vec:
5164 vtable = &PL_vtbl_vec;
5166 case PERL_MAGIC_arylen_p:
5167 case PERL_MAGIC_rhash:
5168 case PERL_MAGIC_symtab:
5169 case PERL_MAGIC_vstring:
5172 case PERL_MAGIC_utf8:
5173 vtable = &PL_vtbl_utf8;
5175 case PERL_MAGIC_substr:
5176 vtable = &PL_vtbl_substr;
5178 case PERL_MAGIC_defelem:
5179 vtable = &PL_vtbl_defelem;
5181 case PERL_MAGIC_arylen:
5182 vtable = &PL_vtbl_arylen;
5184 case PERL_MAGIC_pos:
5185 vtable = &PL_vtbl_pos;
5187 case PERL_MAGIC_backref:
5188 vtable = &PL_vtbl_backref;
5190 case PERL_MAGIC_hintselem:
5191 vtable = &PL_vtbl_hintselem;
5193 case PERL_MAGIC_hints:
5194 vtable = &PL_vtbl_hints;
5196 case PERL_MAGIC_ext:
5197 /* Reserved for use by extensions not perl internals. */
5198 /* Useful for attaching extension internal data to perl vars. */
5199 /* Note that multiple extensions may clash if magical scalars */
5200 /* etc holding private data from one are passed to another. */
5204 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5207 /* Rest of work is done else where */
5208 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5211 case PERL_MAGIC_taint:
5214 case PERL_MAGIC_ext:
5215 case PERL_MAGIC_dbfile:
5222 =for apidoc sv_unmagic
5224 Removes all magic of type C<type> from an SV.
5230 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5235 PERL_ARGS_ASSERT_SV_UNMAGIC;
5237 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5239 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5240 for (mg = *mgp; mg; mg = *mgp) {
5241 if (mg->mg_type == type) {
5242 const MGVTBL* const vtbl = mg->mg_virtual;
5243 *mgp = mg->mg_moremagic;
5244 if (vtbl && vtbl->svt_free)
5245 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
5246 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5248 Safefree(mg->mg_ptr);
5249 else if (mg->mg_len == HEf_SVKEY)
5250 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5251 else if (mg->mg_type == PERL_MAGIC_utf8)
5252 Safefree(mg->mg_ptr);
5254 if (mg->mg_flags & MGf_REFCOUNTED)
5255 SvREFCNT_dec(mg->mg_obj);
5259 mgp = &mg->mg_moremagic;
5262 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5263 mg_magical(sv); /* else fix the flags now */
5267 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5273 =for apidoc sv_rvweaken
5275 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5276 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5277 push a back-reference to this RV onto the array of backreferences
5278 associated with that magic. If the RV is magical, set magic will be
5279 called after the RV is cleared.
5285 Perl_sv_rvweaken(pTHX_ SV *const sv)
5289 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5291 if (!SvOK(sv)) /* let undefs pass */
5294 Perl_croak(aTHX_ "Can't weaken a nonreference");
5295 else if (SvWEAKREF(sv)) {
5296 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5300 Perl_sv_add_backref(aTHX_ tsv, sv);
5306 /* Give tsv backref magic if it hasn't already got it, then push a
5307 * back-reference to sv onto the array associated with the backref magic.
5310 /* A discussion about the backreferences array and its refcount:
5312 * The AV holding the backreferences is pointed to either as the mg_obj of
5313 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5314 * xhv_backreferences field of the HvAUX structure. The array is created
5315 * with a refcount of 2. This means that if during global destruction the
5316 * array gets picked on before its parent to have its refcount decremented
5317 * by the random zapper, it won't actually be freed, meaning it's still
5318 * there for when its parent gets freed.
5319 * When the parent SV is freed, in the case of magic, the magic is freed,
5320 * Perl_magic_killbackrefs is called which decrements one refcount, then
5321 * mg_obj is freed which kills the second count.
5322 * In the vase of a HV being freed, one ref is removed by S_hfreeentries,
5323 * the other by Perl_sv_kill_backrefs, which it calls.
5327 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5332 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5334 if (SvTYPE(tsv) == SVt_PVHV) {
5335 AV **const avp = Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5341 SvREFCNT_inc_simple_void(av); /* see discussion above */
5345 const MAGIC *const mg
5346 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5348 av = MUTABLE_AV(mg->mg_obj);
5352 sv_magic(tsv, MUTABLE_SV(av), PERL_MAGIC_backref, NULL, 0);
5353 /* av now has a refcnt of 2; see discussion above */
5356 if (AvFILLp(av) >= AvMAX(av)) {
5357 av_extend(av, AvFILLp(av)+1);
5359 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5362 /* delete a back-reference to ourselves from the backref magic associated
5363 * with the SV we point to.
5367 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5374 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5376 if (SvTYPE(tsv) == SVt_PVHV) {
5378 /* SvOOK: We must avoid creating the hv_aux structure if its
5379 * not already there, as that is stored in the main HvARRAY(),
5380 * and hfreentries assumes that no-one reallocates HvARRAY()
5381 * while it is running. */
5382 av = *Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5386 const MAGIC *const mg
5387 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5389 av = MUTABLE_AV(mg->mg_obj);
5393 Perl_croak(aTHX_ "panic: del_backref");
5395 assert(!SvIS_FREED(av));
5398 /* We shouldn't be in here more than once, but for paranoia reasons lets
5400 for (i = AvFILLp(av); i >= 0; i--) {
5402 const SSize_t fill = AvFILLp(av);
5404 /* We weren't the last entry.
5405 An unordered list has this property that you can take the
5406 last element off the end to fill the hole, and it's still
5407 an unordered list :-)
5412 AvFILLp(av) = fill - 1;
5418 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5420 SV **svp = AvARRAY(av);
5422 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5425 SV *const *const last = svp + AvFILLp(av);
5427 assert(!SvIS_FREED(av));
5428 while (svp <= last) {
5430 SV *const referrer = *svp;
5431 if (SvWEAKREF(referrer)) {
5432 /* XXX Should we check that it hasn't changed? */
5433 assert(SvROK(referrer));
5434 SvRV_set(referrer, 0);
5436 SvWEAKREF_off(referrer);
5437 SvSETMAGIC(referrer);
5438 } else if (SvTYPE(referrer) == SVt_PVGV ||
5439 SvTYPE(referrer) == SVt_PVLV) {
5440 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5441 /* You lookin' at me? */
5442 assert(GvSTASH(referrer));
5443 assert(GvSTASH(referrer) == (const HV *)sv);
5444 GvSTASH(referrer) = 0;
5445 } else if (SvTYPE(referrer) == SVt_PVCV ||
5446 SvTYPE(referrer) == SVt_PVFM) {
5447 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5448 /* You lookin' at me? */
5449 assert(CvSTASH(referrer));
5450 assert(CvSTASH(referrer) == (const HV *)sv);
5451 CvSTASH(referrer) = 0;
5454 assert(SvTYPE(sv) == SVt_PVGV);
5455 /* You lookin' at me? */
5456 assert(CvGV(referrer));
5457 assert(CvGV(referrer) == (const GV *)sv);
5458 anonymise_cv_maybe(MUTABLE_GV(sv),
5459 MUTABLE_CV(referrer));
5464 "panic: magic_killbackrefs (flags=%"UVxf")",
5465 (UV)SvFLAGS(referrer));
5474 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5479 =for apidoc sv_insert
5481 Inserts a string at the specified offset/length within the SV. Similar to
5482 the Perl substr() function. Handles get magic.
5484 =for apidoc sv_insert_flags
5486 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5492 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5497 register char *midend;
5498 register char *bigend;
5502 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5505 Perl_croak(aTHX_ "Can't modify non-existent substring");
5506 SvPV_force_flags(bigstr, curlen, flags);
5507 (void)SvPOK_only_UTF8(bigstr);
5508 if (offset + len > curlen) {
5509 SvGROW(bigstr, offset+len+1);
5510 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5511 SvCUR_set(bigstr, offset+len);
5515 i = littlelen - len;
5516 if (i > 0) { /* string might grow */
5517 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5518 mid = big + offset + len;
5519 midend = bigend = big + SvCUR(bigstr);
5522 while (midend > mid) /* shove everything down */
5523 *--bigend = *--midend;
5524 Move(little,big+offset,littlelen,char);
5525 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5530 Move(little,SvPVX(bigstr)+offset,len,char);
5535 big = SvPVX(bigstr);
5538 bigend = big + SvCUR(bigstr);
5540 if (midend > bigend)
5541 Perl_croak(aTHX_ "panic: sv_insert");
5543 if (mid - big > bigend - midend) { /* faster to shorten from end */
5545 Move(little, mid, littlelen,char);
5548 i = bigend - midend;
5550 Move(midend, mid, i,char);
5554 SvCUR_set(bigstr, mid - big);
5556 else if ((i = mid - big)) { /* faster from front */
5557 midend -= littlelen;
5559 Move(big, midend - i, i, char);
5560 sv_chop(bigstr,midend-i);
5562 Move(little, mid, littlelen,char);
5564 else if (littlelen) {
5565 midend -= littlelen;
5566 sv_chop(bigstr,midend);
5567 Move(little,midend,littlelen,char);
5570 sv_chop(bigstr,midend);
5576 =for apidoc sv_replace
5578 Make the first argument a copy of the second, then delete the original.
5579 The target SV physically takes over ownership of the body of the source SV
5580 and inherits its flags; however, the target keeps any magic it owns,
5581 and any magic in the source is discarded.
5582 Note that this is a rather specialist SV copying operation; most of the
5583 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5589 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5592 const U32 refcnt = SvREFCNT(sv);
5594 PERL_ARGS_ASSERT_SV_REPLACE;
5596 SV_CHECK_THINKFIRST_COW_DROP(sv);
5597 if (SvREFCNT(nsv) != 1) {
5598 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5599 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5601 if (SvMAGICAL(sv)) {
5605 sv_upgrade(nsv, SVt_PVMG);
5606 SvMAGIC_set(nsv, SvMAGIC(sv));
5607 SvFLAGS(nsv) |= SvMAGICAL(sv);
5609 SvMAGIC_set(sv, NULL);
5613 assert(!SvREFCNT(sv));
5614 #ifdef DEBUG_LEAKING_SCALARS
5615 sv->sv_flags = nsv->sv_flags;
5616 sv->sv_any = nsv->sv_any;
5617 sv->sv_refcnt = nsv->sv_refcnt;
5618 sv->sv_u = nsv->sv_u;
5620 StructCopy(nsv,sv,SV);
5622 if(SvTYPE(sv) == SVt_IV) {
5624 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5628 #ifdef PERL_OLD_COPY_ON_WRITE
5629 if (SvIsCOW_normal(nsv)) {
5630 /* We need to follow the pointers around the loop to make the
5631 previous SV point to sv, rather than nsv. */
5634 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5637 assert(SvPVX_const(current) == SvPVX_const(nsv));
5639 /* Make the SV before us point to the SV after us. */
5641 PerlIO_printf(Perl_debug_log, "previous is\n");
5643 PerlIO_printf(Perl_debug_log,
5644 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5645 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5647 SV_COW_NEXT_SV_SET(current, sv);
5650 SvREFCNT(sv) = refcnt;
5651 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5656 /* We're about to free a GV which has a CV that refers back to us.
5657 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5661 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5667 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5670 assert(SvREFCNT(gv) == 0);
5671 assert(isGV(gv) && isGV_with_GP(gv));
5673 assert(!CvANON(cv));
5674 assert(CvGV(cv) == gv);
5676 /* will the CV shortly be freed by gp_free() ? */
5677 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5678 SvANY(cv)->xcv_gv = NULL;
5682 /* if not, anonymise: */
5683 stash = GvSTASH(gv) ? HvNAME(GvSTASH(gv)) : NULL;
5684 gvname = Perl_newSVpvf(aTHX_ "%s::__ANON__",
5685 stash ? stash : "__ANON__");
5686 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5687 SvREFCNT_dec(gvname);
5691 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5696 =for apidoc sv_clear
5698 Clear an SV: call any destructors, free up any memory used by the body,
5699 and free the body itself. The SV's head is I<not> freed, although
5700 its type is set to all 1's so that it won't inadvertently be assumed
5701 to be live during global destruction etc.
5702 This function should only be called when REFCNT is zero. Most of the time
5703 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5710 Perl_sv_clear(pTHX_ register SV *const sv)
5713 const U32 type = SvTYPE(sv);
5714 const struct body_details *const sv_type_details
5715 = bodies_by_type + type;
5718 PERL_ARGS_ASSERT_SV_CLEAR;
5719 assert(SvREFCNT(sv) == 0);
5720 assert(SvTYPE(sv) != SVTYPEMASK);
5722 if (type <= SVt_IV) {
5723 /* See the comment in sv.h about the collusion between this early
5724 return and the overloading of the NULL slots in the size table. */
5727 SvFLAGS(sv) &= SVf_BREAK;
5728 SvFLAGS(sv) |= SVTYPEMASK;
5733 if (PL_defstash && /* Still have a symbol table? */
5740 stash = SvSTASH(sv);
5741 destructor = StashHANDLER(stash,DESTROY);
5743 /* A constant subroutine can have no side effects, so
5744 don't bother calling it. */
5745 && !CvCONST(destructor)
5746 /* Don't bother calling an empty destructor */
5747 && (CvISXSUB(destructor)
5748 || (CvSTART(destructor)
5749 && (CvSTART(destructor)->op_next->op_type != OP_LEAVESUB))))
5751 SV* const tmpref = newRV(sv);
5752 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
5754 PUSHSTACKi(PERLSI_DESTROY);
5759 call_sv(MUTABLE_SV(destructor), G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
5765 if(SvREFCNT(tmpref) < 2) {
5766 /* tmpref is not kept alive! */
5768 SvRV_set(tmpref, NULL);
5771 SvREFCNT_dec(tmpref);
5773 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
5777 if (PL_in_clean_objs)
5778 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
5780 /* DESTROY gave object new lease on life */
5786 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
5787 SvOBJECT_off(sv); /* Curse the object. */
5788 if (type != SVt_PVIO)
5789 --PL_sv_objcount; /* XXX Might want something more general */
5792 if (type >= SVt_PVMG) {
5793 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5794 SvREFCNT_dec(SvOURSTASH(sv));
5795 } else if (SvMAGIC(sv))
5797 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5798 SvREFCNT_dec(SvSTASH(sv));
5801 /* case SVt_BIND: */
5804 IoIFP(sv) != PerlIO_stdin() &&
5805 IoIFP(sv) != PerlIO_stdout() &&
5806 IoIFP(sv) != PerlIO_stderr() &&
5807 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5809 io_close(MUTABLE_IO(sv), FALSE);
5811 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5812 PerlDir_close(IoDIRP(sv));
5813 IoDIRP(sv) = (DIR*)NULL;
5814 Safefree(IoTOP_NAME(sv));
5815 Safefree(IoFMT_NAME(sv));
5816 Safefree(IoBOTTOM_NAME(sv));
5819 /* FIXME for plugins */
5820 pregfree2((REGEXP*) sv);
5824 cv_undef(MUTABLE_CV(sv));
5825 /* If we're in a stash, we don't own a reference to it. However it does
5826 have a back reference to us, which needs to be cleared. */
5827 if ((stash = CvSTASH(sv)))
5828 sv_del_backref(MUTABLE_SV(stash), sv);
5831 if (PL_last_swash_hv == (const HV *)sv) {
5832 PL_last_swash_hv = NULL;
5834 hv_undef(MUTABLE_HV(sv));
5837 if (PL_comppad == MUTABLE_AV(sv)) {
5841 av_undef(MUTABLE_AV(sv));
5844 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5845 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5846 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5847 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5849 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5850 SvREFCNT_dec(LvTARG(sv));
5852 if (isGV_with_GP(sv)) {
5853 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
5854 && HvNAME_get(stash))
5855 mro_method_changed_in(stash);
5856 gp_free(MUTABLE_GV(sv));
5858 unshare_hek(GvNAME_HEK(sv));
5859 /* If we're in a stash, we don't own a reference to it. However it does
5860 have a back reference to us, which needs to be cleared. */
5861 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
5862 sv_del_backref(MUTABLE_SV(stash), sv);
5864 /* FIXME. There are probably more unreferenced pointers to SVs in the
5865 interpreter struct that we should check and tidy in a similar
5867 if ((const GV *)sv == PL_last_in_gv)
5868 PL_last_in_gv = NULL;
5874 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5877 SvOOK_offset(sv, offset);
5878 SvPV_set(sv, SvPVX_mutable(sv) - offset);
5879 /* Don't even bother with turning off the OOK flag. */
5884 SV * const target = SvRV(sv);
5886 sv_del_backref(target, sv);
5888 SvREFCNT_dec(target);
5891 #ifdef PERL_OLD_COPY_ON_WRITE
5892 else if (SvPVX_const(sv)
5893 && !(SvTYPE(sv) == SVt_PVIO && !(IoFLAGS(sv) & IOf_FAKE_DIRP))) {
5896 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
5900 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
5902 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5906 } else if (SvLEN(sv)) {
5907 Safefree(SvPVX_const(sv));
5911 else if (SvPVX_const(sv) && SvLEN(sv)
5912 && !(SvTYPE(sv) == SVt_PVIO && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
5913 Safefree(SvPVX_mutable(sv));
5914 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
5915 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5924 SvFLAGS(sv) &= SVf_BREAK;
5925 SvFLAGS(sv) |= SVTYPEMASK;
5927 if (sv_type_details->arena) {
5928 del_body(((char *)SvANY(sv) + sv_type_details->offset),
5929 &PL_body_roots[type]);
5931 else if (sv_type_details->body_size) {
5932 my_safefree(SvANY(sv));
5937 =for apidoc sv_newref
5939 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5946 Perl_sv_newref(pTHX_ SV *const sv)
5948 PERL_UNUSED_CONTEXT;
5957 Decrement an SV's reference count, and if it drops to zero, call
5958 C<sv_clear> to invoke destructors and free up any memory used by
5959 the body; finally, deallocate the SV's head itself.
5960 Normally called via a wrapper macro C<SvREFCNT_dec>.
5966 Perl_sv_free(pTHX_ SV *const sv)
5971 if (SvREFCNT(sv) == 0) {
5972 if (SvFLAGS(sv) & SVf_BREAK)
5973 /* this SV's refcnt has been artificially decremented to
5974 * trigger cleanup */
5976 if (PL_in_clean_all) /* All is fair */
5978 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5979 /* make sure SvREFCNT(sv)==0 happens very seldom */
5980 SvREFCNT(sv) = (~(U32)0)/2;
5983 if (ckWARN_d(WARN_INTERNAL)) {
5984 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5985 Perl_dump_sv_child(aTHX_ sv);
5987 #ifdef DEBUG_LEAKING_SCALARS
5990 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5991 if (PL_warnhook == PERL_WARNHOOK_FATAL
5992 || ckDEAD(packWARN(WARN_INTERNAL))) {
5993 /* Don't let Perl_warner cause us to escape our fate: */
5997 /* This may not return: */
5998 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
5999 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6000 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6003 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6008 if (--(SvREFCNT(sv)) > 0)
6010 Perl_sv_free2(aTHX_ sv);
6014 Perl_sv_free2(pTHX_ SV *const sv)
6018 PERL_ARGS_ASSERT_SV_FREE2;
6022 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6023 "Attempt to free temp prematurely: SV 0x%"UVxf
6024 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6028 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6029 /* make sure SvREFCNT(sv)==0 happens very seldom */
6030 SvREFCNT(sv) = (~(U32)0)/2;
6041 Returns the length of the string in the SV. Handles magic and type
6042 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6048 Perl_sv_len(pTHX_ register SV *const sv)
6056 len = mg_length(sv);
6058 (void)SvPV_const(sv, len);
6063 =for apidoc sv_len_utf8
6065 Returns the number of characters in the string in an SV, counting wide
6066 UTF-8 bytes as a single character. Handles magic and type coercion.
6072 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6073 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6074 * (Note that the mg_len is not the length of the mg_ptr field.
6075 * This allows the cache to store the character length of the string without
6076 * needing to malloc() extra storage to attach to the mg_ptr.)
6081 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6087 return mg_length(sv);
6091 const U8 *s = (U8*)SvPV_const(sv, len);
6095 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6097 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6098 if (mg->mg_len != -1)
6101 /* We can use the offset cache for a headstart.
