3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009 by Larry Wall
7 * You may distribute under the terms of either the GNU General Public
8 * License or the Artistic License, as specified in the README file.
13 * 'I wonder what the Entish is for "yes" and "no",' he thought.
16 * [p.480 of _The Lord of the Rings_, III/iv: "Treebeard"]
22 * This file contains the code that creates, manipulates and destroys
23 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
24 * structure of an SV, so their creation and destruction is handled
25 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
26 * level functions (eg. substr, split, join) for each of the types are
36 # if __STDC_VERSION__ >= 199901L && !defined(VMS)
45 /* Missing proto on LynxOS */
46 char *gconvert(double, int, int, char *);
49 /* void Gconvert: on Linux at least, gcvt (which Gconvert gets deffed to),
50 * has a mandatory return value, even though that value is just the same
53 #define V_Gconvert(x,n,t,b) \
55 char *rc = (char *)Gconvert(x,n,t,b); \
56 PERL_UNUSED_VAR(rc); \
60 #ifdef PERL_UTF8_CACHE_ASSERT
61 /* if adding more checks watch out for the following tests:
62 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
63 * lib/utf8.t lib/Unicode/Collate/t/index.t
66 # define ASSERT_UTF8_CACHE(cache) \
67 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
68 assert((cache)[2] <= (cache)[3]); \
69 assert((cache)[3] <= (cache)[1]);} \
72 # define ASSERT_UTF8_CACHE(cache) NOOP
75 #ifdef PERL_OLD_COPY_ON_WRITE
76 #define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
77 #define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
80 /* ============================================================================
82 =head1 Allocation and deallocation of SVs.
84 An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
85 sv, av, hv...) contains type and reference count information, and for
86 many types, a pointer to the body (struct xrv, xpv, xpviv...), which
87 contains fields specific to each type. Some types store all they need
88 in the head, so don't have a body.
90 In all but the most memory-paranoid configurations (ex: PURIFY), heads
91 and bodies are allocated out of arenas, which by default are
92 approximately 4K chunks of memory parcelled up into N heads or bodies.
93 Sv-bodies are allocated by their sv-type, guaranteeing size
94 consistency needed to allocate safely from arrays.
96 For SV-heads, the first slot in each arena is reserved, and holds a
97 link to the next arena, some flags, and a note of the number of slots.
98 Snaked through each arena chain is a linked list of free items; when
99 this becomes empty, an extra arena is allocated and divided up into N
100 items which are threaded into the free list.
102 SV-bodies are similar, but they use arena-sets by default, which
103 separate the link and info from the arena itself, and reclaim the 1st
104 slot in the arena. SV-bodies are further described later.
106 The following global variables are associated with arenas:
108 PL_sv_arenaroot pointer to list of SV arenas
109 PL_sv_root pointer to list of free SV structures
111 PL_body_arenas head of linked-list of body arenas
112 PL_body_roots[] array of pointers to list of free bodies of svtype
113 arrays are indexed by the svtype needed
115 A few special SV heads are not allocated from an arena, but are
116 instead directly created in the interpreter structure, eg PL_sv_undef.
117 The size of arenas can be changed from the default by setting
118 PERL_ARENA_SIZE appropriately at compile time.
120 The SV arena serves the secondary purpose of allowing still-live SVs
121 to be located and destroyed during final cleanup.
123 At the lowest level, the macros new_SV() and del_SV() grab and free
124 an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
125 to return the SV to the free list with error checking.) new_SV() calls
126 more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
127 SVs in the free list have their SvTYPE field set to all ones.
129 At the time of very final cleanup, sv_free_arenas() is called from
130 perl_destruct() to physically free all the arenas allocated since the
131 start of the interpreter.
133 The function visit() scans the SV arenas list, and calls a specified
134 function for each SV it finds which is still live - ie which has an SvTYPE
135 other than all 1's, and a non-zero SvREFCNT. visit() is used by the
136 following functions (specified as [function that calls visit()] / [function
137 called by visit() for each SV]):
139 sv_report_used() / do_report_used()
140 dump all remaining SVs (debugging aid)
142 sv_clean_objs() / do_clean_objs(),do_clean_named_objs(),
143 do_clean_named_io_objs(),do_curse()
144 Attempt to free all objects pointed to by RVs,
145 try to do the same for all objects indir-
146 ectly referenced by typeglobs too, and
147 then do a final sweep, cursing any
148 objects that remain. Called once from
149 perl_destruct(), prior to calling sv_clean_all()
152 sv_clean_all() / do_clean_all()
153 SvREFCNT_dec(sv) each remaining SV, possibly
154 triggering an sv_free(). It also sets the
155 SVf_BREAK flag on the SV to indicate that the
156 refcnt has been artificially lowered, and thus
157 stopping sv_free() from giving spurious warnings
158 about SVs which unexpectedly have a refcnt
159 of zero. called repeatedly from perl_destruct()
160 until there are no SVs left.
162 =head2 Arena allocator API Summary
164 Private API to rest of sv.c
168 new_XPVNV(), del_XPVGV(),
173 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
177 * ========================================================================= */
180 * "A time to plant, and a time to uproot what was planted..."
184 # define MEM_LOG_NEW_SV(sv, file, line, func) \
185 Perl_mem_log_new_sv(sv, file, line, func)
186 # define MEM_LOG_DEL_SV(sv, file, line, func) \
187 Perl_mem_log_del_sv(sv, file, line, func)
189 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
190 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
193 #ifdef DEBUG_LEAKING_SCALARS
194 # define FREE_SV_DEBUG_FILE(sv) STMT_START { \
195 if ((sv)->sv_debug_file) PerlMemShared_free((sv)->sv_debug_file); \
197 # define DEBUG_SV_SERIAL(sv) \
198 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
199 PTR2UV(sv), (long)(sv)->sv_debug_serial))
201 # define FREE_SV_DEBUG_FILE(sv)
202 # define DEBUG_SV_SERIAL(sv) NOOP
206 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
207 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
208 /* Whilst I'd love to do this, it seems that things like to check on
210 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
212 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
213 PoisonNew(&SvREFCNT(sv), 1, U32)
215 # define SvARENA_CHAIN(sv) SvANY(sv)
216 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
217 # define POSION_SV_HEAD(sv)
220 /* Mark an SV head as unused, and add to free list.
222 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
223 * its refcount artificially decremented during global destruction, so
224 * there may be dangling pointers to it. The last thing we want in that
225 * case is for it to be reused. */
227 #define plant_SV(p) \
229 const U32 old_flags = SvFLAGS(p); \
230 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
231 DEBUG_SV_SERIAL(p); \
232 FREE_SV_DEBUG_FILE(p); \
234 SvFLAGS(p) = SVTYPEMASK; \
235 if (!(old_flags & SVf_BREAK)) { \
236 SvARENA_CHAIN_SET(p, PL_sv_root); \
242 #define uproot_SV(p) \
245 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
250 /* make some more SVs by adding another arena */
257 char *chunk; /* must use New here to match call to */
258 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
259 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
264 /* new_SV(): return a new, empty SV head */
266 #ifdef DEBUG_LEAKING_SCALARS
267 /* provide a real function for a debugger to play with */
269 S_new_SV(pTHX_ const char *file, int line, const char *func)
276 sv = S_more_sv(aTHX);
280 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
281 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
287 sv->sv_debug_inpad = 0;
288 sv->sv_debug_parent = NULL;
289 sv->sv_debug_file = PL_curcop ? savesharedpv(CopFILE(PL_curcop)): NULL;
291 sv->sv_debug_serial = PL_sv_serial++;
293 MEM_LOG_NEW_SV(sv, file, line, func);
294 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
295 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
299 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
307 (p) = S_more_sv(aTHX); \
311 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
316 /* del_SV(): return an empty SV head to the free list */
329 S_del_sv(pTHX_ SV *p)
333 PERL_ARGS_ASSERT_DEL_SV;
338 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
339 const SV * const sv = sva + 1;
340 const SV * const svend = &sva[SvREFCNT(sva)];
341 if (p >= sv && p < svend) {
347 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
348 "Attempt to free non-arena SV: 0x%"UVxf
349 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
356 #else /* ! DEBUGGING */
358 #define del_SV(p) plant_SV(p)
360 #endif /* DEBUGGING */
364 =head1 SV Manipulation Functions
366 =for apidoc sv_add_arena
368 Given a chunk of memory, link it to the head of the list of arenas,
369 and split it into a list of free SVs.
375 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
378 SV *const sva = MUTABLE_SV(ptr);
382 PERL_ARGS_ASSERT_SV_ADD_ARENA;
384 /* The first SV in an arena isn't an SV. */
385 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
386 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
387 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
389 PL_sv_arenaroot = sva;
390 PL_sv_root = sva + 1;
392 svend = &sva[SvREFCNT(sva) - 1];
395 SvARENA_CHAIN_SET(sv, (sv + 1));
399 /* Must always set typemask because it's always checked in on cleanup
400 when the arenas are walked looking for objects. */
401 SvFLAGS(sv) = SVTYPEMASK;
404 SvARENA_CHAIN_SET(sv, 0);
408 SvFLAGS(sv) = SVTYPEMASK;
411 /* visit(): call the named function for each non-free SV in the arenas
412 * whose flags field matches the flags/mask args. */
415 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
421 PERL_ARGS_ASSERT_VISIT;
423 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
424 const SV * const svend = &sva[SvREFCNT(sva)];
426 for (sv = sva + 1; sv < svend; ++sv) {
427 if (SvTYPE(sv) != (svtype)SVTYPEMASK
428 && (sv->sv_flags & mask) == flags
441 /* called by sv_report_used() for each live SV */
444 do_report_used(pTHX_ SV *const sv)