6102 The longer value is stored in the first pair. */
6103 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6105 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6109 if (PL_utf8cache < 0) {
6110 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6111 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6115 ulen = Perl_utf8_length(aTHX_ s, s + len);
6116 utf8_mg_len_cache_update(sv, &mg, ulen);
6120 return Perl_utf8_length(aTHX_ s, s + len);
6124 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6127 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6128 STRLEN *const uoffset_p, bool *const at_end)
6130 const U8 *s = start;
6131 STRLEN uoffset = *uoffset_p;
6133 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6135 while (s < send && uoffset) {
6142 else if (s > send) {
6144 /* This is the existing behaviour. Possibly it should be a croak, as
6145 it's actually a bounds error */
6148 *uoffset_p -= uoffset;
6152 /* Given the length of the string in both bytes and UTF-8 characters, decide
6153 whether to walk forwards or backwards to find the byte corresponding to
6154 the passed in UTF-8 offset. */
6156 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6157 STRLEN uoffset, const STRLEN uend)
6159 STRLEN backw = uend - uoffset;
6161 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6163 if (uoffset < 2 * backw) {
6164 /* The assumption is that going forwards is twice the speed of going
6165 forward (that's where the 2 * backw comes from).
6166 (The real figure of course depends on the UTF-8 data.) */
6167 const U8 *s = start;
6169 while (s < send && uoffset--)
6179 while (UTF8_IS_CONTINUATION(*send))
6182 return send - start;
6185 /* For the string representation of the given scalar, find the byte
6186 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6187 give another position in the string, *before* the sought offset, which
6188 (which is always true, as 0, 0 is a valid pair of positions), which should
6189 help reduce the amount of linear searching.
6190 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6191 will be used to reduce the amount of linear searching. The cache will be
6192 created if necessary, and the found value offered to it for update. */
6194 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6195 const U8 *const send, STRLEN uoffset,
6196 STRLEN uoffset0, STRLEN boffset0)
6198 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6200 bool at_end = FALSE;
6202 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6204 assert (uoffset >= uoffset0);
6211 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6212 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6213 if ((*mgp)->mg_ptr) {
6214 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6215 if (cache[0] == uoffset) {
6216 /* An exact match. */
6219 if (cache[2] == uoffset) {
6220 /* An exact match. */
6224 if (cache[0] < uoffset) {
6225 /* The cache already knows part of the way. */
6226 if (cache[0] > uoffset0) {
6227 /* The cache knows more than the passed in pair */
6228 uoffset0 = cache[0];
6229 boffset0 = cache[1];
6231 if ((*mgp)->mg_len != -1) {
6232 /* And we know the end too. */
6234 + sv_pos_u2b_midway(start + boffset0, send,
6236 (*mgp)->mg_len - uoffset0);
6238 uoffset -= uoffset0;
6240 + sv_pos_u2b_forwards(start + boffset0,
6241 send, &uoffset, &at_end);
6242 uoffset += uoffset0;
6245 else if (cache[2] < uoffset) {
6246 /* We're between the two cache entries. */
6247 if (cache[2] > uoffset0) {
6248 /* and the cache knows more than the passed in pair */
6249 uoffset0 = cache[2];
6250 boffset0 = cache[3];
6254 + sv_pos_u2b_midway(start + boffset0,
6257 cache[0] - uoffset0);
6260 + sv_pos_u2b_midway(start + boffset0,
6263 cache[2] - uoffset0);
6267 else if ((*mgp)->mg_len != -1) {
6268 /* If we can take advantage of a passed in offset, do so. */
6269 /* In fact, offset0 is either 0, or less than offset, so don't
6270 need to worry about the other possibility. */
6272 + sv_pos_u2b_midway(start + boffset0, send,
6274 (*mgp)->mg_len - uoffset0);
6279 if (!found || PL_utf8cache < 0) {
6280 STRLEN real_boffset;
6281 uoffset -= uoffset0;
6282 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6283 send, &uoffset, &at_end);
6284 uoffset += uoffset0;
6286 if (found && PL_utf8cache < 0)
6287 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6289 boffset = real_boffset;
6294 utf8_mg_len_cache_update(sv, mgp, uoffset);
6296 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6303 =for apidoc sv_pos_u2b_flags
6305 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6306 the start of the string, to a count of the equivalent number of bytes; if
6307 lenp is non-zero, it does the same to lenp, but this time starting from
6308 the offset, rather than from the start of the string. Handles type coercion.
6309 I<flags> is passed to C<SvPV_flags>, and usually should be
6310 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6316 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6317 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6318 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6323 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6330 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6332 start = (U8*)SvPV_flags(sv, len, flags);
6334 const U8 * const send = start + len;
6336 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6339 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6340 is 0, and *lenp is already set to that. */) {
6341 /* Convert the relative offset to absolute. */
6342 const STRLEN uoffset2 = uoffset + *lenp;
6343 const STRLEN boffset2
6344 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6345 uoffset, boffset) - boffset;
6359 =for apidoc sv_pos_u2b
6361 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6362 the start of the string, to a count of the equivalent number of bytes; if
6363 lenp is non-zero, it does the same to lenp, but this time starting from
6364 the offset, rather than from the start of the string. Handles magic and
6367 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6374 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6375 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6376 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6380 /* This function is subject to size and sign problems */
6383 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6385 PERL_ARGS_ASSERT_SV_POS_U2B;
6388 STRLEN ulen = (STRLEN)*lenp;
6389 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6390 SV_GMAGIC|SV_CONST_RETURN);
6393 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6394 SV_GMAGIC|SV_CONST_RETURN);
6399 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6402 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6406 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6407 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6408 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6412 (*mgp)->mg_len = ulen;
6413 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6414 if (ulen != (STRLEN) (*mgp)->mg_len)
6415 (*mgp)->mg_len = -1;
6418 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6419 byte length pairing. The (byte) length of the total SV is passed in too,
6420 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6421 may not have updated SvCUR, so we can't rely on reading it directly.
6423 The proffered utf8/byte length pairing isn't used if the cache already has
6424 two pairs, and swapping either for the proffered pair would increase the
6425 RMS of the intervals between known byte offsets.
6427 The cache itself consists of 4 STRLEN values
6428 0: larger UTF-8 offset
6429 1: corresponding byte offset
6430 2: smaller UTF-8 offset
6431 3: corresponding byte offset
6433 Unused cache pairs have the value 0, 0.
6434 Keeping the cache "backwards" means that the invariant of
6435 cache[0] >= cache[2] is maintained even with empty slots, which means that
6436 the code that uses it doesn't need to worry if only 1 entry has actually
6437 been set to non-zero. It also makes the "position beyond the end of the
6438 cache" logic much simpler, as the first slot is always the one to start
6442 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6443 const STRLEN utf8, const STRLEN blen)
6447 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6452 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6453 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6454 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6456 (*mgp)->mg_len = -1;
6460 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6461 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6462 (*mgp)->mg_ptr = (char *) cache;
6466 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6467 /* SvPOKp() because it's possible that sv has string overloading, and
6468 therefore is a reference, hence SvPVX() is actually a pointer.
6469 This cures the (very real) symptoms of RT 69422, but I'm not actually
6470 sure whether we should even be caching the results of UTF-8
6471 operations on overloading, given that nothing stops overloading
6472 returning a different value every time it's called. */
6473 const U8 *start = (const U8 *) SvPVX_const(sv);
6474 const STRLEN realutf8 = utf8_length(start, start + byte);
6476 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6480 /* Cache is held with the later position first, to simplify the code
6481 that deals with unbounded ends. */
6483 ASSERT_UTF8_CACHE(cache);
6484 if (cache[1] == 0) {
6485 /* Cache is totally empty */
6488 } else if (cache[3] == 0) {
6489 if (byte > cache[1]) {
6490 /* New one is larger, so goes first. */
6491 cache[2] = cache[0];
6492 cache[3] = cache[1];
6500 #define THREEWAY_SQUARE(a,b,c,d) \
6501 ((float)((d) - (c))) * ((float)((d) - (c))) \
6502 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6503 + ((float)((b) - (a))) * ((float)((b) - (a)))
6505 /* Cache has 2 slots in use, and we know three potential pairs.
6506 Keep the two that give the lowest RMS distance. Do the
6507 calcualation in bytes simply because we always know the byte
6508 length. squareroot has the same ordering as the positive value,
6509 so don't bother with the actual square root. */
6510 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6511 if (byte > cache[1]) {
6512 /* New position is after the existing pair of pairs. */
6513 const float keep_earlier
6514 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6515 const float keep_later
6516 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6518 if (keep_later < keep_earlier) {
6519 if (keep_later < existing) {
6520 cache[2] = cache[0];
6521 cache[3] = cache[1];
6527 if (keep_earlier < existing) {
6533 else if (byte > cache[3]) {
6534 /* New position is between the existing pair of pairs. */
6535 const float keep_earlier
6536 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6537 const float keep_later
6538 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6540 if (keep_later < keep_earlier) {
6541 if (keep_later < existing) {
6547 if (keep_earlier < existing) {
6554 /* New position is before the existing pair of pairs. */
6555 const float keep_earlier
6556 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6557 const float keep_later
6558 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6560 if (keep_later < keep_earlier) {
6561 if (keep_later < existing) {
6567 if (keep_earlier < existing) {
6568 cache[0] = cache[2];
6569 cache[1] = cache[3];
6576 ASSERT_UTF8_CACHE(cache);
6579 /* We already know all of the way, now we may be able to walk back. The same
6580 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6581 backward is half the speed of walking forward. */
6583 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6584 const U8 *end, STRLEN endu)
6586 const STRLEN forw = target - s;
6587 STRLEN backw = end - target;
6589 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6591 if (forw < 2 * backw) {
6592 return utf8_length(s, target);
6595 while (end > target) {
6597 while (UTF8_IS_CONTINUATION(*end)) {
6606 =for apidoc sv_pos_b2u
6608 Converts the value pointed to by offsetp from a count of bytes from the
6609 start of the string, to a count of the equivalent number of UTF-8 chars.
6610 Handles magic and type coercion.
6616 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6617 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6622 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6625 const STRLEN byte = *offsetp;
6626 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6632 PERL_ARGS_ASSERT_SV_POS_B2U;
6637 s = (const U8*)SvPV_const(sv, blen);
6640 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
6646 && SvTYPE(sv) >= SVt_PVMG
6647 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
6650 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
6651 if (cache[1] == byte) {
6652 /* An exact match. */
6653 *offsetp = cache[0];
6656 if (cache[3] == byte) {
6657 /* An exact match. */
6658 *offsetp = cache[2];
6662 if (cache[1] < byte) {
6663 /* We already know part of the way. */
6664 if (mg->mg_len != -1) {
6665 /* Actually, we know the end too. */
6667 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
6668 s + blen, mg->mg_len - cache[0]);
6670 len = cache[0] + utf8_length(s + cache[1], send);
6673 else if (cache[3] < byte) {
6674 /* We're between the two cached pairs, so we do the calculation
6675 offset by the byte/utf-8 positions for the earlier pair,
6676 then add the utf-8 characters from the string start to
6678 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
6679 s + cache[1], cache[0] - cache[2])
6683 else { /* cache[3] > byte */
6684 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
6688 ASSERT_UTF8_CACHE(cache);
6690 } else if (mg->mg_len != -1) {
6691 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
6695 if (!found || PL_utf8cache < 0) {
6696 const STRLEN real_len = utf8_length(s, send);
6698 if (found && PL_utf8cache < 0)
6699 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
6706 utf8_mg_len_cache_update(sv, &mg, len);
6708 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
6713 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
6714 STRLEN real, SV *const sv)
6716 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
6718 /* As this is debugging only code, save space by keeping this test here,
6719 rather than inlining it in all the callers. */
6720 if (from_cache == real)
6723 /* Need to turn the assertions off otherwise we may recurse infinitely
6724 while printing error messages. */
6725 SAVEI8(PL_utf8cache);
6727 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
6728 func, (UV) from_cache, (UV) real, SVfARG(sv));
6734 Returns a boolean indicating whether the strings in the two SVs are
6735 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6736 coerce its args to strings if necessary.
6742 Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
6751 SV* svrecode = NULL;
6758 /* if pv1 and pv2 are the same, second SvPV_const call may
6759 * invalidate pv1, so we may need to make a copy */
6760 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
6761 pv1 = SvPV_const(sv1, cur1);
6762 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
6764 pv1 = SvPV_const(sv1, cur1);
6772 pv2 = SvPV_const(sv2, cur2);
6774 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6775 /* Differing utf8ness.
6776 * Do not UTF8size the comparands as a side-effect. */
6779 svrecode = newSVpvn(pv2, cur2);
6780 sv_recode_to_utf8(svrecode, PL_encoding);
6781 pv2 = SvPV_const(svrecode, cur2);
6784 svrecode = newSVpvn(pv1, cur1);
6785 sv_recode_to_utf8(svrecode, PL_encoding);
6786 pv1 = SvPV_const(svrecode, cur1);
6788 /* Now both are in UTF-8. */
6790 SvREFCNT_dec(svrecode);
6795 bool is_utf8 = TRUE;
6798 /* sv1 is the UTF-8 one,
6799 * if is equal it must be downgrade-able */
6800 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
6806 /* sv2 is the UTF-8 one,
6807 * if is equal it must be downgrade-able */
6808 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
6814 /* Downgrade not possible - cannot be eq */
6822 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
6824 SvREFCNT_dec(svrecode);
6834 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6835 string in C<sv1> is less than, equal to, or greater than the string in
6836 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6837 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
6843 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
6847 const char *pv1, *pv2;
6850 SV *svrecode = NULL;
6857 pv1 = SvPV_const(sv1, cur1);
6864 pv2 = SvPV_const(sv2, cur2);
6866 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6867 /* Differing utf8ness.
6868 * Do not UTF8size the comparands as a side-effect. */
6871 svrecode = newSVpvn(pv2, cur2);
6872 sv_recode_to_utf8(svrecode, PL_encoding);
6873 pv2 = SvPV_const(svrecode, cur2);
6876 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
6881 svrecode = newSVpvn(pv1, cur1);
6882 sv_recode_to_utf8(svrecode, PL_encoding);
6883 pv1 = SvPV_const(svrecode, cur1);
6886 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
6892 cmp = cur2 ? -1 : 0;
6896 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
6899 cmp = retval < 0 ? -1 : 1;
6900 } else if (cur1 == cur2) {
6903 cmp = cur1 < cur2 ? -1 : 1;
6907 SvREFCNT_dec(svrecode);
6915 =for apidoc sv_cmp_locale
6917 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6918 'use bytes' aware, handles get magic, and will coerce its args to strings
6919 if necessary. See also C<sv_cmp>.
6925 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
6928 #ifdef USE_LOCALE_COLLATE
6934 if (PL_collation_standard)
6938 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
6940 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
6942 if (!pv1 || !len1) {
6953 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
6956 return retval < 0 ? -1 : 1;
6959 * When the result of collation is equality, that doesn't mean
6960 * that there are no differences -- some locales exclude some
6961 * characters from consideration. So to avoid false equalities,
6962 * we use the raw string as a tiebreaker.
6968 #endif /* USE_LOCALE_COLLATE */
6970 return sv_cmp(sv1, sv2);
6974 #ifdef USE_LOCALE_COLLATE
6977 =for apidoc sv_collxfrm
6979 Add Collate Transform magic to an SV if it doesn't already have it.
6981 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6982 scalar data of the variable, but transformed to such a format that a normal
6983 memory comparison can be used to compare the data according to the locale
6990 Perl_sv_collxfrm(pTHX_ SV *const sv, STRLEN *const nxp)
6995 PERL_ARGS_ASSERT_SV_COLLXFRM;
6997 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
6998 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7004 Safefree(mg->mg_ptr);
7005 s = SvPV_const(sv, len);
7006 if ((xf = mem_collxfrm(s, len, &xlen))) {
7008 #ifdef PERL_OLD_COPY_ON_WRITE
7010 sv_force_normal_flags(sv, 0);
7012 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7026 if (mg && mg->mg_ptr) {
7028 return mg->mg_ptr + sizeof(PL_collation_ix);
7036 #endif /* USE_LOCALE_COLLATE */
7041 Get a line from the filehandle and store it into the SV, optionally
7042 appending to the currently-stored string.
7048 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7053 register STDCHAR rslast;
7054 register STDCHAR *bp;
7059 PERL_ARGS_ASSERT_SV_GETS;
7061 if (SvTHINKFIRST(sv))
7062 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7063 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7065 However, perlbench says it's slower, because the existing swipe code
7066 is faster than copy on write.
7067 Swings and roundabouts. */
7068 SvUPGRADE(sv, SVt_PV);
7073 if (PerlIO_isutf8(fp)) {
7075 sv_utf8_upgrade_nomg(sv);
7076 sv_pos_u2b(sv,&append,0);
7078 } else if (SvUTF8(sv)) {
7079 SV * const tsv = newSV(0);
7080 sv_gets(tsv, fp, 0);
7081 sv_utf8_upgrade_nomg(tsv);
7082 SvCUR_set(sv,append);
7085 goto return_string_or_null;
7093 if (PerlIO_isutf8(fp))
7096 if (IN_PERL_COMPILETIME) {
7097 /* we always read code in line mode */
7101 else if (RsSNARF(PL_rs)) {
7102 /* If it is a regular disk file use size from stat() as estimate
7103 of amount we are going to read -- may result in mallocing
7104 more memory than we really need if the layers below reduce
7105 the size we read (e.g. CRLF or a gzip layer).
7108 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7109 const Off_t offset = PerlIO_tell(fp);
7110 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7111 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7117 else if (RsRECORD(PL_rs)) {
7125 /* Grab the size of the record we're getting */
7126 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7127 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7130 /* VMS wants read instead of fread, because fread doesn't respect */
7131 /* RMS record boundaries. This is not necessarily a good thing to be */
7132 /* doing, but we've got no other real choice - except avoid stdio
7133 as implementation - perhaps write a :vms layer ?
7135 fd = PerlIO_fileno(fp);
7136 if (fd == -1) { /* in-memory file from PerlIO::Scalar */
7137 bytesread = PerlIO_read(fp, buffer, recsize);
7140 bytesread = PerlLIO_read(fd, buffer, recsize);
7143 bytesread = PerlIO_read(fp, buffer, recsize);
7147 SvCUR_set(sv, bytesread + append);
7148 buffer[bytesread] = '\0';
7149 goto return_string_or_null;
7151 else if (RsPARA(PL_rs)) {
7157 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7158 if (PerlIO_isutf8(fp)) {
7159 rsptr = SvPVutf8(PL_rs, rslen);
7162 if (SvUTF8(PL_rs)) {
7163 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7164 Perl_croak(aTHX_ "Wide character in $/");
7167 rsptr = SvPV_const(PL_rs, rslen);
7171 rslast = rslen ? rsptr[rslen - 1] : '\0';
7173 if (rspara) { /* have to do this both before and after */
7174 do { /* to make sure file boundaries work right */
7177 i = PerlIO_getc(fp);
7181 PerlIO_ungetc(fp,i);
7187 /* See if we know enough about I/O mechanism to cheat it ! */
7189 /* This used to be #ifdef test - it is made run-time test for ease
7190 of abstracting out stdio interface. One call should be cheap
7191 enough here - and may even be a macro allowing compile
7195 if (PerlIO_fast_gets(fp)) {
7198 * We're going to steal some values from the stdio struct
7199 * and put EVERYTHING in the innermost loop into registers.
7201 register STDCHAR *ptr;
7205 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7206 /* An ungetc()d char is handled separately from the regular
7207 * buffer, so we getc() it back out and stuff it in the buffer.