446 if (SvTYPE(sv) != (svtype)SVTYPEMASK) {
447 PerlIO_printf(Perl_debug_log, "****\n");
454 =for apidoc sv_report_used
456 Dump the contents of all SVs not yet freed (debugging aid).
462 Perl_sv_report_used(pTHX)
465 visit(do_report_used, 0, 0);
471 /* called by sv_clean_objs() for each live SV */
474 do_clean_objs(pTHX_ SV *const ref)
479 SV * const target = SvRV(ref);
480 if (SvOBJECT(target)) {
481 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
482 if (SvWEAKREF(ref)) {
483 sv_del_backref(target, ref);
489 SvREFCNT_dec_NN(target);
496 /* clear any slots in a GV which hold objects - except IO;
497 * called by sv_clean_objs() for each live GV */
500 do_clean_named_objs(pTHX_ SV *const sv)
504 assert(SvTYPE(sv) == SVt_PVGV);
505 assert(isGV_with_GP(sv));
509 /* freeing GP entries may indirectly free the current GV;
510 * hold onto it while we mess with the GP slots */
513 if ( ((obj = GvSV(sv) )) && SvOBJECT(obj)) {
514 DEBUG_D((PerlIO_printf(Perl_debug_log,
515 "Cleaning named glob SV object:\n "), sv_dump(obj)));
517 SvREFCNT_dec_NN(obj);
519 if ( ((obj = MUTABLE_SV(GvAV(sv)) )) && SvOBJECT(obj)) {
520 DEBUG_D((PerlIO_printf(Perl_debug_log,
521 "Cleaning named glob AV object:\n "), sv_dump(obj)));
523 SvREFCNT_dec_NN(obj);
525 if ( ((obj = MUTABLE_SV(GvHV(sv)) )) && SvOBJECT(obj)) {
526 DEBUG_D((PerlIO_printf(Perl_debug_log,
527 "Cleaning named glob HV object:\n "), sv_dump(obj)));
529 SvREFCNT_dec_NN(obj);
531 if ( ((obj = MUTABLE_SV(GvCV(sv)) )) && SvOBJECT(obj)) {
532 DEBUG_D((PerlIO_printf(Perl_debug_log,
533 "Cleaning named glob CV object:\n "), sv_dump(obj)));
535 SvREFCNT_dec_NN(obj);
537 SvREFCNT_dec_NN(sv); /* undo the inc above */
540 /* clear any IO slots in a GV which hold objects (except stderr, defout);
541 * called by sv_clean_objs() for each live GV */
544 do_clean_named_io_objs(pTHX_ SV *const sv)
548 assert(SvTYPE(sv) == SVt_PVGV);
549 assert(isGV_with_GP(sv));
550 if (!GvGP(sv) || sv == (SV*)PL_stderrgv || sv == (SV*)PL_defoutgv)
554 if ( ((obj = MUTABLE_SV(GvIO(sv)) )) && SvOBJECT(obj)) {
555 DEBUG_D((PerlIO_printf(Perl_debug_log,
556 "Cleaning named glob IO object:\n "), sv_dump(obj)));
558 SvREFCNT_dec_NN(obj);
560 SvREFCNT_dec_NN(sv); /* undo the inc above */
563 /* Void wrapper to pass to visit() */
565 do_curse(pTHX_ SV * const sv) {
566 if ((PL_stderrgv && GvGP(PL_stderrgv) && (SV*)GvIO(PL_stderrgv) == sv)
567 || (PL_defoutgv && GvGP(PL_defoutgv) && (SV*)GvIO(PL_defoutgv) == sv))
573 =for apidoc sv_clean_objs
575 Attempt to destroy all objects not yet freed.
581 Perl_sv_clean_objs(pTHX)
585 PL_in_clean_objs = TRUE;
586 visit(do_clean_objs, SVf_ROK, SVf_ROK);
587 /* Some barnacles may yet remain, clinging to typeglobs.
588 * Run the non-IO destructors first: they may want to output
589 * error messages, close files etc */
590 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
591 visit(do_clean_named_io_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
592 /* And if there are some very tenacious barnacles clinging to arrays,
593 closures, or what have you.... */
594 visit(do_curse, SVs_OBJECT, SVs_OBJECT);
595 olddef = PL_defoutgv;
596 PL_defoutgv = NULL; /* disable skip of PL_defoutgv */
597 if (olddef && isGV_with_GP(olddef))
598 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olddef));
599 olderr = PL_stderrgv;
600 PL_stderrgv = NULL; /* disable skip of PL_stderrgv */
601 if (olderr && isGV_with_GP(olderr))
602 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olderr));
603 SvREFCNT_dec(olddef);
604 PL_in_clean_objs = FALSE;
607 /* called by sv_clean_all() for each live SV */
610 do_clean_all(pTHX_ SV *const sv)
613 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
614 /* don't clean pid table and strtab */
617 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
618 SvFLAGS(sv) |= SVf_BREAK;
623 =for apidoc sv_clean_all
625 Decrement the refcnt of each remaining SV, possibly triggering a
626 cleanup. This function may have to be called multiple times to free
627 SVs which are in complex self-referential hierarchies.
633 Perl_sv_clean_all(pTHX)
637 PL_in_clean_all = TRUE;
638 cleaned = visit(do_clean_all, 0,0);
643 ARENASETS: a meta-arena implementation which separates arena-info
644 into struct arena_set, which contains an array of struct
645 arena_descs, each holding info for a single arena. By separating
646 the meta-info from the arena, we recover the 1st slot, formerly
647 borrowed for list management. The arena_set is about the size of an
648 arena, avoiding the needless malloc overhead of a naive linked-list.
650 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
651 memory in the last arena-set (1/2 on average). In trade, we get
652 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
653 smaller types). The recovery of the wasted space allows use of
654 small arenas for large, rare body types, by changing array* fields
655 in body_details_by_type[] below.
658 char *arena; /* the raw storage, allocated aligned */
659 size_t size; /* its size ~4k typ */
660 svtype utype; /* bodytype stored in arena */
665 /* Get the maximum number of elements in set[] such that struct arena_set
666 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
667 therefore likely to be 1 aligned memory page. */
669 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
670 - 2 * sizeof(int)) / sizeof (struct arena_desc))
673 struct arena_set* next;
674 unsigned int set_size; /* ie ARENAS_PER_SET */
675 unsigned int curr; /* index of next available arena-desc */
676 struct arena_desc set[ARENAS_PER_SET];
680 =for apidoc sv_free_arenas
682 Deallocate the memory used by all arenas. Note that all the individual SV
683 heads and bodies within the arenas must already have been freed.