7209 i = PerlIO_getc(fp);
7210 if (i == EOF) return 0;
7211 *(--((*fp)->_ptr)) = (unsigned char) i;
7215 /* Here is some breathtakingly efficient cheating */
7217 cnt = PerlIO_get_cnt(fp); /* get count into register */
7218 /* make sure we have the room */
7219 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7220 /* Not room for all of it
7221 if we are looking for a separator and room for some
7223 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7224 /* just process what we have room for */
7225 shortbuffered = cnt - SvLEN(sv) + append + 1;
7226 cnt -= shortbuffered;
7230 /* remember that cnt can be negative */
7231 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7236 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7237 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7238 DEBUG_P(PerlIO_printf(Perl_debug_log,
7239 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7240 DEBUG_P(PerlIO_printf(Perl_debug_log,
7241 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7242 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7243 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7248 while (cnt > 0) { /* this | eat */
7250 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7251 goto thats_all_folks; /* screams | sed :-) */
7255 Copy(ptr, bp, cnt, char); /* this | eat */
7256 bp += cnt; /* screams | dust */
7257 ptr += cnt; /* louder | sed :-) */
7262 if (shortbuffered) { /* oh well, must extend */
7263 cnt = shortbuffered;
7265 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7267 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7268 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7272 DEBUG_P(PerlIO_printf(Perl_debug_log,
7273 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7274 PTR2UV(ptr),(long)cnt));
7275 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7277 DEBUG_P(PerlIO_printf(Perl_debug_log,
7278 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7279 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7280 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7282 /* This used to call 'filbuf' in stdio form, but as that behaves like
7283 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7284 another abstraction. */
7285 i = PerlIO_getc(fp); /* get more characters */
7287 DEBUG_P(PerlIO_printf(Perl_debug_log,
7288 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7289 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7290 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7292 cnt = PerlIO_get_cnt(fp);
7293 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7294 DEBUG_P(PerlIO_printf(Perl_debug_log,
7295 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7297 if (i == EOF) /* all done for ever? */
7298 goto thats_really_all_folks;
7300 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7302 SvGROW(sv, bpx + cnt + 2);
7303 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7305 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7307 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7308 goto thats_all_folks;
7312 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7313 memNE((char*)bp - rslen, rsptr, rslen))
7314 goto screamer; /* go back to the fray */
7315 thats_really_all_folks:
7317 cnt += shortbuffered;
7318 DEBUG_P(PerlIO_printf(Perl_debug_log,
7319 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7320 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7321 DEBUG_P(PerlIO_printf(Perl_debug_log,
7322 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7323 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7324 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7326 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7327 DEBUG_P(PerlIO_printf(Perl_debug_log,
7328 "Screamer: done, len=%ld, string=|%.*s|\n",
7329 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7333 /*The big, slow, and stupid way. */
7334 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7335 STDCHAR *buf = NULL;
7336 Newx(buf, 8192, STDCHAR);
7344 register const STDCHAR * const bpe = buf + sizeof(buf);
7346 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7347 ; /* keep reading */
7351 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7352 /* Accomodate broken VAXC compiler, which applies U8 cast to
7353 * both args of ?: operator, causing EOF to change into 255
7356 i = (U8)buf[cnt - 1];
7362 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7364 sv_catpvn(sv, (char *) buf, cnt);
7366 sv_setpvn(sv, (char *) buf, cnt);
7368 if (i != EOF && /* joy */
7370 SvCUR(sv) < rslen ||
7371 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7375 * If we're reading from a TTY and we get a short read,
7376 * indicating that the user hit his EOF character, we need
7377 * to notice it now, because if we try to read from the TTY
7378 * again, the EOF condition will disappear.
7380 * The comparison of cnt to sizeof(buf) is an optimization
7381 * that prevents unnecessary calls to feof().
7385 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7389 #ifdef USE_HEAP_INSTEAD_OF_STACK
7394 if (rspara) { /* have to do this both before and after */
7395 while (i != EOF) { /* to make sure file boundaries work right */
7396 i = PerlIO_getc(fp);
7398 PerlIO_ungetc(fp,i);
7404 return_string_or_null:
7405 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7411 Auto-increment of the value in the SV, doing string to numeric conversion
7412 if necessary. Handles 'get' magic and operator overloading.
7418 Perl_sv_inc(pTHX_ register SV *const sv)
7427 =for apidoc sv_inc_nomg
7429 Auto-increment of the value in the SV, doing string to numeric conversion
7430 if necessary. Handles operator overloading. Skips handling 'get' magic.
7436 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7444 if (SvTHINKFIRST(sv)) {
7446 sv_force_normal_flags(sv, 0);
7447 if (SvREADONLY(sv)) {
7448 if (IN_PERL_RUNTIME)
7449 Perl_croak_no_modify(aTHX);
7453 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
7455 i = PTR2IV(SvRV(sv));
7460 flags = SvFLAGS(sv);
7461 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7462 /* It's (privately or publicly) a float, but not tested as an
7463 integer, so test it to see. */
7465 flags = SvFLAGS(sv);
7467 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7468 /* It's publicly an integer, or privately an integer-not-float */
7469 #ifdef PERL_PRESERVE_IVUV
7473 if (SvUVX(sv) == UV_MAX)
7474 sv_setnv(sv, UV_MAX_P1);
7476 (void)SvIOK_only_UV(sv);
7477 SvUV_set(sv, SvUVX(sv) + 1);
7479 if (SvIVX(sv) == IV_MAX)
7480 sv_setuv(sv, (UV)IV_MAX + 1);
7482 (void)SvIOK_only(sv);
7483 SvIV_set(sv, SvIVX(sv) + 1);
7488 if (flags & SVp_NOK) {
7489 const NV was = SvNVX(sv);
7490 if (NV_OVERFLOWS_INTEGERS_AT &&
7491 was >= NV_OVERFLOWS_INTEGERS_AT) {
7492 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7493 "Lost precision when incrementing %" NVff " by 1",
7496 (void)SvNOK_only(sv);
7497 SvNV_set(sv, was + 1.0);
7501 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7502 if ((flags & SVTYPEMASK) < SVt_PVIV)
7503 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7504 (void)SvIOK_only(sv);
7509 while (isALPHA(*d)) d++;
7510 while (isDIGIT(*d)) d++;
7511 if (d < SvEND(sv)) {
7512 #ifdef PERL_PRESERVE_IVUV
7513 /* Got to punt this as an integer if needs be, but we don't issue
7514 warnings. Probably ought to make the sv_iv_please() that does
7515 the conversion if possible, and silently. */
7516 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7517 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7518 /* Need to try really hard to see if it's an integer.
7519 9.22337203685478e+18 is an integer.
7520 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7521 so $a="9.22337203685478e+18"; $a+0; $a++
7522 needs to be the same as $a="9.22337203685478e+18"; $a++
7529 /* sv_2iv *should* have made this an NV */
7530 if (flags & SVp_NOK) {
7531 (void)SvNOK_only(sv);
7532 SvNV_set(sv, SvNVX(sv) + 1.0);
7535 /* I don't think we can get here. Maybe I should assert this
7536 And if we do get here I suspect that sv_setnv will croak. NWC
7538 #if defined(USE_LONG_DOUBLE)
7539 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",
7540 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7542 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7543 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7546 #endif /* PERL_PRESERVE_IVUV */
7547 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7551 while (d >= SvPVX_const(sv)) {
7559 /* MKS: The original code here died if letters weren't consecutive.
7560 * at least it didn't have to worry about non-C locales. The
7561 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7562 * arranged in order (although not consecutively) and that only
7563 * [A-Za-z] are accepted by isALPHA in the C locale.
7565 if (*d != 'z' && *d != 'Z') {
7566 do { ++*d; } while (!isALPHA(*d));
7569 *(d--) -= 'z' - 'a';
7574 *(d--) -= 'z' - 'a' + 1;
7578 /* oh,oh, the number grew */
7579 SvGROW(sv, SvCUR(sv) + 2);
7580 SvCUR_set(sv, SvCUR(sv) + 1);
7581 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7592 Auto-decrement of the value in the SV, doing string to numeric conversion
7593 if necessary. Handles 'get' magic and operator overloading.
7599 Perl_sv_dec(pTHX_ register SV *const sv)
7609 =for apidoc sv_dec_nomg
7611 Auto-decrement of the value in the SV, doing string to numeric conversion
7612 if necessary. Handles operator overloading. Skips handling 'get' magic.
7618 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
7625 if (SvTHINKFIRST(sv)) {
7627 sv_force_normal_flags(sv, 0);
7628 if (SvREADONLY(sv)) {
7629 if (IN_PERL_RUNTIME)
7630 Perl_croak_no_modify(aTHX);
7634 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
7636 i = PTR2IV(SvRV(sv));
7641 /* Unlike sv_inc we don't have to worry about string-never-numbers
7642 and keeping them magic. But we mustn't warn on punting */
7643 flags = SvFLAGS(sv);
7644 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7645 /* It's publicly an integer, or privately an integer-not-float */
7646 #ifdef PERL_PRESERVE_IVUV
7650 if (SvUVX(sv) == 0) {
7651 (void)SvIOK_only(sv);
7655 (void)SvIOK_only_UV(sv);
7656 SvUV_set(sv, SvUVX(sv) - 1);
7659 if (SvIVX(sv) == IV_MIN) {
7660 sv_setnv(sv, (NV)IV_MIN);
7664 (void)SvIOK_only(sv);
7665 SvIV_set(sv, SvIVX(sv) - 1);
7670 if (flags & SVp_NOK) {
7673 const NV was = SvNVX(sv);
7674 if (NV_OVERFLOWS_INTEGERS_AT &&
7675 was <= -NV_OVERFLOWS_INTEGERS_AT) {
7676 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7677 "Lost precision when decrementing %" NVff " by 1",
7680 (void)SvNOK_only(sv);
7681 SvNV_set(sv, was - 1.0);
7685 if (!(flags & SVp_POK)) {
7686 if ((flags & SVTYPEMASK) < SVt_PVIV)
7687 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
7689 (void)SvIOK_only(sv);
7692 #ifdef PERL_PRESERVE_IVUV
7694 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7695 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7696 /* Need to try really hard to see if it's an integer.
7697 9.22337203685478e+18 is an integer.
7698 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7699 so $a="9.22337203685478e+18"; $a+0; $a--
7700 needs to be the same as $a="9.22337203685478e+18"; $a--
7707 /* sv_2iv *should* have made this an NV */
7708 if (flags & SVp_NOK) {
7709 (void)SvNOK_only(sv);
7710 SvNV_set(sv, SvNVX(sv) - 1.0);
7713 /* I don't think we can get here. Maybe I should assert this
7714 And if we do get here I suspect that sv_setnv will croak. NWC
7716 #if defined(USE_LONG_DOUBLE)
7717 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",
7718 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7720 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7721 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7725 #endif /* PERL_PRESERVE_IVUV */
7726 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
7729 /* this define is used to eliminate a chunk of duplicated but shared logic
7730 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
7731 * used anywhere but here - yves
7733 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
7736 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
7740 =for apidoc sv_mortalcopy
7742 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
7743 The new SV is marked as mortal. It will be destroyed "soon", either by an
7744 explicit call to FREETMPS, or by an implicit call at places such as
7745 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
7750 /* Make a string that will exist for the duration of the expression
7751 * evaluation. Actually, it may have to last longer than that, but
7752 * hopefully we won't free it until it has been assigned to a
7753 * permanent location. */
7756 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
7762 sv_setsv(sv,oldstr);
7763 PUSH_EXTEND_MORTAL__SV_C(sv);
7769 =for apidoc sv_newmortal
7771 Creates a new null SV which is mortal. The reference count of the SV is
7772 set to 1. It will be destroyed "soon", either by an explicit call to
7773 FREETMPS, or by an implicit call at places such as statement boundaries.
7774 See also C<sv_mortalcopy> and C<sv_2mortal>.
7780 Perl_sv_newmortal(pTHX)
7786 SvFLAGS(sv) = SVs_TEMP;
7787 PUSH_EXTEND_MORTAL__SV_C(sv);
7793 =for apidoc newSVpvn_flags
7795 Creates a new SV and copies a string into it. The reference count for the
7796 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7797 string. You are responsible for ensuring that the source string is at least
7798 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7799 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
7800 If C<SVs_TEMP> is set, then C<sv2mortal()> is called on the result before
7801 returning. If C<SVf_UTF8> is set, C<s> is considered to be in UTF-8 and the
7802 C<SVf_UTF8> flag will be set on the new SV.
7803 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
7805 #define newSVpvn_utf8(s, len, u) \
7806 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
7812 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
7817 /* All the flags we don't support must be zero.
7818 And we're new code so I'm going to assert this from the start. */
7819 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
7821 sv_setpvn(sv,s,len);
7823 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
7824 * and do what it does outselves here.
7825 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
7826 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
7827 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
7828 * eleminate quite a few steps than it looks - Yves (explaining patch by gfx)
7831 SvFLAGS(sv) |= flags;
7833 if(flags & SVs_TEMP){
7834 PUSH_EXTEND_MORTAL__SV_C(sv);
7841 =for apidoc sv_2mortal
7843 Marks an existing SV as mortal. The SV will be destroyed "soon", either
7844 by an explicit call to FREETMPS, or by an implicit call at places such as
7845 statement boundaries. SvTEMP() is turned on which means that the SV's
7846 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
7847 and C<sv_mortalcopy>.
7853 Perl_sv_2mortal(pTHX_ register SV *const sv)
7858 if (SvREADONLY(sv) && SvIMMORTAL(sv))
7860 PUSH_EXTEND_MORTAL__SV_C(sv);
7868 Creates a new SV and copies a string into it. The reference count for the
7869 SV is set to 1. If C<len> is zero, Perl will compute the length using
7870 strlen(). For efficiency, consider using C<newSVpvn> instead.
7876 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
7882 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
7887 =for apidoc newSVpvn
7889 Creates a new SV and copies a string into it. The reference count for the
7890 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7891 string. You are responsible for ensuring that the source string is at least
7892 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7898 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
7904 sv_setpvn(sv,s,len);
7909 =for apidoc newSVhek
7911 Creates a new SV from the hash key structure. It will generate scalars that
7912 point to the shared string table where possible. Returns a new (undefined)
7913 SV if the hek is NULL.
7919 Perl_newSVhek(pTHX_ const HEK *const hek)
7929 if (HEK_LEN(hek) == HEf_SVKEY) {
7930 return newSVsv(*(SV**)HEK_KEY(hek));
7932 const int flags = HEK_FLAGS(hek);
7933 if (flags & HVhek_WASUTF8) {
7935 Andreas would like keys he put in as utf8 to come back as utf8
7937 STRLEN utf8_len = HEK_LEN(hek);
7938 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
7939 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
7942 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
7944 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
7945 /* We don't have a pointer to the hv, so we have to replicate the
7946 flag into every HEK. This hv is using custom a hasing
7947 algorithm. Hence we can't return a shared string scalar, as
7948 that would contain the (wrong) hash value, and might get passed
7949 into an hv routine with a regular hash.
7950 Similarly, a hash that isn't using shared hash keys has to have
7951 the flag in every key so that we know not to try to call
7952 share_hek_kek on it. */
7954 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
7959 /* This will be overwhelminly the most common case. */
7961 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
7962 more efficient than sharepvn(). */
7966 sv_upgrade(sv, SVt_PV);
7967 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7968 SvCUR_set(sv, HEK_LEN(hek));
7981 =for apidoc newSVpvn_share
7983 Creates a new SV with its SvPVX_const pointing to a shared string in the string
7984 table. If the string does not already exist in the table, it is created
7985 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
7986 value is used; otherwise the hash is computed. The string's hash can be later
7987 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
7988 that as the string table is used for shared hash keys these strings will have
7989 SvPVX_const == HeKEY and hash lookup will avoid string compare.
7995 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
7999 bool is_utf8 = FALSE;
8000 const char *const orig_src = src;
8003 STRLEN tmplen = -len;
8005 /* See the note in hv.c:hv_fetch() --jhi */
8006 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8010 PERL_HASH(hash, src, len);
8012 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8013 changes here, update it there too. */
8014 sv_upgrade(sv, SVt_PV);
8015 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8023 if (src != orig_src)
8029 #if defined(PERL_IMPLICIT_CONTEXT)
8031 /* pTHX_ magic can't cope with varargs, so this is a no-context
8032 * version of the main function, (which may itself be aliased to us).
8033 * Don't access this version directly.
8037 Perl_newSVpvf_nocontext(const char *const pat, ...)
8043 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8045 va_start(args, pat);
8046 sv = vnewSVpvf(pat, &args);
8053 =for apidoc newSVpvf
8055 Creates a new SV and initializes it with the string formatted like
8062 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8067 PERL_ARGS_ASSERT_NEWSVPVF;
8069 va_start(args, pat);
8070 sv = vnewSVpvf(pat, &args);
8075 /* backend for newSVpvf() and newSVpvf_nocontext() */
8078 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8083 PERL_ARGS_ASSERT_VNEWSVPVF;
8086 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8093 Creates a new SV and copies a floating point value into it.
8094 The reference count for the SV is set to 1.
8100 Perl_newSVnv(pTHX_ const NV n)
8113 Creates a new SV and copies an integer into it. The reference count for the
8120 Perl_newSViv(pTHX_ const IV i)
8133 Creates a new SV and copies an unsigned integer into it.
8134 The reference count for the SV is set to 1.
8140 Perl_newSVuv(pTHX_ const UV u)
8151 =for apidoc newSV_type
8153 Creates a new SV, of the type specified. The reference count for the new SV
8160 Perl_newSV_type(pTHX_ const svtype type)
8165 sv_upgrade(sv, type);
8170 =for apidoc newRV_noinc
8172 Creates an RV wrapper for an SV. The reference count for the original
8173 SV is B<not> incremented.
8179 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8182 register SV *sv = newSV_type(SVt_IV);
8184 PERL_ARGS_ASSERT_NEWRV_NOINC;
8187 SvRV_set(sv, tmpRef);
8192 /* newRV_inc is the official function name to use now.
8193 * newRV_inc is in fact #defined to newRV in sv.h
8197 Perl_newRV(pTHX_ SV *const sv)
8201 PERL_ARGS_ASSERT_NEWRV;
8203 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8209 Creates a new SV which is an exact duplicate of the original SV.
8216 Perl_newSVsv(pTHX_ register SV *const old)
8223 if (SvTYPE(old) == SVTYPEMASK) {
8224 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8228 /* SV_GMAGIC is the default for sv_setv()
8229 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8230 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8231 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8236 =for apidoc sv_reset
8238 Underlying implementation for the C<reset> Perl function.
8239 Note that the perl-level function is vaguely deprecated.
8245 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8248 char todo[PERL_UCHAR_MAX+1];
8250 PERL_ARGS_ASSERT_SV_RESET;
8255 if (!*s) { /* reset ?? searches */
8256 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8258 const U32 count = mg->mg_len / sizeof(PMOP**);
8259 PMOP **pmp = (PMOP**) mg->mg_ptr;
8260 PMOP *const *const end = pmp + count;
8264 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8266 (*pmp)->op_pmflags &= ~PMf_USED;
8274 /* reset variables */
8276 if (!HvARRAY(stash))
8279 Zero(todo, 256, char);
8282 I32 i = (unsigned char)*s;
8286 max = (unsigned char)*s++;
8287 for ( ; i <= max; i++) {
8290 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8292 for (entry = HvARRAY(stash)[i];
8294 entry = HeNEXT(entry))
8299 if (!todo[(U8)*HeKEY(entry)])
8301 gv = MUTABLE_GV(HeVAL(entry));
8304 if (SvTHINKFIRST(sv)) {
8305 if (!SvREADONLY(sv) && SvROK(sv))
8307 /* XXX Is this continue a bug? Why should THINKFIRST
8308 exempt us from resetting arrays and hashes? */
8312 if (SvTYPE(sv) >= SVt_PV) {
8314 if (SvPVX_const(sv) != NULL)
8322 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8324 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8327 # if defined(USE_ENVIRON_ARRAY)
8330 # endif /* USE_ENVIRON_ARRAY */
8341 Using various gambits, try to get an IO from an SV: the IO slot if its a
8342 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8343 named after the PV if we're a string.
8349 Perl_sv_2io(pTHX_ SV *const sv)
8354 PERL_ARGS_ASSERT_SV_2IO;
8356 switch (SvTYPE(sv)) {
8358 io = MUTABLE_IO(sv);
8361 if (isGV_with_GP(sv)) {
8362 gv = MUTABLE_GV(sv);
8365 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8371 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8373 return sv_2io(SvRV(sv));
8374 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8380 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8389 Using various gambits, try to get a CV from an SV; in addition, try if
8390 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8391 The flags in C<lref> are passed to gv_fetchsv.
8397 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8403 PERL_ARGS_ASSERT_SV_2CV;
8410 switch (SvTYPE(sv)) {
8414 return MUTABLE_CV(sv);
8421 if (isGV_with_GP(sv)) {
8422 gv = MUTABLE_GV(sv);
8431 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
8433 tryAMAGICunDEREF(to_cv);
8436 if (SvTYPE(sv) == SVt_PVCV) {
8437 cv = MUTABLE_CV(sv);
8442 else if(isGV_with_GP(sv))
8443 gv = MUTABLE_GV(sv);
8445 Perl_croak(aTHX_ "Not a subroutine reference");
8447 else if (isGV_with_GP(sv)) {
8449 gv = MUTABLE_GV(sv);
8452 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8458 /* Some flags to gv_fetchsv mean don't really create the GV */
8459 if (!isGV_with_GP(gv)) {
8465 if (lref && !GvCVu(gv)) {
8469 gv_efullname3(tmpsv, gv, NULL);
8470 /* XXX this is probably not what they think they're getting.
8471 * It has the same effect as "sub name;", i.e. just a forward
8473 newSUB(start_subparse(FALSE, 0),
8474 newSVOP(OP_CONST, 0, tmpsv),
8478 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8479 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8488 Returns true if the SV has a true value by Perl's rules.
8489 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8490 instead use an in-line version.
8496 Perl_sv_true(pTHX_ register SV *const sv)
8501 register const XPV* const tXpv = (XPV*)SvANY(sv);
8503 (tXpv->xpv_cur > 1 ||
8504 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8511 return SvIVX(sv) != 0;
8514 return SvNVX(sv) != 0.0;
8516 return sv_2bool(sv);
8522 =for apidoc sv_pvn_force
8524 Get a sensible string out of the SV somehow.
8525 A private implementation of the C<SvPV_force> macro for compilers which
8526 can't cope with complex macro expressions. Always use the macro instead.
8528 =for apidoc sv_pvn_force_flags
8530 Get a sensible string out of the SV somehow.
8531 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8532 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8533 implemented in terms of this function.
8534 You normally want to use the various wrapper macros instead: see
8535 C<SvPV_force> and C<SvPV_force_nomg>
8541 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8545 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8547 if (SvTHINKFIRST(sv) && !SvROK(sv))
8548 sv_force_normal_flags(sv, 0);
8558 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8559 const char * const ref = sv_reftype(sv,0);
8561 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8562 ref, OP_DESC(PL_op));
8564 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8566 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8567 || isGV_with_GP(sv))
8568 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8570 s = sv_2pv_flags(sv, &len, flags);
8574 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8577 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8578 SvGROW(sv, len + 1);
8579 Move(s,SvPVX(sv),len,char);
8581 SvPVX(sv)[len] = '\0';
8584 SvPOK_on(sv); /* validate pointer */
8586 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
8587 PTR2UV(sv),SvPVX_const(sv)));
8590 return SvPVX_mutable(sv);
8594 =for apidoc sv_pvbyten_force
8596 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
8602 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
8604 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
8606 sv_pvn_force(sv,lp);
8607 sv_utf8_downgrade(sv,0);
8613 =for apidoc sv_pvutf8n_force
8615 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
8621 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
8623 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
8625 sv_pvn_force(sv,lp);
8626 sv_utf8_upgrade(sv);
8632 =for apidoc sv_reftype
8634 Returns a string describing what the SV is a reference to.
8640 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
8642 PERL_ARGS_ASSERT_SV_REFTYPE;
8644 /* The fact that I don't need to downcast to char * everywhere, only in ?:
8645 inside return suggests a const propagation bug in g++. */
8646 if (ob && SvOBJECT(sv)) {
8647 char * const name = HvNAME_get(SvSTASH(sv));
8648 return name ? name : (char *) "__ANON__";
8651 switch (SvTYPE(sv)) {
8666 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
8667 /* tied lvalues should appear to be
8668 * scalars for backwards compatitbility */
8669 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
8670 ? "SCALAR" : "LVALUE");
8671 case SVt_PVAV: return "ARRAY";
8672 case SVt_PVHV: return "HASH";
8673 case SVt_PVCV: return "CODE";
8674 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
8675 ? "GLOB" : "SCALAR");
8676 case SVt_PVFM: return "FORMAT";
8677 case SVt_PVIO: return "IO";
8678 case SVt_BIND: return "BIND";
8679 case SVt_REGEXP: return "REGEXP";
8680 default: return "UNKNOWN";
8686 =for apidoc sv_isobject
8688 Returns a boolean indicating whether the SV is an RV pointing to a blessed
8689 object. If the SV is not an RV, or if the object is not blessed, then this
8696 Perl_sv_isobject(pTHX_ SV *sv)
8712 Returns a boolean indicating whether the SV is blessed into the specified
8713 class. This does not check for subtypes; use C<sv_derived_from> to verify
8714 an inheritance relationship.
8720 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
8724 PERL_ARGS_ASSERT_SV_ISA;
8734 hvname = HvNAME_get(SvSTASH(sv));
8738 return strEQ(hvname, name);
8744 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
8745 it will be upgraded to one. If C<classname> is non-null then the new SV will
8746 be blessed in the specified package. The new SV is returned and its
8747 reference count is 1.
8753 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
8758 PERL_ARGS_ASSERT_NEWSVRV;
8762 SV_CHECK_THINKFIRST_COW_DROP(rv);
8763 (void)SvAMAGIC_off(rv);
8765 if (SvTYPE(rv) >= SVt_PVMG) {
8766 const U32 refcnt = SvREFCNT(rv);
8770 SvREFCNT(rv) = refcnt;
8772 sv_upgrade(rv, SVt_IV);
8773 } else if (SvROK(rv)) {
8774 SvREFCNT_dec(SvRV(rv));
8776 prepare_SV_for_RV(rv);
8784 HV* const stash = gv_stashpv(classname, GV_ADD);
8785 (void)sv_bless(rv, stash);
8791 =for apidoc sv_setref_pv
8793 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
8794 argument will be upgraded to an RV. That RV will be modified to point to
8795 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
8796 into the SV. The C<classname> argument indicates the package for the
8797 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8798 will have a reference count of 1, and the RV will be returned.
8800 Do not use with other Perl types such as HV, AV, SV, CV, because those
8801 objects will become corrupted by the pointer copy process.
8803 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
8809 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
8813 PERL_ARGS_ASSERT_SV_SETREF_PV;
8816 sv_setsv(rv, &PL_sv_undef);
8820 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
8825 =for apidoc sv_setref_iv
8827 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
8828 argument will be upgraded to an RV. That RV will be modified to point to
8829 the new SV. The C<classname> argument indicates the package for the
8830 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8831 will have a reference count of 1, and the RV will be returned.
8837 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
8839 PERL_ARGS_ASSERT_SV_SETREF_IV;
8841 sv_setiv(newSVrv(rv,classname), iv);
8846 =for apidoc sv_setref_uv
8848 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
8849 argument will be upgraded to an RV. That RV will be modified to point to
8850 the new SV. The C<classname> argument indicates the package for the
8851 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8852 will have a reference count of 1, and the RV will be returned.
8858 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
8860 PERL_ARGS_ASSERT_SV_SETREF_UV;
8862 sv_setuv(newSVrv(rv,classname), uv);
8867 =for apidoc sv_setref_nv
8869 Copies a double into a new SV, optionally blessing the SV. The C<rv>
8870 argument will be upgraded to an RV. That RV will be modified to point to
8871 the new SV. The C<classname> argument indicates the package for the
8872 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8873 will have a reference count of 1, and the RV will be returned.
8879 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
8881 PERL_ARGS_ASSERT_SV_SETREF_NV;
8883 sv_setnv(newSVrv(rv,classname), nv);
8888 =for apidoc sv_setref_pvn
8890 Copies a string into a new SV, optionally blessing the SV. The length of the
8891 string must be specified with C<n>. The C<rv> argument will be upgraded to
8892 an RV. That RV will be modified to point to the new SV. The C<classname>
8893 argument indicates the package for the blessing. Set C<classname> to
8894 C<NULL> to avoid the blessing. The new SV will have a reference count
8895 of 1, and the RV will be returned.
8897 Note that C<sv_setref_pv> copies the pointer while this copies the string.
8903 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
8904 const char *const pv, const STRLEN n)
8906 PERL_ARGS_ASSERT_SV_SETREF_PVN;
8908 sv_setpvn(newSVrv(rv,classname), pv, n);
8913 =for apidoc sv_bless
8915 Blesses an SV into a specified package. The SV must be an RV. The package
8916 must be designated by its stash (see C<gv_stashpv()>). The reference count
8917 of the SV is unaffected.
8923 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
8928 PERL_ARGS_ASSERT_SV_BLESS;
8931 Perl_croak(aTHX_ "Can't bless non-reference value");
8933 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
8934 if (SvIsCOW(tmpRef))
8935 sv_force_normal_flags(tmpRef, 0);
8936 if (SvREADONLY(tmpRef))
8937 Perl_croak_no_modify(aTHX);
8938 if (SvOBJECT(tmpRef)) {
8939 if (SvTYPE(tmpRef) != SVt_PVIO)
8941 SvREFCNT_dec(SvSTASH(tmpRef));
8944 SvOBJECT_on(tmpRef);
8945 if (SvTYPE(tmpRef) != SVt_PVIO)
8947 SvUPGRADE(tmpRef, SVt_PVMG);
8948 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
8953 (void)SvAMAGIC_off(sv);
8955 if(SvSMAGICAL(tmpRef))
8956 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
8964 /* Downgrades a PVGV to a PVMG.
8968 S_sv_unglob(pTHX_ SV *const sv)
8973 SV * const temp = sv_newmortal();
8975 PERL_ARGS_ASSERT_SV_UNGLOB;
8977 assert(SvTYPE(sv) == SVt_PVGV);
8979 gv_efullname3(temp, MUTABLE_GV(sv), "*");
8982 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
8983 && HvNAME_get(stash))
8984 mro_method_changed_in(stash);
8985 gp_free(MUTABLE_GV(sv));
8988 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
8992 if (GvNAME_HEK(sv)) {
8993 unshare_hek(GvNAME_HEK(sv));
8995 isGV_with_GP_off(sv);
8997 /* need to keep SvANY(sv) in the right arena */
8998 xpvmg = new_XPVMG();
8999 StructCopy(SvANY(sv), xpvmg, XPVMG);
9000 del_XPVGV(SvANY(sv));
9003 SvFLAGS(sv) &= ~SVTYPEMASK;
9004 SvFLAGS(sv) |= SVt_PVMG;
9006 /* Intentionally not calling any local SET magic, as this isn't so much a
9007 set operation as merely an internal storage change. */
9008 sv_setsv_flags(sv, temp, 0);
9012 =for apidoc sv_unref_flags
9014 Unsets the RV status of the SV, and decrements the reference count of
9015 whatever was being referenced by the RV. This can almost be thought of
9016 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9017 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9018 (otherwise the decrementing is conditional on the reference count being
9019 different from one or the reference being a readonly SV).
9026 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9028 SV* const target = SvRV(ref);
9030 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9032 if (SvWEAKREF(ref)) {
9033 sv_del_backref(target, ref);
9035 SvRV_set(ref, NULL);
9038 SvRV_set(ref, NULL);
9040 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9041 assigned to as BEGIN {$a = \"Foo"} will fail. */
9042 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9043 SvREFCNT_dec(target);
9044 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9045 sv_2mortal(target); /* Schedule for freeing later */
9049 =for apidoc sv_untaint
9051 Untaint an SV. Use C<SvTAINTED_off> instead.
9056 Perl_sv_untaint(pTHX_ SV *const sv)
9058 PERL_ARGS_ASSERT_SV_UNTAINT;
9060 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9061 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9068 =for apidoc sv_tainted
9070 Test an SV for taintedness. Use C<SvTAINTED> instead.
9075 Perl_sv_tainted(pTHX_ SV *const sv)
9077 PERL_ARGS_ASSERT_SV_TAINTED;
9079 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9080 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9081 if (mg && (mg->mg_len & 1) )
9088 =for apidoc sv_setpviv
9090 Copies an integer into the given SV, also updating its string value.
9091 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9097 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9099 char buf[TYPE_CHARS(UV)];
9101 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9103 PERL_ARGS_ASSERT_SV_SETPVIV;
9105 sv_setpvn(sv, ptr, ebuf - ptr);
9109 =for apidoc sv_setpviv_mg
9111 Like C<sv_setpviv>, but also handles 'set' magic.
9117 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9119 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9125 #if defined(PERL_IMPLICIT_CONTEXT)
9127 /* pTHX_ magic can't cope with varargs, so this is a no-context
9128 * version of the main function, (which may itself be aliased to us).
9129 * Don't access this version directly.
9133 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9138 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9140 va_start(args, pat);
9141 sv_vsetpvf(sv, pat, &args);
9145 /* pTHX_ magic can't cope with varargs, so this is a no-context
9146 * version of the main function, (which may itself be aliased to us).
9147 * Don't access this version directly.
9151 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9156 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9158 va_start(args, pat);
9159 sv_vsetpvf_mg(sv, pat, &args);
9165 =for apidoc sv_setpvf
9167 Works like C<sv_catpvf> but copies the text into the SV instead of
9168 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9174 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9178 PERL_ARGS_ASSERT_SV_SETPVF;
9180 va_start(args, pat);
9181 sv_vsetpvf(sv, pat, &args);
9186 =for apidoc sv_vsetpvf
9188 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9189 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9191 Usually used via its frontend C<sv_setpvf>.
9197 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9199 PERL_ARGS_ASSERT_SV_VSETPVF;
9201 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9205 =for apidoc sv_setpvf_mg
9207 Like C<sv_setpvf>, but also handles 'set' magic.
9213 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9217 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9219 va_start(args, pat);
9220 sv_vsetpvf_mg(sv, pat, &args);
9225 =for apidoc sv_vsetpvf_mg
9227 Like C<sv_vsetpvf>, but also handles 'set' magic.
9229 Usually used via its frontend C<sv_setpvf_mg>.
9235 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9237 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9239 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9243 #if defined(PERL_IMPLICIT_CONTEXT)
9245 /* pTHX_ magic can't cope with varargs, so this is a no-context
9246 * version of the main function, (which may itself be aliased to us).
9247 * Don't access this version directly.
9251 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9256 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9258 va_start(args, pat);
9259 sv_vcatpvf(sv, pat, &args);
9263 /* pTHX_ magic can't cope with varargs, so this is a no-context
9264 * version of the main function, (which may itself be aliased to us).
9265 * Don't access this version directly.
9269 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9274 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9276 va_start(args, pat);
9277 sv_vcatpvf_mg(sv, pat, &args);
9283 =for apidoc sv_catpvf
9285 Processes its arguments like C<sprintf> and appends the formatted
9286 output to an SV. If the appended data contains "wide" characters
9287 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9288 and characters >255 formatted with %c), the original SV might get
9289 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9290 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9291 valid UTF-8; if the original SV was bytes, the pattern should be too.
9296 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9300 PERL_ARGS_ASSERT_SV_CATPVF;
9302 va_start(args, pat);
9303 sv_vcatpvf(sv, pat, &args);
9308 =for apidoc sv_vcatpvf
9310 Processes its arguments like C<vsprintf> and appends the formatted output
9311 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9313 Usually used via its frontend C<sv_catpvf>.
9319 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9321 PERL_ARGS_ASSERT_SV_VCATPVF;
9323 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9327 =for apidoc sv_catpvf_mg
9329 Like C<sv_catpvf>, but also handles 'set' magic.
9335 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9339 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9341 va_start(args, pat);
9342 sv_vcatpvf_mg(sv, pat, &args);
9347 =for apidoc sv_vcatpvf_mg
9349 Like C<sv_vcatpvf>, but also handles 'set' magic.
9351 Usually used via its frontend C<sv_catpvf_mg>.
9357 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9359 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9361 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9366 =for apidoc sv_vsetpvfn
9368 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9371 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9377 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9378 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9380 PERL_ARGS_ASSERT_SV_VSETPVFN;
9383 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9388 * Warn of missing argument to sprintf, and then return a defined value
9389 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9391 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9393 S_vcatpvfn_missing_argument(pTHX) {
9394 if (ckWARN(WARN_MISSING)) {
9395 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9396 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9403 S_expect_number(pTHX_ char **const pattern)
9408 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9410 switch (**pattern) {
9411 case '1': case '2': case '3':
9412 case '4': case '5': case '6':
9413 case '7': case '8': case '9':
9414 var = *(*pattern)++ - '0';
9415 while (isDIGIT(**pattern)) {
9416 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9418 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9426 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9428 const int neg = nv < 0;
9431 PERL_ARGS_ASSERT_F0CONVERT;
9439 if (uv & 1 && uv == nv)
9440 uv--; /* Round to even */
9442 const unsigned dig = uv % 10;
9455 =for apidoc sv_vcatpvfn
9457 Processes its arguments like C<vsprintf> and appends the formatted output
9458 to an SV. Uses an array of SVs if the C style variable argument list is
9459 missing (NULL). When running with taint checks enabled, indicates via
9460 C<maybe_tainted> if results are untrustworthy (often due to the use of
9463 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9469 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9470 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9471 vec_utf8 = DO_UTF8(vecsv);
9473 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9476 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9477 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9485 static const char nullstr[] = "(null)";
9487 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9488 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9490 /* Times 4: a decimal digit takes more than 3 binary digits.
9491 * NV_DIG: mantissa takes than many decimal digits.
9492 * Plus 32: Playing safe. */
9493 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9494 /* large enough for "%#.#f" --chip */
9495 /* what about long double NVs? --jhi */
9497 PERL_ARGS_ASSERT_SV_VCATPVFN;
9498 PERL_UNUSED_ARG(maybe_tainted);
9500 /* no matter what, this is a string now */
9501 (void)SvPV_force(sv, origlen);
9503 /* special-case "", "%s", and "%-p" (SVf - see below) */
9506 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9508 const char * const s = va_arg(*args, char*);
9509 sv_catpv(sv, s ? s : nullstr);
9511 else if (svix < svmax) {
9512 sv_catsv(sv, *svargs);
9515 S_vcatpvfn_missing_argument(aTHX);
9518 if (args && patlen == 3 && pat[0] == '%' &&
9519 pat[1] == '-' && pat[2] == 'p') {
9520 argsv = MUTABLE_SV(va_arg(*args, void*));
9521 sv_catsv(sv, argsv);
9525 #ifndef USE_LONG_DOUBLE
9526 /* special-case "%.<number>[gf]" */
9527 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9528 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9529 unsigned digits = 0;
9533 while (*pp >= '0' && *pp <= '9')
9534 digits = 10 * digits + (*pp++ - '0');
9535 if (pp - pat == (int)patlen - 1 && svix < svmax) {
9536 const NV nv = SvNV(*svargs);
9538 /* Add check for digits != 0 because it seems that some
9539 gconverts are buggy in this case, and we don't yet have
9540 a Configure test for this. */
9541 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9542 /* 0, point, slack */
9543 Gconvert(nv, (int)digits, 0, ebuf);
9545 if (*ebuf) /* May return an empty string for digits==0 */
9548 } else if (!digits) {
9551 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9552 sv_catpvn(sv, p, l);
9558 #endif /* !USE_LONG_DOUBLE */
9560 if (!args && svix < svmax && DO_UTF8(*svargs))
9563 patend = (char*)pat + patlen;
9564 for (p = (char*)pat; p < patend; p = q) {
9567 bool vectorize = FALSE;
9568 bool vectorarg = FALSE;
9569 bool vec_utf8 = FALSE;
9575 bool has_precis = FALSE;
9577 const I32 osvix = svix;
9578 bool is_utf8 = FALSE; /* is this item utf8? */
9579 #ifdef HAS_LDBL_SPRINTF_BUG
9580 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
9581 with sfio - Allen <allens@cpan.org> */
9582 bool fix_ldbl_sprintf_bug = FALSE;
9586 U8 utf8buf[UTF8_MAXBYTES+1];
9587 STRLEN esignlen = 0;
9589 const char *eptr = NULL;
9590 const char *fmtstart;
9593 const U8 *vecstr = NULL;
9600 /* we need a long double target in case HAS_LONG_DOUBLE but
9603 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9611 const char *dotstr = ".";
9612 STRLEN dotstrlen = 1;
9613 I32 efix = 0; /* explicit format parameter index */
9614 I32 ewix = 0; /* explicit width index */
9615 I32 epix = 0; /* explicit precision index */
9616 I32 evix = 0; /* explicit vector index */
9617 bool asterisk = FALSE;
9619 /* echo everything up to the next format specification */
9620 for (q = p; q < patend && *q != '%'; ++q) ;
9622 if (has_utf8 && !pat_utf8)
9623 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
9625 sv_catpvn(sv, p, q - p);
9634 We allow format specification elements in this order:
9635 \d+\$ explicit format parameter index
9637 v|\*(\d+\$)?v vector with optional (optionally specified) arg
9638 0 flag (as above): repeated to allow "v02"
9639 \d+|\*(\d+\$)? width using optional (optionally specified) arg
9640 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
9642 [%bcdefginopsuxDFOUX] format (mandatory)
9647 As of perl5.9.3, printf format checking is on by default.
9648 Internally, perl uses %p formats to provide an escape to
9649 some extended formatting. This block deals with those
9650 extensions: if it does not match, (char*)q is reset and
9651 the normal format processing code is used.