688 Perl_sv_free_arenas(pTHX)
695 /* Free arenas here, but be careful about fake ones. (We assume
696 contiguity of the fake ones with the corresponding real ones.) */
698 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
699 svanext = MUTABLE_SV(SvANY(sva));
700 while (svanext && SvFAKE(svanext))
701 svanext = MUTABLE_SV(SvANY(svanext));
708 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
711 struct arena_set *current = aroot;
714 assert(aroot->set[i].arena);
715 Safefree(aroot->set[i].arena);
723 i = PERL_ARENA_ROOTS_SIZE;
725 PL_body_roots[i] = 0;
732 Here are mid-level routines that manage the allocation of bodies out
733 of the various arenas. There are 5 kinds of arenas:
735 1. SV-head arenas, which are discussed and handled above
736 2. regular body arenas
737 3. arenas for reduced-size bodies
740 Arena types 2 & 3 are chained by body-type off an array of
741 arena-root pointers, which is indexed by svtype. Some of the
742 larger/less used body types are malloced singly, since a large
743 unused block of them is wasteful. Also, several svtypes dont have
744 bodies; the data fits into the sv-head itself. The arena-root
745 pointer thus has a few unused root-pointers (which may be hijacked
746 later for arena types 4,5)
748 3 differs from 2 as an optimization; some body types have several
749 unused fields in the front of the structure (which are kept in-place
750 for consistency). These bodies can be allocated in smaller chunks,
751 because the leading fields arent accessed. Pointers to such bodies
752 are decremented to point at the unused 'ghost' memory, knowing that
753 the pointers are used with offsets to the real memory.
756 =head1 SV-Body Allocation
758 Allocation of SV-bodies is similar to SV-heads, differing as follows;
759 the allocation mechanism is used for many body types, so is somewhat
760 more complicated, it uses arena-sets, and has no need for still-live
763 At the outermost level, (new|del)_X*V macros return bodies of the
764 appropriate type. These macros call either (new|del)_body_type or
765 (new|del)_body_allocated macro pairs, depending on specifics of the
766 type. Most body types use the former pair, the latter pair is used to
767 allocate body types with "ghost fields".
769 "ghost fields" are fields that are unused in certain types, and
770 consequently don't need to actually exist. They are declared because
771 they're part of a "base type", which allows use of functions as
772 methods. The simplest examples are AVs and HVs, 2 aggregate types
773 which don't use the fields which support SCALAR semantics.
775 For these types, the arenas are carved up into appropriately sized
776 chunks, we thus avoid wasted memory for those unaccessed members.
777 When bodies are allocated, we adjust the pointer back in memory by the
778 size of the part not allocated, so it's as if we allocated the full
779 structure. (But things will all go boom if you write to the part that
780 is "not there", because you'll be overwriting the last members of the
781 preceding structure in memory.)
783 We calculate the correction using the STRUCT_OFFSET macro on the first
784 member present. If the allocated structure is smaller (no initial NV
785 actually allocated) then the net effect is to subtract the size of the NV
786 from the pointer, to return a new pointer as if an initial NV were actually
787 allocated. (We were using structures named *_allocated for this, but
788 this turned out to be a subtle bug, because a structure without an NV
789 could have a lower alignment constraint, but the compiler is allowed to
790 optimised accesses based on the alignment constraint of the actual pointer
791 to the full structure, for example, using a single 64 bit load instruction
792 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
794 This is the same trick as was used for NV and IV bodies. Ironically it
795 doesn't need to be used for NV bodies any more, because NV is now at
796 the start of the structure. IV bodies don't need it either, because
797 they are no longer allocated.
799 In turn, the new_body_* allocators call S_new_body(), which invokes
800 new_body_inline macro, which takes a lock, and takes a body off the
801 linked list at PL_body_roots[sv_type], calling Perl_more_bodies() if
802 necessary to refresh an empty list. Then the lock is released, and
803 the body is returned.
805 Perl_more_bodies allocates a new arena, and carves it up into an array of N
806 bodies, which it strings into a linked list. It looks up arena-size
807 and body-size from the body_details table described below, thus
808 supporting the multiple body-types.
810 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
811 the (new|del)_X*V macros are mapped directly to malloc/free.
813 For each sv-type, struct body_details bodies_by_type[] carries
814 parameters which control these aspects of SV handling:
816 Arena_size determines whether arenas are used for this body type, and if
817 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
818 zero, forcing individual mallocs and frees.
820 Body_size determines how big a body is, and therefore how many fit into
821 each arena. Offset carries the body-pointer adjustment needed for
822 "ghost fields", and is used in *_allocated macros.
824 But its main purpose is to parameterize info needed in
825 Perl_sv_upgrade(). The info here dramatically simplifies the function
826 vs the implementation in 5.8.8, making it table-driven. All fields
827 are used for this, except for arena_size.
829 For the sv-types that have no bodies, arenas are not used, so those
830 PL_body_roots[sv_type] are unused, and can be overloaded. In
831 something of a special case, SVt_NULL is borrowed for HE arenas;
832 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
833 bodies_by_type[SVt_NULL] slot is not used, as the table is not
838 struct body_details {
839 U8 body_size; /* Size to allocate */
840 U8 copy; /* Size of structure to copy (may be shorter) */
842 unsigned int type : 4; /* We have space for a sanity check. */
843 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
844 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
845 unsigned int arena : 1; /* Allocated from an arena */
846 size_t arena_size; /* Size of arena to allocate */
854 /* With -DPURFIY we allocate everything directly, and don't use arenas.