9653 Currently defined extensions are:
9654 %p include pointer address (standard)
9655 %-p (SVf) include an SV (previously %_)
9656 %-<num>p include an SV with precision <num>
9657 %<num>p reserved for future extensions
9659 Robin Barker 2005-07-14
9661 %1p (VDf) removed. RMB 2007-10-19
9668 n = expect_number(&q);
9675 argsv = MUTABLE_SV(va_arg(*args, void*));
9676 eptr = SvPV_const(argsv, elen);
9682 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
9683 "internal %%<num>p might conflict with future printf extensions");
9689 if ( (width = expect_number(&q)) ) {
9704 if (plus == '+' && *q == ' ') /* '+' over ' ' */
9733 if ( (ewix = expect_number(&q)) )
9742 if ((vectorarg = asterisk)) {
9755 width = expect_number(&q);
9761 vecsv = va_arg(*args, SV*);
9763 vecsv = (evix > 0 && evix <= svmax)
9764 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
9766 vecsv = svix < svmax
9767 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
9769 dotstr = SvPV_const(vecsv, dotstrlen);
9770 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
9771 bad with tied or overloaded values that return UTF8. */
9774 else if (has_utf8) {
9775 vecsv = sv_mortalcopy(vecsv);
9776 sv_utf8_upgrade(vecsv);
9777 dotstr = SvPV_const(vecsv, dotstrlen);
9784 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
9785 vecsv = svargs[efix ? efix-1 : svix++];
9786 vecstr = (U8*)SvPV_const(vecsv,veclen);
9787 vec_utf8 = DO_UTF8(vecsv);
9789 /* if this is a version object, we need to convert
9790 * back into v-string notation and then let the
9791 * vectorize happen normally
9793 if (sv_derived_from(vecsv, "version")) {
9794 char *version = savesvpv(vecsv);
9795 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
9796 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9797 "vector argument not supported with alpha versions");
9800 vecsv = sv_newmortal();
9801 scan_vstring(version, version + veclen, vecsv);
9802 vecstr = (U8*)SvPV_const(vecsv, veclen);
9803 vec_utf8 = DO_UTF8(vecsv);
9815 i = va_arg(*args, int);
9817 i = (ewix ? ewix <= svmax : svix < svmax) ?
9818 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9820 width = (i < 0) ? -i : i;
9830 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
9832 /* XXX: todo, support specified precision parameter */
9836 i = va_arg(*args, int);
9838 i = (ewix ? ewix <= svmax : svix < svmax)
9839 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9841 has_precis = !(i < 0);
9846 precis = precis * 10 + (*q++ - '0');
9855 case 'I': /* Ix, I32x, and I64x */
9857 if (q[1] == '6' && q[2] == '4') {
9863 if (q[1] == '3' && q[2] == '2') {
9873 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9884 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9885 if (*(q + 1) == 'l') { /* lld, llf */
9911 if (!vectorize && !args) {
9913 const I32 i = efix-1;
9914 argsv = (i >= 0 && i < svmax)
9915 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
9917 argsv = (svix >= 0 && svix < svmax)
9918 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
9929 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
9931 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
9933 eptr = (char*)utf8buf;
9934 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
9948 eptr = va_arg(*args, char*);
9950 elen = strlen(eptr);
9952 eptr = (char *)nullstr;
9953 elen = sizeof nullstr - 1;
9957 eptr = SvPV_const(argsv, elen);
9958 if (DO_UTF8(argsv)) {
9959 STRLEN old_precis = precis;
9960 if (has_precis && precis < elen) {
9961 STRLEN ulen = sv_len_utf8(argsv);
9962 I32 p = precis > ulen ? ulen : precis;
9963 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
9966 if (width) { /* fudge width (can't fudge elen) */
9967 if (has_precis && precis < elen)
9968 width += precis - old_precis;
9970 width += elen - sv_len_utf8(argsv);
9977 if (has_precis && precis < elen)
9984 if (alt || vectorize)
9986 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10007 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10016 esignbuf[esignlen++] = plus;
10020 case 'h': iv = (short)va_arg(*args, int); break;
10021 case 'l': iv = va_arg(*args, long); break;
10022 case 'V': iv = va_arg(*args, IV); break;
10023 default: iv = va_arg(*args, int); break;
10026 iv = va_arg(*args, Quad_t); break;
10033 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10035 case 'h': iv = (short)tiv; break;
10036 case 'l': iv = (long)tiv; break;
10038 default: iv = tiv; break;
10041 iv = (Quad_t)tiv; break;
10047 if ( !vectorize ) /* we already set uv above */
10052 esignbuf[esignlen++] = plus;
10056 esignbuf[esignlen++] = '-';
10100 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10111 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10112 case 'l': uv = va_arg(*args, unsigned long); break;
10113 case 'V': uv = va_arg(*args, UV); break;
10114 default: uv = va_arg(*args, unsigned); break;
10117 uv = va_arg(*args, Uquad_t); break;
10124 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10126 case 'h': uv = (unsigned short)tuv; break;
10127 case 'l': uv = (unsigned long)tuv; break;
10129 default: uv = tuv; break;
10132 uv = (Uquad_t)tuv; break;
10141 char *ptr = ebuf + sizeof ebuf;
10142 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10148 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10152 } while (uv >>= 4);
10154 esignbuf[esignlen++] = '0';
10155 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10161 *--ptr = '0' + dig;
10162 } while (uv >>= 3);
10163 if (alt && *ptr != '0')
10169 *--ptr = '0' + dig;
10170 } while (uv >>= 1);
10172 esignbuf[esignlen++] = '0';
10173 esignbuf[esignlen++] = c;
10176 default: /* it had better be ten or less */
10179 *--ptr = '0' + dig;
10180 } while (uv /= base);
10183 elen = (ebuf + sizeof ebuf) - ptr;
10187 zeros = precis - elen;
10188 else if (precis == 0 && elen == 1 && *eptr == '0'
10189 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10192 /* a precision nullifies the 0 flag. */
10199 /* FLOATING POINT */
10202 c = 'f'; /* maybe %F isn't supported here */
10204 case 'e': case 'E':
10206 case 'g': case 'G':
10210 /* This is evil, but floating point is even more evil */
10212 /* for SV-style calling, we can only get NV
10213 for C-style calling, we assume %f is double;
10214 for simplicity we allow any of %Lf, %llf, %qf for long double
10218 #if defined(USE_LONG_DOUBLE)
10222 /* [perl #20339] - we should accept and ignore %lf rather than die */
10226 #if defined(USE_LONG_DOUBLE)
10227 intsize = args ? 0 : 'q';
10231 #if defined(HAS_LONG_DOUBLE)
10240 /* now we need (long double) if intsize == 'q', else (double) */
10242 #if LONG_DOUBLESIZE > DOUBLESIZE
10244 va_arg(*args, long double) :
10245 va_arg(*args, double)
10247 va_arg(*args, double)
10252 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10253 else. frexp() has some unspecified behaviour for those three */
10254 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10256 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10257 will cast our (long double) to (double) */
10258 (void)Perl_frexp(nv, &i);
10259 if (i == PERL_INT_MIN)
10260 Perl_die(aTHX_ "panic: frexp");
10262 need = BIT_DIGITS(i);
10264 need += has_precis ? precis : 6; /* known default */
10269 #ifdef HAS_LDBL_SPRINTF_BUG
10270 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10271 with sfio - Allen <allens@cpan.org> */
10274 # define MY_DBL_MAX DBL_MAX
10275 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10276 # if DOUBLESIZE >= 8
10277 # define MY_DBL_MAX 1.7976931348623157E+308L
10279 # define MY_DBL_MAX 3.40282347E+38L
10283 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10284 # define MY_DBL_MAX_BUG 1L
10286 # define MY_DBL_MAX_BUG MY_DBL_MAX
10290 # define MY_DBL_MIN DBL_MIN
10291 # else /* XXX guessing! -Allen */
10292 # if DOUBLESIZE >= 8
10293 # define MY_DBL_MIN 2.2250738585072014E-308L
10295 # define MY_DBL_MIN 1.17549435E-38L
10299 if ((intsize == 'q') && (c == 'f') &&
10300 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10301 (need < DBL_DIG)) {
10302 /* it's going to be short enough that
10303 * long double precision is not needed */
10305 if ((nv <= 0L) && (nv >= -0L))
10306 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10308 /* would use Perl_fp_class as a double-check but not
10309 * functional on IRIX - see perl.h comments */
10311 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10312 /* It's within the range that a double can represent */
10313 #if defined(DBL_MAX) && !defined(DBL_MIN)
10314 if ((nv >= ((long double)1/DBL_MAX)) ||
10315 (nv <= (-(long double)1/DBL_MAX)))
10317 fix_ldbl_sprintf_bug = TRUE;
10320 if (fix_ldbl_sprintf_bug == TRUE) {
10330 # undef MY_DBL_MAX_BUG
10333 #endif /* HAS_LDBL_SPRINTF_BUG */
10335 need += 20; /* fudge factor */
10336 if (PL_efloatsize < need) {
10337 Safefree(PL_efloatbuf);
10338 PL_efloatsize = need + 20; /* more fudge */
10339 Newx(PL_efloatbuf, PL_efloatsize, char);
10340 PL_efloatbuf[0] = '\0';
10343 if ( !(width || left || plus || alt) && fill != '0'
10344 && has_precis && intsize != 'q' ) { /* Shortcuts */
10345 /* See earlier comment about buggy Gconvert when digits,
10347 if ( c == 'g' && precis) {
10348 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10349 /* May return an empty string for digits==0 */
10350 if (*PL_efloatbuf) {
10351 elen = strlen(PL_efloatbuf);
10352 goto float_converted;
10354 } else if ( c == 'f' && !precis) {
10355 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10360 char *ptr = ebuf + sizeof ebuf;
10363 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10364 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10365 if (intsize == 'q') {
10366 /* Copy the one or more characters in a long double
10367 * format before the 'base' ([efgEFG]) character to
10368 * the format string. */
10369 static char const prifldbl[] = PERL_PRIfldbl;
10370 char const *p = prifldbl + sizeof(prifldbl) - 3;
10371 while (p >= prifldbl) { *--ptr = *p--; }
10376 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10381 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10393 /* No taint. Otherwise we are in the strange situation
10394 * where printf() taints but print($float) doesn't.
10396 #if defined(HAS_LONG_DOUBLE)
10397 elen = ((intsize == 'q')
10398 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10399 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10401 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10405 eptr = PL_efloatbuf;
10413 i = SvCUR(sv) - origlen;
10416 case 'h': *(va_arg(*args, short*)) = i; break;
10417 default: *(va_arg(*args, int*)) = i; break;
10418 case 'l': *(va_arg(*args, long*)) = i; break;
10419 case 'V': *(va_arg(*args, IV*)) = i; break;
10422 *(va_arg(*args, Quad_t*)) = i; break;
10429 sv_setuv_mg(argsv, (UV)i);
10430 continue; /* not "break" */
10437 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10438 && ckWARN(WARN_PRINTF))
10440 SV * const msg = sv_newmortal();
10441 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10442 (PL_op->op_type == OP_PRTF) ? "" : "s");
10443 if (fmtstart < patend) {
10444 const char * const fmtend = q < patend ? q : patend;
10446 sv_catpvs(msg, "\"%");
10447 for (f = fmtstart; f < fmtend; f++) {
10449 sv_catpvn(msg, f, 1);
10451 Perl_sv_catpvf(aTHX_ msg,
10452 "\\%03"UVof, (UV)*f & 0xFF);
10455 sv_catpvs(msg, "\"");
10457 sv_catpvs(msg, "end of string");
10459 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10462 /* output mangled stuff ... */
10468 /* ... right here, because formatting flags should not apply */
10469 SvGROW(sv, SvCUR(sv) + elen + 1);
10471 Copy(eptr, p, elen, char);
10474 SvCUR_set(sv, p - SvPVX_const(sv));
10476 continue; /* not "break" */
10479 if (is_utf8 != has_utf8) {
10482 sv_utf8_upgrade(sv);
10485 const STRLEN old_elen = elen;
10486 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10487 sv_utf8_upgrade(nsv);
10488 eptr = SvPVX_const(nsv);
10491 if (width) { /* fudge width (can't fudge elen) */
10492 width += elen - old_elen;
10498 have = esignlen + zeros + elen;
10500 Perl_croak_nocontext("%s", PL_memory_wrap);
10502 need = (have > width ? have : width);
10505 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10506 Perl_croak_nocontext("%s", PL_memory_wrap);
10507 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10509 if (esignlen && fill == '0') {
10511 for (i = 0; i < (int)esignlen; i++)
10512 *p++ = esignbuf[i];
10514 if (gap && !left) {
10515 memset(p, fill, gap);
10518 if (esignlen && fill != '0') {
10520 for (i = 0; i < (int)esignlen; i++)
10521 *p++ = esignbuf[i];
10525 for (i = zeros; i; i--)
10529 Copy(eptr, p, elen, char);
10533 memset(p, ' ', gap);
10538 Copy(dotstr, p, dotstrlen, char);
10542 vectorize = FALSE; /* done iterating over vecstr */
10549 SvCUR_set(sv, p - SvPVX_const(sv));
10558 /* =========================================================================
10560 =head1 Cloning an interpreter
10562 All the macros and functions in this section are for the private use of
10563 the main function, perl_clone().
10565 The foo_dup() functions make an exact copy of an existing foo thingy.
10566 During the course of a cloning, a hash table is used to map old addresses
10567 to new addresses. The table is created and manipulated with the
10568 ptr_table_* functions.
10572 * =========================================================================*/
10575 #if defined(USE_ITHREADS)
10577 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
10578 #ifndef GpREFCNT_inc
10579 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
10583 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
10584 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
10585 If this changes, please unmerge ss_dup.
10586 Likewise, sv_dup_inc_multiple() relies on this fact. */
10587 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
10588 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
10589 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
10590 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
10591 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
10592 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
10593 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
10594 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
10595 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
10596 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
10597 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
10598 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
10599 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
10601 /* clone a parser */
10604 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
10608 PERL_ARGS_ASSERT_PARSER_DUP;
10613 /* look for it in the table first */
10614 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
10618 /* create anew and remember what it is */
10619 Newxz(parser, 1, yy_parser);
10620 ptr_table_store(PL_ptr_table, proto, parser);
10622 parser->yyerrstatus = 0;
10623 parser->yychar = YYEMPTY; /* Cause a token to be read. */
10625 /* XXX these not yet duped */
10626 parser->old_parser = NULL;
10627 parser->stack = NULL;
10629 parser->stack_size = 0;
10630 /* XXX parser->stack->state = 0; */
10632 /* XXX eventually, just Copy() most of the parser struct ? */
10634 parser->lex_brackets = proto->lex_brackets;
10635 parser->lex_casemods = proto->lex_casemods;
10636 parser->lex_brackstack = savepvn(proto->lex_brackstack,
10637 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
10638 parser->lex_casestack = savepvn(proto->lex_casestack,
10639 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
10640 parser->lex_defer = proto->lex_defer;
10641 parser->lex_dojoin = proto->lex_dojoin;
10642 parser->lex_expect = proto->lex_expect;
10643 parser->lex_formbrack = proto->lex_formbrack;
10644 parser->lex_inpat = proto->lex_inpat;
10645 parser->lex_inwhat = proto->lex_inwhat;
10646 parser->lex_op = proto->lex_op;
10647 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
10648 parser->lex_starts = proto->lex_starts;
10649 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
10650 parser->multi_close = proto->multi_close;
10651 parser->multi_open = proto->multi_open;
10652 parser->multi_start = proto->multi_start;
10653 parser->multi_end = proto->multi_end;
10654 parser->pending_ident = proto->pending_ident;
10655 parser->preambled = proto->preambled;
10656 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
10657 parser->linestr = sv_dup_inc(proto->linestr, param);
10658 parser->expect = proto->expect;
10659 parser->copline = proto->copline;
10660 parser->last_lop_op = proto->last_lop_op;
10661 parser->lex_state = proto->lex_state;
10662 parser->rsfp = fp_dup(proto->rsfp, '<', param);
10663 /* rsfp_filters entries have fake IoDIRP() */
10664 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
10665 parser->in_my = proto->in_my;
10666 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
10667 parser->error_count = proto->error_count;
10670 parser->linestr = sv_dup_inc(proto->linestr, param);
10673 char * const ols = SvPVX(proto->linestr);
10674 char * const ls = SvPVX(parser->linestr);
10676 parser->bufptr = ls + (proto->bufptr >= ols ?
10677 proto->bufptr - ols : 0);
10678 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
10679 proto->oldbufptr - ols : 0);
10680 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
10681 proto->oldoldbufptr - ols : 0);
10682 parser->linestart = ls + (proto->linestart >= ols ?
10683 proto->linestart - ols : 0);
10684 parser->last_uni = ls + (proto->last_uni >= ols ?
10685 proto->last_uni - ols : 0);
10686 parser->last_lop = ls + (proto->last_lop >= ols ?
10687 proto->last_lop - ols : 0);
10689 parser->bufend = ls + SvCUR(parser->linestr);
10692 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
10696 parser->endwhite = proto->endwhite;
10697 parser->faketokens = proto->faketokens;
10698 parser->lasttoke = proto->lasttoke;
10699 parser->nextwhite = proto->nextwhite;
10700 parser->realtokenstart = proto->realtokenstart;
10701 parser->skipwhite = proto->skipwhite;
10702 parser->thisclose = proto->thisclose;
10703 parser->thismad = proto->thismad;
10704 parser->thisopen = proto->thisopen;
10705 parser->thisstuff = proto->thisstuff;
10706 parser->thistoken = proto->thistoken;
10707 parser->thiswhite = proto->thiswhite;
10709 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
10710 parser->curforce = proto->curforce;
10712 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
10713 Copy(proto->nexttype, parser->nexttype, 5, I32);
10714 parser->nexttoke = proto->nexttoke;
10717 /* XXX should clone saved_curcop here, but we aren't passed
10718 * proto_perl; so do it in perl_clone_using instead */
10724 /* duplicate a file handle */
10727 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
10731 PERL_ARGS_ASSERT_FP_DUP;
10732 PERL_UNUSED_ARG(type);
10735 return (PerlIO*)NULL;
10737 /* look for it in the table first */
10738 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
10742 /* create anew and remember what it is */
10743 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
10744 ptr_table_store(PL_ptr_table, fp, ret);
10748 /* duplicate a directory handle */
10751 Perl_dirp_dup(pTHX_ DIR *const dp)
10753 PERL_UNUSED_CONTEXT;
10760 /* duplicate a typeglob */
10763 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
10767 PERL_ARGS_ASSERT_GP_DUP;
10771 /* look for it in the table first */
10772 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
10776 /* create anew and remember what it is */
10778 ptr_table_store(PL_ptr_table, gp, ret);
10781 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
10782 on Newxz() to do this for us. */
10783 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
10784 ret->gp_io = io_dup_inc(gp->gp_io, param);
10785 ret->gp_form = cv_dup_inc(gp->gp_form, param);
10786 ret->gp_av = av_dup_inc(gp->gp_av, param);
10787 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
10788 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
10789 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
10790 ret->gp_cvgen = gp->gp_cvgen;
10791 ret->gp_line = gp->gp_line;
10792 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
10796 /* duplicate a chain of magic */
10799 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
10801 MAGIC *mgret = NULL;
10802 MAGIC **mgprev_p = &mgret;
10804 PERL_ARGS_ASSERT_MG_DUP;
10806 for (; mg; mg = mg->mg_moremagic) {
10809 if ((param->flags & CLONEf_JOIN_IN)
10810 && mg->mg_type == PERL_MAGIC_backref)
10811 /* when joining, we let the individual SVs add themselves to
10812 * backref as needed. */
10815 Newx(nmg, 1, MAGIC);
10817 mgprev_p = &(nmg->mg_moremagic);
10819 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
10820 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
10821 from the original commit adding Perl_mg_dup() - revision 4538.
10822 Similarly there is the annotation "XXX random ptr?" next to the
10823 assignment to nmg->mg_ptr. */
10826 /* FIXME for plugins
10827 if (nmg->mg_type == PERL_MAGIC_qr) {
10828 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
10832 if(nmg->mg_type == PERL_MAGIC_backref) {
10833 /* The backref AV has its reference count deliberately bumped by
10836 = SvREFCNT_inc(av_dup_inc((const AV *) nmg->mg_obj, param));
10839 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
10840 ? sv_dup_inc(nmg->mg_obj, param)
10841 : sv_dup(nmg->mg_obj, param);
10844 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
10845 if (nmg->mg_len > 0) {
10846 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
10847 if (nmg->mg_type == PERL_MAGIC_overload_table &&
10848 AMT_AMAGIC((AMT*)nmg->mg_ptr))
10850 AMT * const namtp = (AMT*)nmg->mg_ptr;
10851 sv_dup_inc_multiple((SV**)(namtp->table),
10852 (SV**)(namtp->table), NofAMmeth, param);
10855 else if (nmg->mg_len == HEf_SVKEY)
10856 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
10858 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
10859 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
10865 #endif /* USE_ITHREADS */
10867 struct ptr_tbl_arena {
10868 struct ptr_tbl_arena *next;
10869 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
10872 /* create a new pointer-mapping table */
10875 Perl_ptr_table_new(pTHX)
10878 PERL_UNUSED_CONTEXT;
10880 Newx(tbl, 1, PTR_TBL_t);
10881 tbl->tbl_max = 511;
10882 tbl->tbl_items = 0;
10883 tbl->tbl_arena = NULL;
10884 tbl->tbl_arena_next = NULL;
10885 tbl->tbl_arena_end = NULL;
10886 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
10890 #define PTR_TABLE_HASH(ptr) \
10891 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
10893 /* map an existing pointer using a table */
10895 STATIC PTR_TBL_ENT_t *
10896 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
10898 PTR_TBL_ENT_t *tblent;
10899 const UV hash = PTR_TABLE_HASH(sv);
10901 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
10903 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
10904 for (; tblent; tblent = tblent->next) {
10905 if (tblent->oldval == sv)
10912 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
10914 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
10916 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
10917 PERL_UNUSED_CONTEXT;
10919 return tblent ? tblent->newval : NULL;
10922 /* add a new entry to a pointer-mapping table */
10925 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
10927 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
10929 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
10930 PERL_UNUSED_CONTEXT;
10933 tblent->newval = newsv;
10935 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
10937 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
10938 struct ptr_tbl_arena *new_arena;
10940 Newx(new_arena, 1, struct ptr_tbl_arena);
10941 new_arena->next = tbl->tbl_arena;
10942 tbl->tbl_arena = new_arena;
10943 tbl->tbl_arena_next = new_arena->array;
10944 tbl->tbl_arena_end = new_arena->array
10945 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
10948 tblent = tbl->tbl_arena_next++;
10950 tblent->oldval = oldsv;
10951 tblent->newval = newsv;
10952 tblent->next = tbl->tbl_ary[entry];
10953 tbl->tbl_ary[entry] = tblent;
10955 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
10956 ptr_table_split(tbl);
10960 /* double the hash bucket size of an existing ptr table */
10963 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
10965 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
10966 const UV oldsize = tbl->tbl_max + 1;
10967 UV newsize = oldsize * 2;
10970 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
10971 PERL_UNUSED_CONTEXT;
10973 Renew(ary, newsize, PTR_TBL_ENT_t*);
10974 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
10975 tbl->tbl_max = --newsize;
10976 tbl->tbl_ary = ary;
10977 for (i=0; i < oldsize; i++, ary++) {
10978 PTR_TBL_ENT_t **entp = ary;
10979 PTR_TBL_ENT_t *ent = *ary;
10980 PTR_TBL_ENT_t **curentp;
10983 curentp = ary + oldsize;
10985 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
10987 ent->next = *curentp;
10997 /* remove all the entries from a ptr table */
10998 /* Deprecated - will be removed post 5.14 */
11001 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11003 if (tbl && tbl->tbl_items) {
11004 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11006 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11009 struct ptr_tbl_arena *next = arena->next;
11015 tbl->tbl_items = 0;
11016 tbl->tbl_arena = NULL;
11017 tbl->tbl_arena_next = NULL;
11018 tbl->tbl_arena_end = NULL;
11022 /* clear and free a ptr table */
11025 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11027 struct ptr_tbl_arena *arena;
11033 arena = tbl->tbl_arena;
11036 struct ptr_tbl_arena *next = arena->next;
11042 Safefree(tbl->tbl_ary);
11046 #if defined(USE_ITHREADS)
11049 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11051 PERL_ARGS_ASSERT_RVPV_DUP;
11054 if (SvWEAKREF(sstr)) {
11055 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11056 if (param->flags & CLONEf_JOIN_IN) {
11057 /* if joining, we add any back references individually rather
11058 * than copying the whole backref array */
11059 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11063 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11065 else if (SvPVX_const(sstr)) {
11066 /* Has something there */
11068 /* Normal PV - clone whole allocated space */
11069 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11070 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11071 /* Not that normal - actually sstr is copy on write.
11072 But we are a true, independant SV, so: */
11073 SvREADONLY_off(dstr);
11078 /* Special case - not normally malloced for some reason */
11079 if (isGV_with_GP(sstr)) {
11080 /* Don't need to do anything here. */
11082 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11083 /* A "shared" PV - clone it as "shared" PV */
11085 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11089 /* Some other special case - random pointer */
11090 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11095 /* Copy the NULL */
11096 SvPV_set(dstr, NULL);
11100 /* duplicate a list of SVs. source and dest may point to the same memory. */
11102 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11103 SSize_t items, CLONE_PARAMS *const param)
11105 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11107 while (items-- > 0) {
11108 *dest++ = sv_dup_inc(*source++, param);
11114 /* duplicate an SV of any type (including AV, HV etc) */
11117 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11122 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11124 if (SvTYPE(sstr) == SVTYPEMASK) {
11125 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11130 /* look for it in the table first */
11131 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11135 if(param->flags & CLONEf_JOIN_IN) {
11136 /** We are joining here so we don't want do clone
11137 something that is bad **/
11138 if (SvTYPE(sstr) == SVt_PVHV) {
11139 const HEK * const hvname = HvNAME_HEK(sstr);
11141 /** don't clone stashes if they already exist **/
11142 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
11143 ptr_table_store(PL_ptr_table, sstr, dstr);
11149 /* create anew and remember what it is */
11152 #ifdef DEBUG_LEAKING_SCALARS
11153 dstr->sv_debug_optype = sstr->sv_debug_optype;
11154 dstr->sv_debug_line = sstr->sv_debug_line;
11155 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11156 dstr->sv_debug_cloned = 1;
11157 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11160 ptr_table_store(PL_ptr_table, sstr, dstr);
11163 SvFLAGS(dstr) = SvFLAGS(sstr);
11164 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11165 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11168 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11169 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11170 (void*)PL_watch_pvx, SvPVX_const(sstr));
11173 /* don't clone objects whose class has asked us not to */
11174 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11179 switch (SvTYPE(sstr)) {
11181 SvANY(dstr) = NULL;
11184 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11186 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11188 SvIV_set(dstr, SvIVX(sstr));
11192 SvANY(dstr) = new_XNV();
11193 SvNV_set(dstr, SvNVX(sstr));
11195 /* case SVt_BIND: */
11198 /* These are all the types that need complex bodies allocating. */
11200 const svtype sv_type = SvTYPE(sstr);
11201 const struct body_details *const sv_type_details
11202 = bodies_by_type + sv_type;
11206 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11221 assert(sv_type_details->body_size);
11222 if (sv_type_details->arena) {
11223 new_body_inline(new_body, sv_type);
11225 = (void*)((char*)new_body - sv_type_details->offset);
11227 new_body = new_NOARENA(sv_type_details);
11231 SvANY(dstr) = new_body;
11234 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11235 ((char*)SvANY(dstr)) + sv_type_details->offset,
11236 sv_type_details->copy, char);
11238 Copy(((char*)SvANY(sstr)),
11239 ((char*)SvANY(dstr)),
11240 sv_type_details->body_size + sv_type_details->offset, char);
11243 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11244 && !isGV_with_GP(dstr)
11245 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11246 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11248 /* The Copy above means that all the source (unduplicated) pointers
11249 are now in the destination. We can check the flags and the
11250 pointers in either, but it's possible that there's less cache
11251 missing by always going for the destination.
11252 FIXME - instrument and check that assumption */
11253 if (sv_type >= SVt_PVMG) {
11254 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11255 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11256 } else if (SvMAGIC(dstr))
11257 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11259 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11262 /* The cast silences a GCC warning about unhandled types. */
11263 switch ((int)sv_type) {
11273 /* FIXME for plugins */
11274 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11277 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11278 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11279 LvTARG(dstr) = dstr;
11280 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11281 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11283 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11285 if(isGV_with_GP(sstr)) {
11286 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11287 /* Don't call sv_add_backref here as it's going to be
11288 created as part of the magic cloning of the symbol
11289 table--unless this is during a join and the stash
11290 is not actually being cloned. */
11291 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11292 at the point of this comment. */
11293 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11294 if (param->flags & CLONEf_JOIN_IN)
11295 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
11296 GvGP(dstr) = gp_dup(GvGP(sstr), param);
11297 (void)GpREFCNT_inc(GvGP(dstr));
11299 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11302 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11303 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11304 /* I have no idea why fake dirp (rsfps)
11305 should be treated differently but otherwise
11306 we end up with leaks -- sky*/
11307 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11308 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11309 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11311 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11312 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11313 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11314 if (IoDIRP(dstr)) {
11315 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
11318 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11320 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
11322 if (IoOFP(dstr) == IoIFP(sstr))
11323 IoOFP(dstr) = IoIFP(dstr);
11325 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11326 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11327 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11328 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11331 /* avoid cloning an empty array */
11332 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11333 SV **dst_ary, **src_ary;
11334 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11336 src_ary = AvARRAY((const AV *)sstr);
11337 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11338 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11339 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11340 AvALLOC((const AV *)dstr) = dst_ary;
11341 if (AvREAL((const AV *)sstr)) {
11342 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11346 while (items-- > 0)
11347 *dst_ary++ = sv_dup(*src_ary++, param);
11349 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11350 while (items-- > 0) {
11351 *dst_ary++ = &PL_sv_undef;
11355 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11356 AvALLOC((const AV *)dstr) = (SV**)NULL;
11357 AvMAX( (const AV *)dstr) = -1;
11358 AvFILLp((const AV *)dstr) = -1;
11362 if (HvARRAY((const HV *)sstr)) {
11364 const bool sharekeys = !!HvSHAREKEYS(sstr);
11365 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11366 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11368 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11369 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11371 HvARRAY(dstr) = (HE**)darray;
11372 while (i <= sxhv->xhv_max) {
11373 const HE * const source = HvARRAY(sstr)[i];
11374 HvARRAY(dstr)[i] = source
11375 ? he_dup(source, sharekeys, param) : 0;
11380 const struct xpvhv_aux * const saux = HvAUX(sstr);
11381 struct xpvhv_aux * const daux = HvAUX(dstr);
11382 /* This flag isn't copied. */
11383 /* SvOOK_on(hv) attacks the IV flags. */
11384 SvFLAGS(dstr) |= SVf_OOK;
11386 hvname = saux->xhv_name;
11387 daux->xhv_name = hek_dup(hvname, param);
11389 daux->xhv_riter = saux->xhv_riter;
11390 daux->xhv_eiter = saux->xhv_eiter
11391 ? he_dup(saux->xhv_eiter,
11392 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
11393 /* backref array needs refcnt=2; see sv_add_backref */
11394 daux->xhv_backreferences =
11395 (param->flags & CLONEf_JOIN_IN)
11396 /* when joining, we let the individual GVs and
11397 * CVs add themselves to backref as
11398 * needed. This avoids pulling in stuff
11399 * that isn't required, and simplifies the
11400 * case where stashes aren't cloned back
11401 * if they already exist in the parent
11404 : saux->xhv_backreferences
11405 ? MUTABLE_AV(SvREFCNT_inc(
11406 sv_dup_inc((const SV *)saux->xhv_backreferences, param)))
11409 daux->xhv_mro_meta = saux->xhv_mro_meta
11410 ? mro_meta_dup(saux->xhv_mro_meta, param)
11413 /* Record stashes for possible cloning in Perl_clone(). */
11415 av_push(param->stashes, dstr);
11419 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11422 if (!(param->flags & CLONEf_COPY_STACKS)) {
11427 /* NOTE: not refcounted */
11428 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
11429 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
11430 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
11432 if (!CvISXSUB(dstr))
11433 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
11435 if (CvCONST(dstr) && CvISXSUB(dstr)) {
11436 CvXSUBANY(dstr).any_ptr =
11437 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
11439 /* don't dup if copying back - CvGV isn't refcounted, so the
11440 * duped GV may never be freed. A bit of a hack! DAPM */
11441 SvANY(MUTABLE_CV(dstr))->xcv_gv =
11443 ? gv_dup_inc(CvGV(sstr), param)
11444 : (param->flags & CLONEf_JOIN_IN)
11446 : gv_dup(CvGV(sstr), param);
11448 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
11450 CvWEAKOUTSIDE(sstr)
11451 ? cv_dup( CvOUTSIDE(dstr), param)
11452 : cv_dup_inc(CvOUTSIDE(dstr), param);
11453 if (!CvISXSUB(dstr))
11454 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
11460 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
11467 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11469 PERL_ARGS_ASSERT_SV_DUP_INC;
11470 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
11474 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11476 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
11477 PERL_ARGS_ASSERT_SV_DUP;
11479 /* Track every SV that (at least initially) had a reference count of 0.
11480 We need to do this by holding an actual reference to it in this array.
11481 If we attempt to cheat, turn AvREAL_off(), and store only pointers
11482 (akin to the stashes hash, and the perl stack), we come unstuck if
11483 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
11484 thread) is manipulated in a CLONE method, because CLONE runs before the
11485 unreferenced array is walked to find SVs still with SvREFCNT() == 0
11486 (and fix things up by giving each a reference via the temps stack).
11487 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
11488 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
11489 before the walk of unreferenced happens and a reference to that is SV
11490 added to the temps stack. At which point we have the same SV considered
11491 to be in use, and free to be re-used. Not good.
11493 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
11494 assert(param->unreferenced);
11495 av_push(param->unreferenced, SvREFCNT_inc(dstr));
11501 /* duplicate a context */
11504 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
11506 PERL_CONTEXT *ncxs;
11508 PERL_ARGS_ASSERT_CX_DUP;
11511 return (PERL_CONTEXT*)NULL;
11513 /* look for it in the table first */
11514 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
11518 /* create anew and remember what it is */
11519 Newx(ncxs, max + 1, PERL_CONTEXT);
11520 ptr_table_store(PL_ptr_table, cxs, ncxs);
11521 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
11524 PERL_CONTEXT * const ncx = &ncxs[ix];
11525 if (CxTYPE(ncx) == CXt_SUBST) {
11526 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
11529 switch (CxTYPE(ncx)) {
11531 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
11532 ? cv_dup_inc(ncx->blk_sub.cv, param)
11533 : cv_dup(ncx->blk_sub.cv,param));
11534 ncx->blk_sub.argarray = (CxHASARGS(ncx)
11535 ? av_dup_inc(ncx->blk_sub.argarray,
11538 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
11540 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
11541 ncx->blk_sub.oldcomppad);
11544 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
11546 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
11548 case CXt_LOOP_LAZYSV:
11549 ncx->blk_loop.state_u.lazysv.end
11550 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
11551 /* We are taking advantage of av_dup_inc and sv_dup_inc
11552 actually being the same function, and order equivalance of
11554 We can assert the later [but only at run time :-(] */
11555 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
11556 (void *) &ncx->blk_loop.state_u.lazysv.cur);
11558 ncx->blk_loop.state_u.ary.ary
11559 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
11560 case CXt_LOOP_LAZYIV:
11561 case CXt_LOOP_PLAIN:
11562 if (CxPADLOOP(ncx)) {
11563 ncx->blk_loop.oldcomppad
11564 = (PAD*)ptr_table_fetch(PL_ptr_table,
11565 ncx->blk_loop.oldcomppad);
11567 ncx->blk_loop.oldcomppad
11568 = (PAD*)gv_dup((const GV *)ncx->blk_loop.oldcomppad,
11573 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
11574 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
11575 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
11588 /* duplicate a stack info structure */
11591 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
11595 PERL_ARGS_ASSERT_SI_DUP;
11598 return (PERL_SI*)NULL;
11600 /* look for it in the table first */
11601 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
11605 /* create anew and remember what it is */
11606 Newxz(nsi, 1, PERL_SI);
11607 ptr_table_store(PL_ptr_table, si, nsi);
11609 nsi->si_stack = av_dup_inc(si->si_stack, param);
11610 nsi->si_cxix = si->si_cxix;
11611 nsi->si_cxmax = si->si_cxmax;
11612 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
11613 nsi->si_type = si->si_type;
11614 nsi->si_prev = si_dup(si->si_prev, param);
11615 nsi->si_next = si_dup(si->si_next, param);
11616 nsi->si_markoff = si->si_markoff;
11621 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
11622 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
11623 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
11624 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
11625 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
11626 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
11627 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
11628 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
11629 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
11630 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
11631 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
11632 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
11633 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
11634 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
11635 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
11636 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
11639 #define pv_dup_inc(p) SAVEPV(p)
11640 #define pv_dup(p) SAVEPV(p)
11641 #define svp_dup_inc(p,pp) any_dup(p,pp)
11643 /* map any object to the new equivent - either something in the
11644 * ptr table, or something in the interpreter structure
11648 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
11652 PERL_ARGS_ASSERT_ANY_DUP;
11655 return (void*)NULL;
11657 /* look for it in the table first */
11658 ret = ptr_table_fetch(PL_ptr_table, v);
11662 /* see if it is part of the interpreter structure */
11663 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
11664 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
11672 /* duplicate the save stack */
11675 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
11678 ANY * const ss = proto_perl->Isavestack;
11679 const I32 max = proto_perl->Isavestack_max;
11680 I32 ix = proto_perl->Isavestack_ix;
11693 void (*dptr) (void*);
11694 void (*dxptr) (pTHX_ void*);
11696 PERL_ARGS_ASSERT_SS_DUP;
11698 Newxz(nss, max, ANY);
11701 const UV uv = POPUV(ss,ix);
11702 const U8 type = (U8)uv & SAVE_MASK;
11704 TOPUV(nss,ix) = uv;
11706 case SAVEt_CLEARSV:
11708 case SAVEt_HELEM: /* hash element */
11709 sv = (const SV *)POPPTR(ss,ix);
11710 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11712 case SAVEt_ITEM: /* normal string */
11713 case SAVEt_SV: /* scalar reference */
11714 sv = (const SV *)POPPTR(ss,ix);
11715 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11718 case SAVEt_MORTALIZESV:
11719 sv = (const SV *)POPPTR(ss,ix);
11720 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11722 case SAVEt_SHARED_PVREF: /* char* in shared space */
11723 c = (char*)POPPTR(ss,ix);
11724 TOPPTR(nss,ix) = savesharedpv(c);
11725 ptr = POPPTR(ss,ix);
11726 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11728 case SAVEt_GENERIC_SVREF: /* generic sv */
11729 case SAVEt_SVREF: /* scalar reference */
11730 sv = (const SV *)POPPTR(ss,ix);
11731 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11732 ptr = POPPTR(ss,ix);
11733 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
11735 case SAVEt_HV: /* hash reference */
11736 case SAVEt_AV: /* array reference */
11737 sv = (const SV *) POPPTR(ss,ix);
11738 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11740 case SAVEt_COMPPAD:
11742 sv = (const SV *) POPPTR(ss,ix);
11743 TOPPTR(nss,ix) = sv_dup(sv, param);
11745 case SAVEt_INT: /* int reference */
11746 ptr = POPPTR(ss,ix);
11747 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11748 intval = (int)POPINT(ss,ix);
11749 TOPINT(nss,ix) = intval;
11751 case SAVEt_LONG: /* long reference */
11752 ptr = POPPTR(ss,ix);
11753 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11754 longval = (long)POPLONG(ss,ix);
11755 TOPLONG(nss,ix) = longval;
11757 case SAVEt_I32: /* I32 reference */
11758 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
11759 ptr = POPPTR(ss,ix);
11760 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11762 TOPINT(nss,ix) = i;
11764 case SAVEt_IV: /* IV reference */
11765 ptr = POPPTR(ss,ix);
11766 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11768 TOPIV(nss,ix) = iv;
11770 case SAVEt_HPTR: /* HV* reference */
11771 case SAVEt_APTR: /* AV* reference */
11772 case SAVEt_SPTR: /* SV* reference */
11773 ptr = POPPTR(ss,ix);
11774 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11775 sv = (const SV *)POPPTR(ss,ix);
11776 TOPPTR(nss,ix) = sv_dup(sv, param);
11778 case SAVEt_VPTR: /* random* reference */
11779 ptr = POPPTR(ss,ix);
11780 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11782 case SAVEt_INT_SMALL:
11783 case SAVEt_I32_SMALL:
11784 case SAVEt_I16: /* I16 reference */
11785 case SAVEt_I8: /* I8 reference */
11787 ptr = POPPTR(ss,ix);
11788 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11790 case SAVEt_GENERIC_PVREF: /* generic char* */
11791 case SAVEt_PPTR: /* char* reference */
11792 ptr = POPPTR(ss,ix);
11793 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11794 c = (char*)POPPTR(ss,ix);
11795 TOPPTR(nss,ix) = pv_dup(c);
11797 case SAVEt_GP: /* scalar reference */
11798 gv = (const GV *)POPPTR(ss,ix);
11799 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
11800 gp = (GP*)POPPTR(ss,ix);
11801 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
11802 (void)GpREFCNT_inc(gp);
11804 TOPINT(nss,ix) = i;
11807 ptr = POPPTR(ss,ix);
11808 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
11809 /* these are assumed to be refcounted properly */
11811 switch (((OP*)ptr)->op_type) {
11813 case OP_LEAVESUBLV:
11817 case OP_LEAVEWRITE:
11818 TOPPTR(nss,ix) = ptr;
11821 (void) OpREFCNT_inc(o);
11825 TOPPTR(nss,ix) = NULL;
11830 TOPPTR(nss,ix) = NULL;
11833 hv = (const HV *)POPPTR(ss,ix);
11834 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11836 TOPINT(nss,ix) = i;
11839 c = (char*)POPPTR(ss,ix);
11840 TOPPTR(nss,ix) = pv_dup_inc(c);
11842 case SAVEt_STACK_POS: /* Position on Perl stack */
11844 TOPINT(nss,ix) = i;
11846 case SAVEt_DESTRUCTOR:
11847 ptr = POPPTR(ss,ix);
11848 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11849 dptr = POPDPTR(ss,ix);
11850 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
11851 any_dup(FPTR2DPTR(void *, dptr),
11854 case SAVEt_DESTRUCTOR_X:
11855 ptr = POPPTR(ss,ix);
11856 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11857 dxptr = POPDXPTR(ss,ix);
11858 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
11859 any_dup(FPTR2DPTR(void *, dxptr),
11862 case SAVEt_REGCONTEXT:
11864 ix -= uv >> SAVE_TIGHT_SHIFT;
11866 case SAVEt_AELEM: /* array element */
11867 sv = (const SV *)POPPTR(ss,ix);
11868 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11870 TOPINT(nss,ix) = i;
11871 av = (const AV *)POPPTR(ss,ix);
11872 TOPPTR(nss,ix) = av_dup_inc(av, param);
11875 ptr = POPPTR(ss,ix);
11876 TOPPTR(nss,ix) = ptr;
11879 ptr = POPPTR(ss,ix);
11882 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
11883 HINTS_REFCNT_UNLOCK;
11885 TOPPTR(nss,ix) = ptr;
11887 TOPINT(nss,ix) = i;
11888 if (i & HINT_LOCALIZE_HH) {
11889 hv = (const HV *)POPPTR(ss,ix);
11890 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11893 case SAVEt_PADSV_AND_MORTALIZE:
11894 longval = (long)POPLONG(ss,ix);
11895 TOPLONG(nss,ix) = longval;
11896 ptr = POPPTR(ss,ix);
11897 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11898 sv = (const SV *)POPPTR(ss,ix);
11899 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11901 case SAVEt_SET_SVFLAGS:
11903 TOPINT(nss,ix) = i;
11905 TOPINT(nss,ix) = i;
11906 sv = (const SV *)POPPTR(ss,ix);
11907 TOPPTR(nss,ix) = sv_dup(sv, param);
11909 case SAVEt_RE_STATE:
11911 const struct re_save_state *const old_state
11912 = (struct re_save_state *)
11913 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11914 struct re_save_state *const new_state
11915 = (struct re_save_state *)
11916 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11918 Copy(old_state, new_state, 1, struct re_save_state);
11919 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
11921 new_state->re_state_bostr
11922 = pv_dup(old_state->re_state_bostr);
11923 new_state->re_state_reginput
11924 = pv_dup(old_state->re_state_reginput);
11925 new_state->re_state_regeol
11926 = pv_dup(old_state->re_state_regeol);
11927 new_state->re_state_regoffs
11928 = (regexp_paren_pair*)
11929 any_dup(old_state->re_state_regoffs, proto_perl);
11930 new_state->re_state_reglastparen
11931 = (U32*) any_dup(old_state->re_state_reglastparen,
11933 new_state->re_state_reglastcloseparen
11934 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
11936 /* XXX This just has to be broken. The old save_re_context
11937 code did SAVEGENERICPV(PL_reg_start_tmp);