855 This seems a rather elegant way to simplify some of the code below. */
856 #define HASARENA FALSE
858 #define HASARENA TRUE
860 #define NOARENA FALSE
862 /* Size the arenas to exactly fit a given number of bodies. A count
863 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
864 simplifying the default. If count > 0, the arena is sized to fit
865 only that many bodies, allowing arenas to be used for large, rare
866 bodies (XPVFM, XPVIO) without undue waste. The arena size is
867 limited by PERL_ARENA_SIZE, so we can safely oversize the
870 #define FIT_ARENA0(body_size) \
871 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
872 #define FIT_ARENAn(count,body_size) \
873 ( count * body_size <= PERL_ARENA_SIZE) \
874 ? count * body_size \
875 : FIT_ARENA0 (body_size)
876 #define FIT_ARENA(count,body_size) \
878 ? FIT_ARENAn (count, body_size) \
879 : FIT_ARENA0 (body_size)
881 /* Calculate the length to copy. Specifically work out the length less any
882 final padding the compiler needed to add. See the comment in sv_upgrade
883 for why copying the padding proved to be a bug. */
885 #define copy_length(type, last_member) \
886 STRUCT_OFFSET(type, last_member) \
887 + sizeof (((type*)SvANY((const SV *)0))->last_member)
889 static const struct body_details bodies_by_type[] = {
890 /* HEs use this offset for their arena. */
891 { 0, 0, 0, SVt_NULL, FALSE, NONV, NOARENA, 0 },
893 /* IVs are in the head, so the allocation size is 0. */
895 sizeof(IV), /* This is used to copy out the IV body. */
896 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
897 NOARENA /* IVS don't need an arena */, 0
900 { sizeof(NV), sizeof(NV),
901 STRUCT_OFFSET(XPVNV, xnv_u),
902 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
904 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
905 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
906 + STRUCT_OFFSET(XPV, xpv_cur),
907 SVt_PV, FALSE, NONV, HASARENA,
908 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
910 { sizeof(XINVLIST) - STRUCT_OFFSET(XPV, xpv_cur),
911 copy_length(XINVLIST, is_offset) - STRUCT_OFFSET(XPV, xpv_cur),
912 + STRUCT_OFFSET(XPV, xpv_cur),
913 SVt_INVLIST, TRUE, NONV, HASARENA,
914 FIT_ARENA(0, sizeof(XINVLIST) - STRUCT_OFFSET(XPV, xpv_cur)) },
916 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
917 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
918 + STRUCT_OFFSET(XPV, xpv_cur),
919 SVt_PVIV, FALSE, NONV, HASARENA,
920 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
922 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
923 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
924 + STRUCT_OFFSET(XPV, xpv_cur),
925 SVt_PVNV, FALSE, HADNV, HASARENA,
926 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
928 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
929 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
934 SVt_REGEXP, TRUE, NONV, HASARENA,
935 FIT_ARENA(0, sizeof(regexp))
938 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
939 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
941 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
942 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
945 copy_length(XPVAV, xav_alloc),
947 SVt_PVAV, TRUE, NONV, HASARENA,
948 FIT_ARENA(0, sizeof(XPVAV)) },
951 copy_length(XPVHV, xhv_max),
953 SVt_PVHV, TRUE, NONV, HASARENA,
954 FIT_ARENA(0, sizeof(XPVHV)) },
959 SVt_PVCV, TRUE, NONV, HASARENA,
960 FIT_ARENA(0, sizeof(XPVCV)) },
965 SVt_PVFM, TRUE, NONV, NOARENA,
966 FIT_ARENA(20, sizeof(XPVFM)) },
971 SVt_PVIO, TRUE, NONV, HASARENA,
972 FIT_ARENA(24, sizeof(XPVIO)) },
975 #define new_body_allocated(sv_type) \
976 (void *)((char *)S_new_body(aTHX_ sv_type) \
977 - bodies_by_type[sv_type].offset)
979 /* return a thing to the free list */
981 #define del_body(thing, root) \
983 void ** const thing_copy = (void **)thing; \
984 *thing_copy = *root; \
985 *root = (void*)thing_copy; \
990 #define new_XNV() safemalloc(sizeof(XPVNV))
991 #define new_XPVNV() safemalloc(sizeof(XPVNV))
992 #define new_XPVMG() safemalloc(sizeof(XPVMG))
994 #define del_XPVGV(p) safefree(p)
998 #define new_XNV() new_body_allocated(SVt_NV)
999 #define new_XPVNV() new_body_allocated(SVt_PVNV)
1000 #define new_XPVMG() new_body_allocated(SVt_PVMG)
1002 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
1003 &PL_body_roots[SVt_PVGV])
1007 /* no arena for you! */
1009 #define new_NOARENA(details) \
1010 safemalloc((details)->body_size + (details)->offset)
1011 #define new_NOARENAZ(details) \
1012 safecalloc((details)->body_size + (details)->offset, 1)
1015 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1016 const size_t arena_size)
1019 void ** const root = &PL_body_roots[sv_type];
1020 struct arena_desc *adesc;
1021 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1025 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1026 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1027 static bool done_sanity_check;
1029 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1030 * variables like done_sanity_check. */
1031 if (!done_sanity_check) {
1032 unsigned int i = SVt_LAST;
1034 done_sanity_check = TRUE;
1037 assert (bodies_by_type[i].type == i);
1043 /* may need new arena-set to hold new arena */
1044 if (!aroot || aroot->curr >= aroot->set_size) {
1045 struct arena_set *newroot;
1046 Newxz(newroot, 1, struct arena_set);
1047 newroot->set_size = ARENAS_PER_SET;
1048 newroot->next = aroot;
1050 PL_body_arenas = (void *) newroot;
1051 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1054 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1055 curr = aroot->curr++;
1056 adesc = &(aroot->set[curr]);
1057 assert(!adesc->arena);
1059 Newx(adesc->arena, good_arena_size, char);
1060 adesc->size = good_arena_size;
1061 adesc->utype = sv_type;
1062 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1063 curr, (void*)adesc->arena, (UV)good_arena_size));
1065 start = (char *) adesc->arena;
1067 /* Get the address of the byte after the end of the last body we can fit.
1068 Remember, this is integer division: */
1069 end = start + good_arena_size / body_size * body_size;
1071 /* computed count doesn't reflect the 1st slot reservation */
1072 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1073 DEBUG_m(PerlIO_printf(Perl_debug_log,
1074 "arena %p end %p arena-size %d (from %d) type %d "
1076 (void*)start, (void*)end, (int)good_arena_size,
1077 (int)arena_size, sv_type, (int)body_size,
1078 (int)good_arena_size / (int)body_size));
1080 DEBUG_m(PerlIO_printf(Perl_debug_log,
1081 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1082 (void*)start, (void*)end,
1083 (int)arena_size, sv_type, (int)body_size,
1084 (int)good_arena_size / (int)body_size));
1086 *root = (void *)start;
1089 /* Where the next body would start: */
1090 char * const next = start + body_size;
1093 /* This is the last body: */
1094 assert(next == end);
1096 *(void **)start = 0;
1100 *(void**) start = (void *)next;
1105 /* grab a new thing from the free list, allocating more if necessary.
1106 The inline version is used for speed in hot routines, and the
1107 function using it serves the rest (unless PURIFY).
1109 #define new_body_inline(xpv, sv_type) \
1111 void ** const r3wt = &PL_body_roots[sv_type]; \
1112 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1113 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1114 bodies_by_type[sv_type].body_size,\
1115 bodies_by_type[sv_type].arena_size)); \
1116 *(r3wt) = *(void**)(xpv); \
1122 S_new_body(pTHX_ const svtype sv_type)
1126 new_body_inline(xpv, sv_type);
1132 static const struct body_details fake_rv =
1133 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1136 =for apidoc sv_upgrade
1138 Upgrade an SV to a more complex form. Generally adds a new body type to the
1139 SV, then copies across as much information as possible from the old body.
1140 It croaks if the SV is already in a more complex form than requested. You
1141 generally want to use the C<SvUPGRADE> macro wrapper, which checks the type
1142 before calling C<sv_upgrade>, and hence does not croak. See also
1149 Perl_sv_upgrade(pTHX_ SV *const sv, svtype new_type)
1154 const svtype old_type = SvTYPE(sv);
1155 const struct body_details *new_type_details;
1156 const struct body_details *old_type_details
1157 = bodies_by_type + old_type;
1158 SV *referant = NULL;
1160 PERL_ARGS_ASSERT_SV_UPGRADE;
1162 if (old_type == new_type)
1165 /* This clause was purposefully added ahead of the early return above to
1166 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1167 inference by Nick I-S that it would fix other troublesome cases. See
1168 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1170 Given that shared hash key scalars are no longer PVIV, but PV, there is
1171 no longer need to unshare so as to free up the IVX slot for its proper
1172 purpose. So it's safe to move the early return earlier. */
1174 if (new_type > SVt_PVMG && SvIsCOW(sv)) {
1175 sv_force_normal_flags(sv, 0);
1178 old_body = SvANY(sv);
1180 /* Copying structures onto other structures that have been neatly zeroed
1181 has a subtle gotcha. Consider XPVMG
1183 +------+------+------+------+------+-------+-------+
1184 | NV | CUR | LEN | IV | MAGIC | STASH |
1185 +------+------+------+------+------+-------+-------+
1186 0 4 8 12 16 20 24 28
1188 where NVs are aligned to 8 bytes, so that sizeof that structure is
1189 actually 32 bytes long, with 4 bytes of padding at the end:
1191 +------+------+------+------+------+-------+-------+------+
1192 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1193 +------+------+------+------+------+-------+-------+------+
1194 0 4 8 12 16 20 24 28 32
1196 so what happens if you allocate memory for this structure:
1198 +------+------+------+------+------+-------+-------+------+------+...
1199 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1200 +------+------+------+------+------+-------+-------+------+------+...
1201 0 4 8 12 16 20 24 28 32 36
1203 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1204 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1205 started out as zero once, but it's quite possible that it isn't. So now,
1206 rather than a nicely zeroed GP, you have it pointing somewhere random.
1209 (In fact, GP ends up pointing at a previous GP structure, because the
1210 principle cause of the padding in XPVMG getting garbage is a copy of
1211 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1212 this happens to be moot because XPVGV has been re-ordered, with GP
1213 no longer after STASH)
1215 So we are careful and work out the size of used parts of all the
1223 referant = SvRV(sv);
1224 old_type_details = &fake_rv;
1225 if (new_type == SVt_NV)
1226 new_type = SVt_PVNV;
1228 if (new_type < SVt_PVIV) {
1229 new_type = (new_type == SVt_NV)
1230 ? SVt_PVNV : SVt_PVIV;
1235 if (new_type < SVt_PVNV) {
1236 new_type = SVt_PVNV;
1240 assert(new_type > SVt_PV);
1241 assert(SVt_IV < SVt_PV);
1242 assert(SVt_NV < SVt_PV);
1249 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1250 there's no way that it can be safely upgraded, because perl.c
1251 expects to Safefree(SvANY(PL_mess_sv)) */
1252 assert(sv != PL_mess_sv);
1253 /* This flag bit is used to mean other things in other scalar types.
1254 Given that it only has meaning inside the pad, it shouldn't be set
1255 on anything that can get upgraded. */
1256 assert(!SvPAD_TYPED(sv));
1259 if (UNLIKELY(old_type_details->cant_upgrade))
1260 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1261 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1264 if (UNLIKELY(old_type > new_type))
1265 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1266 (int)old_type, (int)new_type);
1268 new_type_details = bodies_by_type + new_type;
1270 SvFLAGS(sv) &= ~SVTYPEMASK;
1271 SvFLAGS(sv) |= new_type;
1273 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1274 the return statements above will have triggered. */
1275 assert (new_type != SVt_NULL);
1278 assert(old_type == SVt_NULL);
1279 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1283 assert(old_type == SVt_NULL);
1284 SvANY(sv) = new_XNV();
1289 assert(new_type_details->body_size);
1292 assert(new_type_details->arena);
1293 assert(new_type_details->arena_size);
1294 /* This points to the start of the allocated area. */
1295 new_body_inline(new_body, new_type);
1296 Zero(new_body, new_type_details->body_size, char);
1297 new_body = ((char *)new_body) - new_type_details->offset;
1299 /* We always allocated the full length item with PURIFY. To do this
1300 we fake things so that arena is false for all 16 types.. */
1301 new_body = new_NOARENAZ(new_type_details);
1303 SvANY(sv) = new_body;
1304 if (new_type == SVt_PVAV) {
1308 if (old_type_details->body_size) {
1311 /* It will have been zeroed when the new body was allocated.