11938 PL_reg_start_tmp is char **.
11939 Look above to what the dup code does for
11940 SAVEt_GENERIC_PVREF
11941 It can never have worked.
11942 So this is merely a faithful copy of the exiting bug: */
11943 new_state->re_state_reg_start_tmp
11944 = (char **) pv_dup((char *)
11945 old_state->re_state_reg_start_tmp);
11946 /* I assume that it only ever "worked" because no-one called
11947 (pseudo)fork while the regexp engine had re-entered itself.
11949 #ifdef PERL_OLD_COPY_ON_WRITE
11950 new_state->re_state_nrs
11951 = sv_dup(old_state->re_state_nrs, param);
11953 new_state->re_state_reg_magic
11954 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
11956 new_state->re_state_reg_oldcurpm
11957 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
11959 new_state->re_state_reg_curpm
11960 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
11962 new_state->re_state_reg_oldsaved
11963 = pv_dup(old_state->re_state_reg_oldsaved);
11964 new_state->re_state_reg_poscache
11965 = pv_dup(old_state->re_state_reg_poscache);
11966 new_state->re_state_reg_starttry
11967 = pv_dup(old_state->re_state_reg_starttry);
11970 case SAVEt_COMPILE_WARNINGS:
11971 ptr = POPPTR(ss,ix);
11972 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
11975 ptr = POPPTR(ss,ix);
11976 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
11980 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
11988 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
11989 * flag to the result. This is done for each stash before cloning starts,
11990 * so we know which stashes want their objects cloned */
11993 do_mark_cloneable_stash(pTHX_ SV *const sv)
11995 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
11997 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
11998 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
11999 if (cloner && GvCV(cloner)) {
12006 mXPUSHs(newSVhek(hvname));
12008 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12015 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12023 =for apidoc perl_clone
12025 Create and return a new interpreter by cloning the current one.
12027 perl_clone takes these flags as parameters:
12029 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12030 without it we only clone the data and zero the stacks,
12031 with it we copy the stacks and the new perl interpreter is
12032 ready to run at the exact same point as the previous one.
12033 The pseudo-fork code uses COPY_STACKS while the
12034 threads->create doesn't.
12036 CLONEf_KEEP_PTR_TABLE
12037 perl_clone keeps a ptr_table with the pointer of the old
12038 variable as a key and the new variable as a value,
12039 this allows it to check if something has been cloned and not
12040 clone it again but rather just use the value and increase the
12041 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12042 the ptr_table using the function
12043 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12044 reason to keep it around is if you want to dup some of your own
12045 variable who are outside the graph perl scans, example of this
12046 code is in threads.xs create
12049 This is a win32 thing, it is ignored on unix, it tells perls
12050 win32host code (which is c++) to clone itself, this is needed on
12051 win32 if you want to run two threads at the same time,
12052 if you just want to do some stuff in a separate perl interpreter
12053 and then throw it away and return to the original one,
12054 you don't need to do anything.
12059 /* XXX the above needs expanding by someone who actually understands it ! */
12060 EXTERN_C PerlInterpreter *
12061 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12064 perl_clone(PerlInterpreter *proto_perl, UV flags)
12067 #ifdef PERL_IMPLICIT_SYS
12069 PERL_ARGS_ASSERT_PERL_CLONE;
12071 /* perlhost.h so we need to call into it
12072 to clone the host, CPerlHost should have a c interface, sky */
12074 if (flags & CLONEf_CLONE_HOST) {
12075 return perl_clone_host(proto_perl,flags);
12077 return perl_clone_using(proto_perl, flags,
12079 proto_perl->IMemShared,
12080 proto_perl->IMemParse,
12082 proto_perl->IStdIO,
12086 proto_perl->IProc);
12090 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12091 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12092 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12093 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12094 struct IPerlDir* ipD, struct IPerlSock* ipS,
12095 struct IPerlProc* ipP)
12097 /* XXX many of the string copies here can be optimized if they're
12098 * constants; they need to be allocated as common memory and just
12099 * their pointers copied. */
12102 CLONE_PARAMS clone_params;
12103 CLONE_PARAMS* const param = &clone_params;
12105 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12107 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12108 #else /* !PERL_IMPLICIT_SYS */
12110 CLONE_PARAMS clone_params;
12111 CLONE_PARAMS* param = &clone_params;
12112 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12114 PERL_ARGS_ASSERT_PERL_CLONE;
12115 #endif /* PERL_IMPLICIT_SYS */
12117 /* for each stash, determine whether its objects should be cloned */
12118 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12119 PERL_SET_THX(my_perl);
12122 PoisonNew(my_perl, 1, PerlInterpreter);
12127 PL_scopestack_name = 0;
12129 PL_savestack_ix = 0;
12130 PL_savestack_max = -1;
12131 PL_sig_pending = 0;
12133 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12134 # ifdef DEBUG_LEAKING_SCALARS
12135 PL_sv_serial = (((U32)my_perl >> 2) & 0xfff) * 1000000;
12137 #else /* !DEBUGGING */
12138 Zero(my_perl, 1, PerlInterpreter);
12139 #endif /* DEBUGGING */
12141 #ifdef PERL_IMPLICIT_SYS
12142 /* host pointers */
12144 PL_MemShared = ipMS;
12145 PL_MemParse = ipMP;
12152 #endif /* PERL_IMPLICIT_SYS */
12154 param->flags = flags;
12155 /* Nothing in the core code uses this, but we make it available to
12156 extensions (using mg_dup). */
12157 param->proto_perl = proto_perl;
12158 /* Likely nothing will use this, but it is initialised to be consistent
12159 with Perl_clone_params_new(). */
12160 param->proto_perl = my_perl;
12161 param->unreferenced = NULL;
12163 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12165 PL_body_arenas = NULL;
12166 Zero(&PL_body_roots, 1, PL_body_roots);
12168 PL_nice_chunk = NULL;
12169 PL_nice_chunk_size = 0;
12171 PL_sv_objcount = 0;
12173 PL_sv_arenaroot = NULL;
12175 PL_debug = proto_perl->Idebug;
12177 PL_hash_seed = proto_perl->Ihash_seed;
12178 PL_rehash_seed = proto_perl->Irehash_seed;
12180 #ifdef USE_REENTRANT_API
12181 /* XXX: things like -Dm will segfault here in perlio, but doing
12182 * PERL_SET_CONTEXT(proto_perl);
12183 * breaks too many other things
12185 Perl_reentrant_init(aTHX);
12188 /* create SV map for pointer relocation */
12189 PL_ptr_table = ptr_table_new();
12191 /* initialize these special pointers as early as possible */
12192 SvANY(&PL_sv_undef) = NULL;
12193 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12194 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12195 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
12197 SvANY(&PL_sv_no) = new_XPVNV();
12198 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12199 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12200 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12201 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
12202 SvCUR_set(&PL_sv_no, 0);
12203 SvLEN_set(&PL_sv_no, 1);
12204 SvIV_set(&PL_sv_no, 0);
12205 SvNV_set(&PL_sv_no, 0);
12206 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
12208 SvANY(&PL_sv_yes) = new_XPVNV();
12209 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12210 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12211 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12212 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
12213 SvCUR_set(&PL_sv_yes, 1);
12214 SvLEN_set(&PL_sv_yes, 2);
12215 SvIV_set(&PL_sv_yes, 1);
12216 SvNV_set(&PL_sv_yes, 1);
12217 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
12219 /* dbargs array probably holds garbage */
12222 /* create (a non-shared!) shared string table */
12223 PL_strtab = newHV();
12224 HvSHAREKEYS_off(PL_strtab);
12225 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
12226 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
12228 PL_compiling = proto_perl->Icompiling;
12230 /* These two PVs will be free'd special way so must set them same way op.c does */
12231 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
12232 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
12234 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
12235 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
12237 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
12238 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
12239 if (PL_compiling.cop_hints_hash) {
12241 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
12242 HINTS_REFCNT_UNLOCK;
12244 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
12245 #ifdef PERL_DEBUG_READONLY_OPS
12250 /* pseudo environmental stuff */
12251 PL_origargc = proto_perl->Iorigargc;
12252 PL_origargv = proto_perl->Iorigargv;
12254 param->stashes = newAV(); /* Setup array of objects to call clone on */
12255 /* This makes no difference to the implementation, as it always pushes
12256 and shifts pointers to other SVs without changing their reference
12257 count, with the array becoming empty before it is freed. However, it
12258 makes it conceptually clear what is going on, and will avoid some
12259 work inside av.c, filling slots between AvFILL() and AvMAX() with
12260 &PL_sv_undef, and SvREFCNT_dec()ing those. */
12261 AvREAL_off(param->stashes);
12263 if (!(flags & CLONEf_COPY_STACKS)) {
12264 param->unreferenced = newAV();
12267 /* Set tainting stuff before PerlIO_debug can possibly get called */
12268 PL_tainting = proto_perl->Itainting;
12269 PL_taint_warn = proto_perl->Itaint_warn;
12271 #ifdef PERLIO_LAYERS
12272 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
12273 PerlIO_clone(aTHX_ proto_perl, param);
12276 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
12277 PL_incgv = gv_dup(proto_perl->Iincgv, param);
12278 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
12279 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
12280 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
12281 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
12284 PL_minus_c = proto_perl->Iminus_c;
12285 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
12286 PL_localpatches = proto_perl->Ilocalpatches;
12287 PL_splitstr = proto_perl->Isplitstr;
12288 PL_minus_n = proto_perl->Iminus_n;
12289 PL_minus_p = proto_perl->Iminus_p;
12290 PL_minus_l = proto_perl->Iminus_l;
12291 PL_minus_a = proto_perl->Iminus_a;
12292 PL_minus_E = proto_perl->Iminus_E;
12293 PL_minus_F = proto_perl->Iminus_F;
12294 PL_doswitches = proto_perl->Idoswitches;
12295 PL_dowarn = proto_perl->Idowarn;
12296 PL_doextract = proto_perl->Idoextract;
12297 PL_sawampersand = proto_perl->Isawampersand;
12298 PL_unsafe = proto_perl->Iunsafe;
12299 PL_inplace = SAVEPV(proto_perl->Iinplace);
12300 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
12301 PL_perldb = proto_perl->Iperldb;
12302 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12303 PL_exit_flags = proto_perl->Iexit_flags;
12305 /* magical thingies */
12306 /* XXX time(&PL_basetime) when asked for? */
12307 PL_basetime = proto_perl->Ibasetime;
12308 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
12310 PL_maxsysfd = proto_perl->Imaxsysfd;
12311 PL_statusvalue = proto_perl->Istatusvalue;
12313 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12315 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12317 PL_encoding = sv_dup(proto_perl->Iencoding, param);
12319 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
12320 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
12321 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
12324 /* RE engine related */
12325 Zero(&PL_reg_state, 1, struct re_save_state);
12326 PL_reginterp_cnt = 0;
12327 PL_regmatch_slab = NULL;
12329 /* Clone the regex array */
12330 /* ORANGE FIXME for plugins, probably in the SV dup code.