1312 Lets not write to it, in case it confuses a write-back
1318 #ifndef NODEFAULT_SHAREKEYS
1319 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1321 /* start with PERL_HASH_DEFAULT_HvMAX+1 buckets: */
1322 HvMAX(sv) = PERL_HASH_DEFAULT_HvMAX;
1325 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1326 The target created by newSVrv also is, and it can have magic.
1327 However, it never has SvPVX set.
1329 if (old_type == SVt_IV) {
1331 } else if (old_type >= SVt_PV) {
1332 assert(SvPVX_const(sv) == 0);
1335 if (old_type >= SVt_PVMG) {
1336 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1337 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1339 sv->sv_u.svu_array = NULL; /* or svu_hash */
1344 /* XXX Is this still needed? Was it ever needed? Surely as there is
1345 no route from NV to PVIV, NOK can never be true */
1346 assert(!SvNOKp(sv));
1359 assert(new_type_details->body_size);
1360 /* We always allocated the full length item with PURIFY. To do this
1361 we fake things so that arena is false for all 16 types.. */
1362 if(new_type_details->arena) {
1363 /* This points to the start of the allocated area. */
1364 new_body_inline(new_body, new_type);
1365 Zero(new_body, new_type_details->body_size, char);
1366 new_body = ((char *)new_body) - new_type_details->offset;
1368 new_body = new_NOARENAZ(new_type_details);
1370 SvANY(sv) = new_body;
1372 if (old_type_details->copy) {
1373 /* There is now the potential for an upgrade from something without
1374 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1375 int offset = old_type_details->offset;
1376 int length = old_type_details->copy;
1378 if (new_type_details->offset > old_type_details->offset) {
1379 const int difference
1380 = new_type_details->offset - old_type_details->offset;
1381 offset += difference;
1382 length -= difference;
1384 assert (length >= 0);
1386 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1390 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1391 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1392 * correct 0.0 for us. Otherwise, if the old body didn't have an
1393 * NV slot, but the new one does, then we need to initialise the
1394 * freshly created NV slot with whatever the correct bit pattern is
1396 if (old_type_details->zero_nv && !new_type_details->zero_nv
1397 && !isGV_with_GP(sv))
1401 if (UNLIKELY(new_type == SVt_PVIO)) {
1402 IO * const io = MUTABLE_IO(sv);
1403 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1406 /* Clear the stashcache because a new IO could overrule a package
1408 DEBUG_o(Perl_deb(aTHX_ "sv_upgrade clearing PL_stashcache\n"));
1409 hv_clear(PL_stashcache);
1411 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1412 IoPAGE_LEN(sv) = 60;
1414 if (UNLIKELY(new_type == SVt_REGEXP))
1415 sv->sv_u.svu_rx = (regexp *)new_body;
1416 else if (old_type < SVt_PV) {
1417 /* referant will be NULL unless the old type was SVt_IV emulating
1419 sv->sv_u.svu_rv = referant;
1423 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1424 (unsigned long)new_type);
1427 if (old_type > SVt_IV) {
1431 /* Note that there is an assumption that all bodies of types that
1432 can be upgraded came from arenas. Only the more complex non-
1433 upgradable types are allowed to be directly malloc()ed. */
1434 assert(old_type_details->arena);
1435 del_body((void*)((char*)old_body + old_type_details->offset),
1436 &PL_body_roots[old_type]);
1442 =for apidoc sv_backoff
1444 Remove any string offset. You should normally use the C<SvOOK_off> macro
1451 Perl_sv_backoff(pTHX_ SV *const sv)
1454 const char * const s = SvPVX_const(sv);
1456 PERL_ARGS_ASSERT_SV_BACKOFF;
1457 PERL_UNUSED_CONTEXT;
1460 assert(SvTYPE(sv) != SVt_PVHV);
1461 assert(SvTYPE(sv) != SVt_PVAV);
1463 SvOOK_offset(sv, delta);
1465 SvLEN_set(sv, SvLEN(sv) + delta);
1466 SvPV_set(sv, SvPVX(sv) - delta);
1467 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1468 SvFLAGS(sv) &= ~SVf_OOK;
1475 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1476 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1477 Use the C<SvGROW> wrapper instead.
1482 static void S_sv_uncow(pTHX_ SV * const sv, const U32 flags);
1485 Perl_sv_grow(pTHX_ SV *const sv, STRLEN newlen)
1489 PERL_ARGS_ASSERT_SV_GROW;
1493 if (SvTYPE(sv) < SVt_PV) {
1494 sv_upgrade(sv, SVt_PV);
1495 s = SvPVX_mutable(sv);
1497 else if (SvOOK(sv)) { /* pv is offset? */
1499 s = SvPVX_mutable(sv);
1500 if (newlen > SvLEN(sv))
1501 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1505 if (SvIsCOW(sv)) S_sv_uncow(aTHX_ sv, 0);
1506 s = SvPVX_mutable(sv);
1509 #ifdef PERL_NEW_COPY_ON_WRITE
1510 /* the new COW scheme uses SvPVX(sv)[SvLEN(sv)-1] (if spare)
1511 * to store the COW count. So in general, allocate one more byte than
1512 * asked for, to make it likely this byte is always spare: and thus
1513 * make more strings COW-able.
1514 * If the new size is a big power of two, don't bother: we assume the
1515 * caller wanted a nice 2^N sized block and will be annoyed at getting
1521 if (newlen > SvLEN(sv)) { /* need more room? */
1522 STRLEN minlen = SvCUR(sv);
1523 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1524 if (newlen < minlen)
1526 #ifndef Perl_safesysmalloc_size
1527 newlen = PERL_STRLEN_ROUNDUP(newlen);
1529 if (SvLEN(sv) && s) {
1530 s = (char*)saferealloc(s, newlen);
1533 s = (char*)safemalloc(newlen);
1534 if (SvPVX_const(sv) && SvCUR(sv)) {
1535 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1539 #ifdef Perl_safesysmalloc_size
1540 /* Do this here, do it once, do it right, and then we will never get
1541 called back into sv_grow() unless there really is some growing
1543 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1545 SvLEN_set(sv, newlen);
1552 =for apidoc sv_setiv
1554 Copies an integer into the given SV, upgrading first if necessary.
1555 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1561 Perl_sv_setiv(pTHX_ SV *const sv, const IV i)
1565 PERL_ARGS_ASSERT_SV_SETIV;
1567 SV_CHECK_THINKFIRST_COW_DROP(sv);
1568 switch (SvTYPE(sv)) {
1571 sv_upgrade(sv, SVt_IV);
1574 sv_upgrade(sv, SVt_PVIV);
1578 if (!isGV_with_GP(sv))
1585 /* diag_listed_as: Can't coerce %s to %s in %s */
1586 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1590 (void)SvIOK_only(sv); /* validate number */
1596 =for apidoc sv_setiv_mg
1598 Like C<sv_setiv>, but also handles 'set' magic.
1604 Perl_sv_setiv_mg(pTHX_ SV *const sv, const IV i)
1606 PERL_ARGS_ASSERT_SV_SETIV_MG;
1613 =for apidoc sv_setuv
1615 Copies an unsigned integer into the given SV, upgrading first if necessary.
1616 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1622 Perl_sv_setuv(pTHX_ SV *const sv, const UV u)
1624 PERL_ARGS_ASSERT_SV_SETUV;
1626 /* With the if statement to ensure that integers are stored as IVs whenever
1628 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1631 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1633 If you wish to remove the following if statement, so that this routine
1634 (and its callers) always return UVs, please benchmark to see what the
1635 effect is. Modern CPUs may be different. Or may not :-)
1637 if (u <= (UV)IV_MAX) {
1638 sv_setiv(sv, (IV)u);
1647 =for apidoc sv_setuv_mg
1649 Like C<sv_setuv>, but also handles 'set' magic.
1655 Perl_sv_setuv_mg(pTHX_ SV *const sv, const UV u)
1657 PERL_ARGS_ASSERT_SV_SETUV_MG;
1664 =for apidoc sv_setnv
1666 Copies a double into the given SV, upgrading first if necessary.
1667 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1673 Perl_sv_setnv(pTHX_ SV *const sv, const NV num)
1677 PERL_ARGS_ASSERT_SV_SETNV;
1679 SV_CHECK_THINKFIRST_COW_DROP(sv);
1680 switch (SvTYPE(sv)) {
1683 sv_upgrade(sv, SVt_NV);
1687 sv_upgrade(sv, SVt_PVNV);
1691 if (!isGV_with_GP(sv))
1698 /* diag_listed_as: Can't coerce %s to %s in %s */
1699 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1704 (void)SvNOK_only(sv); /* validate number */
1709 =for apidoc sv_setnv_mg
1711 Like C<sv_setnv>, but also handles 'set' magic.
1717 Perl_sv_setnv_mg(pTHX_ SV *const sv, const NV num)
1719 PERL_ARGS_ASSERT_SV_SETNV_MG;
1725 /* Return a cleaned-up, printable version of sv, for non-numeric, or
1726 * not incrementable warning display.
1727 * Originally part of S_not_a_number().
1728 * The return value may be != tmpbuf.
1732 S_sv_display(pTHX_ SV *const sv, char *tmpbuf, STRLEN tmpbuf_size) {
1735 PERL_ARGS_ASSERT_SV_DISPLAY;
1738 SV *dsv = newSVpvs_flags("", SVs_TEMP);
1739 pv = sv_uni_display(dsv, sv, 10, UNI_DISPLAY_ISPRINT);
1742 const char * const limit = tmpbuf + tmpbuf_size - 8;
1743 /* each *s can expand to 4 chars + "...\0",
1744 i.e. need room for 8 chars */
1746 const char *s = SvPVX_const(sv);
1747 const char * const end = s + SvCUR(sv);
1748 for ( ; s < end && d < limit; s++ ) {
1750 if (! isASCII(ch) && !isPRINT_LC(ch)) {
1754 /* Map to ASCII "equivalent" of Latin1 */
1755 ch = LATIN1_TO_NATIVE(NATIVE_TO_LATIN1(ch) & 127);
1761 else if (ch == '\r') {
1765 else if (ch == '\f') {
1769 else if (ch == '\\') {
1773 else if (ch == '\0') {
1777 else if (isPRINT_LC(ch))
1796 /* Print an "isn't numeric" warning, using a cleaned-up,
1797 * printable version of the offending string
1801 S_not_a_number(pTHX_ SV *const sv)
1807 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1809 pv = sv_display(sv, tmpbuf, sizeof(tmpbuf));
1812 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1813 /* diag_listed_as: Argument "%s" isn't numeric%s */
1814 "Argument \"%s\" isn't numeric in %s", pv,
1817 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1818 /* diag_listed_as: Argument "%s" isn't numeric%s */
1819 "Argument \"%s\" isn't numeric", pv);
1823 S_not_incrementable(pTHX_ SV *const sv) {
1828 PERL_