12331 newSViv(PTR2IV(CALLREGDUPE(
12332 INT2PTR(REGEXP *, SvIVX(regex)), param))))
12334 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
12335 PL_regex_pad = AvARRAY(PL_regex_padav);
12337 /* shortcuts to various I/O objects */
12338 PL_ofsgv = gv_dup(proto_perl->Iofsgv, param);
12339 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
12340 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
12341 PL_defgv = gv_dup(proto_perl->Idefgv, param);
12342 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
12343 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
12344 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
12346 /* shortcuts to regexp stuff */
12347 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
12349 /* shortcuts to misc objects */
12350 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
12352 /* shortcuts to debugging objects */
12353 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
12354 PL_DBline = gv_dup(proto_perl->IDBline, param);
12355 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
12356 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
12357 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
12358 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
12360 /* symbol tables */
12361 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
12362 PL_curstash = hv_dup(proto_perl->Icurstash, param);
12363 PL_debstash = hv_dup(proto_perl->Idebstash, param);
12364 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
12365 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
12367 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
12368 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
12369 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
12370 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
12371 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
12372 PL_endav = av_dup_inc(proto_perl->Iendav, param);
12373 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
12374 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
12376 PL_sub_generation = proto_perl->Isub_generation;
12377 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
12379 /* funky return mechanisms */
12380 PL_forkprocess = proto_perl->Iforkprocess;
12382 /* subprocess state */
12383 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
12385 /* internal state */
12386 PL_maxo = proto_perl->Imaxo;
12387 if (proto_perl->Iop_mask)
12388 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
12391 /* PL_asserting = proto_perl->Iasserting; */
12393 /* current interpreter roots */
12394 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
12396 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
12398 PL_main_start = proto_perl->Imain_start;
12399 PL_eval_root = proto_perl->Ieval_root;
12400 PL_eval_start = proto_perl->Ieval_start;
12402 /* runtime control stuff */
12403 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
12405 PL_filemode = proto_perl->Ifilemode;
12406 PL_lastfd = proto_perl->Ilastfd;
12407 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12410 PL_gensym = proto_perl->Igensym;
12411 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
12412 PL_laststatval = proto_perl->Ilaststatval;
12413 PL_laststype = proto_perl->Ilaststype;
12416 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
12418 /* interpreter atexit processing */
12419 PL_exitlistlen = proto_perl->Iexitlistlen;
12420 if (PL_exitlistlen) {
12421 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12422 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12425 PL_exitlist = (PerlExitListEntry*)NULL;
12427 PL_my_cxt_size = proto_perl->Imy_cxt_size;
12428 if (PL_my_cxt_size) {
12429 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
12430 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
12431 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12432 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
12433 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
12437 PL_my_cxt_list = (void**)NULL;
12438 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12439 PL_my_cxt_keys = (const char**)NULL;
12442 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
12443 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
12444 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
12446 PL_profiledata = NULL;
12448 PL_compcv = cv_dup(proto_perl->Icompcv, param);
12450 PAD_CLONE_VARS(proto_perl, param);
12452 #ifdef HAVE_INTERP_INTERN
12453 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
12456 /* more statics moved here */
12457 PL_generation = proto_perl->Igeneration;
12458 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
12460 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12461 PL_in_clean_all = proto_perl->Iin_clean_all;
12463 PL_uid = proto_perl->Iuid;
12464 PL_euid = proto_perl->Ieuid;
12465 PL_gid = proto_perl->Igid;
12466 PL_egid = proto_perl->Iegid;
12467 PL_nomemok = proto_perl->Inomemok;
12468 PL_an = proto_perl->Ian;
12469 PL_evalseq = proto_perl->Ievalseq;
12470 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12471 PL_origalen = proto_perl->Iorigalen;
12472 #ifdef PERL_USES_PL_PIDSTATUS
12473 PL_pidstatus = newHV(); /* XXX flag for cloning? */
12475 PL_osname = SAVEPV(proto_perl->Iosname);
12476 PL_sighandlerp = proto_perl->Isighandlerp;
12478 PL_runops = proto_perl->Irunops;
12480 PL_parser = parser_dup(proto_perl->Iparser, param);
12482 /* XXX this only works if the saved cop has already been cloned */
12483 if (proto_perl->Iparser) {
12484 PL_parser->saved_curcop = (COP*)any_dup(
12485 proto_perl->Iparser->saved_curcop,
12489 PL_subline = proto_perl->Isubline;
12490 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
12493 PL_cryptseen = proto_perl->Icryptseen;
12496 PL_hints = proto_perl->Ihints;
12498 PL_amagic_generation = proto_perl->Iamagic_generation;
12500 #ifdef USE_LOCALE_COLLATE
12501 PL_collation_ix = proto_perl->Icollation_ix;
12502 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
12503 PL_collation_standard = proto_perl->Icollation_standard;
12504 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12505 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12506 #endif /* USE_LOCALE_COLLATE */
12508 #ifdef USE_LOCALE_NUMERIC
12509 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
12510 PL_numeric_standard = proto_perl->Inumeric_standard;
12511 PL_numeric_local = proto_perl->Inumeric_local;
12512 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
12513 #endif /* !USE_LOCALE_NUMERIC */
12515 /* utf8 character classes */
12516 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
12517 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
12518 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
12519 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
12520 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
12521 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
12522 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
12523 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
12524 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
12525 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
12526 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
12527 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
12528 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
12529 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
12530 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
12531 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
12532 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
12533 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
12534 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
12535 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
12536 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
12537 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
12538 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
12539 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
12540 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
12541 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
12542 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
12543 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
12544 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
12546 /* Did the locale setup indicate UTF-8? */
12547 PL_utf8locale = proto_perl->Iutf8locale;
12548 /* Unicode features (see perlrun/-C) */
12549 PL_unicode = proto_perl->Iunicode;
12551 /* Pre-5.8 signals control */
12552 PL_signals = proto_perl->Isignals;
12554 /* times() ticks per second */
12555 PL_clocktick = proto_perl->Iclocktick;
12557 /* Recursion stopper for PerlIO_find_layer */
12558 PL_in_load_module = proto_perl->Iin_load_module;
12560 /* sort() routine */
12561 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12563 /* Not really needed/useful since the reenrant_retint is "volatile",
12564 * but do it for consistency's sake. */
12565 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12567 /* Hooks to shared SVs and locks. */
12568 PL_sharehook = proto_perl->Isharehook;
12569 PL_lockhook = proto_perl->Ilockhook;
12570 PL_unlockhook = proto_perl->Iunlockhook;
12571 PL_threadhook = proto_perl->Ithreadhook;
12572 PL_destroyhook = proto_perl->Idestroyhook;
12573 PL_signalhook = proto_perl->Isignalhook;
12575 #ifdef THREADS_HAVE_PIDS
12576 PL_ppid = proto_perl->Ippid;
12580 PL_last_swash_hv = NULL; /* reinits on demand */
12581 PL_last_swash_klen = 0;
12582 PL_last_swash_key[0]= '\0';
12583 PL_last_swash_tmps = (U8*)NULL;
12584 PL_last_swash_slen = 0;
12586 PL_glob_index = proto_perl->Iglob_index;
12587 PL_srand_called = proto_perl->Isrand_called;
12589 if (proto_perl->Ipsig_pend) {
12590 Newxz(PL_psig_pend, SIG_SIZE, int);
12593 PL_psig_pend = (int*)NULL;
12596 if (proto_perl->Ipsig_name) {
12597 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
12598 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
12600 PL_psig_ptr = PL_psig_name + SIG_SIZE;
12603 PL_psig_ptr = (SV**)NULL;
12604 PL_psig_name = (SV**)NULL;
12607 /* intrpvar.h stuff */
12609 if (flags & CLONEf_COPY_STACKS) {
12610 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12611 PL_tmps_ix = proto_perl->Itmps_ix;
12612 PL_tmps_max = proto_perl->Itmps_max;
12613 PL_tmps_floor = proto_perl->Itmps_floor;
12614 Newx(PL_tmps_stack, PL_tmps_max, SV*);
12615 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
12616 PL_tmps_ix+1, param);
12618 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
12619 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
12620 Newxz(PL_markstack, i, I32);
12621 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
12622 - proto_perl->Imarkstack);
12623 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
12624 - proto_perl->Imarkstack);
12625 Copy(proto_perl->Imarkstack, PL_markstack,
12626 PL_markstack_ptr - PL_markstack + 1, I32);
12628 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12629 * NOTE: unlike the others! */
12630 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12631 PL_scopestack_max = proto_perl->Iscopestack_max;
12632 Newxz(PL_scopestack, PL_scopestack_max, I32);
12633 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
12636 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
12637 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
12639 /* NOTE: si_dup() looks at PL_markstack */
12640 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
12642 /* PL_curstack = PL_curstackinfo->si_stack; */
12643 PL_curstack = av_dup(proto_perl->Icurstack, param);
12644 PL_mainstack = av_dup(proto_perl->Imainstack, param);
12646 /* next PUSHs() etc. set *(PL_stack_sp+1) */
12647 PL_stack_base = AvARRAY(PL_curstack);
12648 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
12649 - proto_perl->Istack_base);
12650 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
12652 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12653 * NOTE: unlike the others! */
12654 PL_savestack_ix = proto_perl->Isavestack_ix;
12655 PL_savestack_max = proto_perl->Isavestack_max;
12656 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
12657 PL_savestack = ss_dup(proto_perl, param);
12661 ENTER; /* perl_destruct() wants to LEAVE; */
12664 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12665 PL_top_env = &PL_start_env;
12667 PL_op = proto_perl->Iop;
12670 PL_Xpv = (XPV*)NULL;
12671 my_perl->Ina = proto_perl->Ina;
12673 PL_statbuf = proto_perl->Istatbuf;
12674 PL_statcache = proto_perl->Istatcache;
12675 PL_statgv = gv_dup(proto_perl->Istatgv, param);
12676 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
12678 PL_timesbuf = proto_perl->Itimesbuf;
12681 PL_tainted = proto_perl->Itainted;
12682 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12683 PL_rs = sv_dup_inc(proto_perl->Irs, param);
12684 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
12685 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
12686 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12687 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
12688 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
12689 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
12691 PL_restartjmpenv = proto_perl->Irestartjmpenv;
12692 PL_restartop = proto_perl->Irestartop;
12693 PL_in_eval = proto_perl->Iin_eval;
12694 PL_delaymagic = proto_perl->Idelaymagic;
12695 PL_dirty = proto_perl->Idirty;
12696 PL_localizing = proto_perl->Ilocalizing;
12698 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
12699 PL_hv_fetch_ent_mh = NULL;
12700 PL_modcount = proto_perl->Imodcount;
12701 PL_lastgotoprobe = NULL;
12702 PL_dumpindent = proto_perl->Idumpindent;
12704 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
12705 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
12706 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
12707 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
12708 PL_efloatbuf = NULL; /* reinits on demand */
12709 PL_efloatsize = 0; /* reinits on demand */
12713 PL_screamfirst = NULL;
12714 PL_screamnext = NULL;
12715 PL_maxscream = -1; /* reinits on demand */
12716 PL_lastscream = NULL;
12719 PL_regdummy = proto_perl->Iregdummy;
12720 PL_colorset = 0; /* reinits PL_colors[] */
12721 /*PL_colors[6] = {0,0,0,0,0,0};*/
12725 /* Pluggable optimizer */
12726 PL_peepp = proto_perl->Ipeepp;
12727 /* op_free() hook */
12728 PL_opfreehook = proto_perl->Iopfreehook;
12730 PL_stashcache = newHV();
12732 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
12733 proto_perl->Iwatchaddr);
12734 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
12735 if (PL_debug && PL_watchaddr) {
12736 PerlIO_printf(Perl_debug_log,
12737 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
12738 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
12739 PTR2UV(PL_watchok));
12742 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
12743 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
12745 /* Call the ->CLONE method, if it exists, for each of the stashes
12746 identified by sv_dup() above.
12748 while(av_len(param->stashes) != -1) {
12749 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
12750 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
12751 if (cloner && GvCV(cloner)) {
12756 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
12758 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
12764 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
12765 ptr_table_free(PL_ptr_table);
12766 PL_ptr_table = NULL;
12769 if (!(flags & CLONEf_COPY_STACKS)) {
12770 unreferenced_to_tmp_stack(param->unreferenced);
12773 SvREFCNT_dec(param->stashes);
12775 /* orphaned? eg threads->new inside BEGIN or use */
12776 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
12777 SvREFCNT_inc_simple_void(PL_compcv);
12778 SAVEFREESV(PL_compcv);
12785 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
12787 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
12789 if (AvFILLp(unreferenced) > -1) {
12790 SV **svp = AvARRAY(unreferenced);
12791 SV **const last = svp + AvFILLp(unreferenced);
12795 if (SvREFCNT(*svp) == 1)
12797 } while (++svp <= last);
12799 EXTEND_MORTAL(count);
12800 svp = AvARRAY(unreferenced);
12803 if (SvREFCNT(*svp) == 1) {
12804 /* Our reference is the only one to this SV. This means that
12805 in this thread, the scalar effectively has a 0 reference.
12806 That doesn't work (cleanup never happens), so donate our
12807 reference to it onto the save stack. */
12808 PL_tmps_stack[++PL_tmps_ix] = *svp;
12810 /* As an optimisation, because we are already walking the
12811 entire array, instead of above doing either
12812 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
12813 release our reference to the scalar, so that at the end of
12814 the array owns zero references to the scalars it happens to
12815 point to. We are effectively converting the array from
12816 AvREAL() on to AvREAL() off. This saves the av_clear()
12817 (triggered by the SvREFCNT_dec(unreferenced) below) from
12818 walking the array a second time. */
12819 SvREFCNT_dec(*svp);
12822 } while (++svp <= last);
12823 AvREAL_off(unreferenced);
12825 SvREFCNT_dec(unreferenced);
12829 Perl_clone_params_del(CLONE_PARAMS *param)
12831 PerlInterpreter *const was = PERL_GET_THX;
12832 PerlInterpreter *const to = param->new_perl;
12835 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
12841 SvREFCNT_dec(param->stashes);
12842 if (param->unreferenced)
12843 unreferenced_to_tmp_stack(param->unreferenced);
12853 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
12855 /* Need to play this game, as newAV() can call safesysmalloc(), and that
12856 does a dTHX; to get the context from thread local storage.
12857 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
12858 a version that passes in my_perl. */
12859 PerlInterpreter *const was = PERL_GET_THX;
12860 CLONE_PARAMS *param;
12862 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
12868 /* Given that we've set the context, we can do this unshared. */
12869 Newx(param, 1, CLONE_PARAMS);
12872 param->proto_perl = from;
12873 param->new_perl = to;
12874 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
12875 AvREAL_off(param->stashes);
12876 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
12884 #endif /* USE_ITHREADS */
12887 =head1 Unicode Support
12889 =for apidoc sv_recode_to_utf8
12891 The encoding is assumed to be an Encode object, on entry the PV
12892 of the sv is assumed to be octets in that encoding, and the sv
12893 will be converted into Unicode (and UTF-8).
12895 If the sv already is UTF-8 (or if it is not POK), or if the encoding
12896 is not a reference, nothing is done to the sv. If the encoding is not
12897 an C<Encode::XS> Encoding object, bad things will happen.
12898 (See F<lib/encoding.pm> and L<Encode>).
12900 The PV of the sv is returned.
12905 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
12909 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
12911 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
12925 Passing sv_yes is wrong - it needs to be or'ed set of constants
12926 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
12927 remove converted chars from source.
12929 Both will default the value - let them.
12931 XPUSHs(&PL_sv_yes);
12934 call_method("decode", G_SCALAR);
12938 s = SvPV_const(uni, len);
12939 if (s != SvPVX_const(sv)) {
12940 SvGROW(sv, len + 1);
12941 Move(s, SvPVX(sv), len + 1, char);
12942 SvCUR_set(sv, len);
12949 return SvPOKp(sv) ? SvPVX(sv) : NULL;
12953 =for apidoc sv_cat_decode
12955 The encoding is assumed to be an Encode object, the PV of the ssv is
12956 assumed to be octets in that encoding and decoding the input starts
12957 from the position which (PV + *offset) pointed to. The dsv will be
12958 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
12959 when the string tstr appears in decoding output or the input ends on
12960 the PV of the ssv. The value which the offset points will be modified
12961 to the last input position on the ssv.
12963 Returns TRUE if the terminator was found, else returns FALSE.
12968 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
12969 SV *ssv, int *offset, char *tstr, int tlen)
12974 PERL_ARGS_ASSERT_SV_CAT_DECODE;
12976 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
12987 offsv = newSViv(*offset);
12989 mXPUSHp(tstr, tlen);
12991 call_method("cat_decode", G_SCALAR);
12993 ret = SvTRUE(TOPs);
12994 *offset = SvIV(offsv);
13000 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13005 /* ---------------------------------------------------------------------
13007 * support functions for report_uninit()
13010 /* the maxiumum size of array or hash where we will scan looking
13011 * for the undefined element that triggered the warning */
13013 #define FUV_MAX_SEARCH_SIZE 1000
13015 /* Look for an entry in the hash whose value has the same SV as val;
13016 * If so, return a mortal copy of the key. */
13019 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13022 register HE **array;
13025 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13027 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13028 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13031 array = HvARRAY(hv);
13033 for (i=HvMAX(hv); i>0; i--) {
13034 register HE *entry;
13035 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13036 if (HeVAL(entry) != val)
13038 if ( HeVAL(entry) == &PL_sv_undef ||
13039 HeVAL(entry) == &PL_sv_placeholder)
13043 if (HeKLEN(entry) == HEf_SVKEY)
13044 return sv_mortalcopy(HeKEY_sv(entry));
13045 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13051 /* Look for an entry in the array whose value has the same SV as val;
13052 * If so, return the index, otherwise return -1. */
13055 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13059 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13061 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13062 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13065 if (val != &PL_sv_undef) {
13066 SV ** const svp = AvARRAY(av);
13069 for (i=AvFILLp(av); i>=0; i--)
13076 /* S_varname(): return the name of a variable, optionally with a subscript.
13077 * If gv is non-zero, use the name of that global, along with gvtype (one
13078 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13079 * targ. Depending on the value of the subscript_type flag, return:
13082 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13083 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13084 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13085 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13088 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13089 const SV *const keyname, I32 aindex, int subscript_type)
13092 SV * const name = sv_newmortal();
13095 buffer[0] = gvtype;
13098 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13100 gv_fullname4(name, gv, buffer, 0);
13102 if ((unsigned int)SvPVX(name)[1] <= 26) {
13104 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13106 /* Swap the 1 unprintable control character for the 2 byte pretty
13107 version - ie substr($name, 1, 1) = $buffer; */
13108 sv_insert(name, 1, 1, buffer, 2);
13112 CV * const cv = find_runcv(NULL);
13116 if (!cv || !CvPADLIST(cv))
13118 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13119 sv = *av_fetch(av, targ, FALSE);
13120 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
13123 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13124 SV * const sv = newSV(0);
13125 *SvPVX(name) = '$';
13126 Perl_sv_catpvf(aTHX_ name, "{%s}",
13127 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
13130 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13131 *SvPVX(name) = '$';
13132 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13134 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13135 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13136 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13144 =for apidoc find_uninit_var
13146 Find the name of the undefined variable (if any) that caused the operator o
13147 to issue a "Use of uninitialized value" warning.
13148 If match is true, only return a name if it's value matches uninit_sv.
13149 So roughly speaking, if a unary operator (such as OP_COS) generates a
13150 warning, then following the direct child of the op may yield an
13151 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13152 other hand, with OP_ADD there are two branches to follow, so we only print
13153 the variable name if we get an exact match.
13155 The name is returned as a mortal SV.
13157 Assumes that PL_op is the op that originally triggered the error, and that
13158 PL_comppad/PL_curpad points to the currently executing pad.
13164 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13170 const OP *o, *o2, *kid;
13172 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13173 uninit_sv == &PL_sv_placeholder)))
13176 switch (obase->op_type) {
13183 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13184 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13187 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13189 if (pad) { /* @lex, %lex */
13190 sv = PAD_SVl(obase->op_targ);
13194 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13195 /* @global, %global */
13196 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13199 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13201 else /* @{expr}, %{expr} */
13202 return find_uninit_var(cUNOPx(obase)->op_first,
13206 /* attempt to find a match within the aggregate */
13208 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13210 subscript_type = FUV_SUBSCRIPT_HASH;
13213 index = find_array_subscript((const AV *)sv, uninit_sv);
13215 subscript_type = FUV_SUBSCRIPT_ARRAY;
13218 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
13221 return varname(gv, hash ? '%' : '@', obase->op_targ,
13222 keysv, index, subscript_type);
13226 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
13228 return varname(NULL, '$', obase->op_targ,
13229 NULL, 0, FUV_SUBSCRIPT_NONE);
13232 gv = cGVOPx_gv(obase);
13233 if (!gv || (match && GvSV(gv) != uninit_sv))
13235 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13238 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
13241 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
13242 if (!av || SvRMAGICAL(av))
13244 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13245 if (!svp || *svp != uninit_sv)
13248 return varname(NULL, '$', obase->op_targ,
13249 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13252 gv = cGVOPx_gv(obase);
13257 AV *const av = GvAV(gv);
13258 if (!av || SvRMAGICAL(av))
13260 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13261 if (!svp || *svp != uninit_sv)
13264 return varname(gv, '$', 0,
13265 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13270 o = cUNOPx(obase)->op_first;
13271 if (!o || o->op_type != OP_NULL ||
13272 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
13274 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
13278 if (PL_op == obase)
13279 /* $a[uninit_expr] or $h{uninit_expr} */
13280 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
13283 o = cBINOPx(obase)->op_first;
13284 kid = cBINOPx(obase)->op_last;
13286 /* get the av or hv, and optionally the gv */
13288 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
13289 sv = PAD_SV(o->op_targ);
13291 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
13292 && cUNOPo->op_first->op_type == OP_GV)
13294 gv = cGVOPx_gv(cUNOPo->op_first);
13298 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
13303 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
13304 /* index is constant */
13308 if (obase->op_type == OP_HELEM) {
13309 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
13310 if (!he || HeVAL(he) != uninit_sv)
13314 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
13315 if (!svp || *svp != uninit_sv)
13319 if (obase->op_type == OP_HELEM)
13320 return varname(gv, '%', o->op_targ,
13321 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
13323 return varname(gv, '@', o->op_targ, NULL,
13324 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
13327 /* index is an expression;
13328 * attempt to find a match within the aggregate */
13329 if (obase->op_type == OP_HELEM) {
13330 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13332 return varname(gv, '%', o->op_targ,
13333 keysv, 0, FUV_SUBSCRIPT_HASH);
13337 = find_array_subscript((const AV *)sv, uninit_sv);
13339 return varname(gv, '@', o->op_targ,
13340 NULL, index, FUV_SUBSCRIPT_ARRAY);
13345 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
13347 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
13352 /* only examine RHS */
13353 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
13356 o = cUNOPx(obase)->op_first;
13357 if (o->op_type == OP_PUSHMARK)
13360 if (!o->op_sibling) {
13361 /* one-arg version of open is highly magical */
13363 if (o->op_type == OP_GV) { /* open FOO; */
13365 if (match && GvSV(gv) != uninit_sv)
13367 return varname(gv, '$', 0,
13368 NULL, 0, FUV_SUBSCRIPT_NONE);
13370 /* other possibilities not handled are:
13371 * open $x; or open my $x; should return '${*$x}'
13372 * open expr; should return '$'.expr ideally
13378 /* ops where $_ may be an implicit arg */
13382 if ( !(obase->op_flags & OPf_STACKED)) {
13383 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
13384 ? PAD_SVl(obase->op_targ)
13387 sv = sv_newmortal();
13388 sv_setpvs(sv, "$_");
13397 match = 1; /* print etc can return undef on defined args */
13398 /* skip filehandle as it can't produce 'undef' warning */
13399 o = cUNOPx(obase)->op_first;
13400 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
13401 o = o->op_sibling->op_sibling;
13405 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
13407 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
13409 /* the following ops are capable of returning PL_sv_undef even for
13410 * defined arg(s) */
13429 case OP_GETPEERNAME:
13477 case OP_SMARTMATCH:
13486 /* XXX tmp hack: these two may call an XS sub, and currently
13487 XS subs don't have a SUB entry on the context stack, so CV and
13488 pad determination goes wrong, and BAD things happen. So, just
13489 don't try to determine the value under those circumstances.
13490 Need a better fix at dome point. DAPM 11/2007 */
13496 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
13497 if (gv && GvSV(gv) == uninit_sv)
13498 return newSVpvs_flags("$.", SVs_TEMP);
13503 /* def-ness of rval pos() is independent of the def-ness of its arg */
13504 if ( !(obase->op_flags & OPf_MOD))
13509 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
13510 return newSVpvs_flags("${$/}", SVs_TEMP);
13515 if (!(obase->op_flags & OPf_KIDS))
13517 o = cUNOPx(obase)->op_first;
13523 /* if all except one arg are constant, or have no side-effects,
13524 * or are optimized away, then it's unambiguous */
13526 for (kid=o; kid; kid = kid->op_sibling) {
13528 const OPCODE type = kid->op_type;
13529 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
13530 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
13531 || (type == OP_PUSHMARK)
13535 if (o2) { /* more than one found */
13542 return find_uninit_var(o2, uninit_sv, match);
13544 /* scan all args */
13546 sv = find_uninit_var(o, uninit_sv, 1);
13558 =for apidoc report_uninit
13560 Print appropriate "Use of uninitialized variable" warning
13566 Perl_report_uninit(pTHX_ const SV *uninit_sv)
13570 SV* varname = NULL;
13572 varname = find_uninit_var(PL_op, uninit_sv,0);
13574 sv_insert(varname, 0, 0, " ", 1);
13576 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13577 varname ? SvPV_nolen_const(varname) : "",
13578 " in ", OP_DESC(PL_op));
13581 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13587 * c-indentation-style: bsd
13588 * c-basic-offset: 4
13589 * indent-tabs-mode: t
13592 * ex: set ts=8 sts=4 sw=4 noet: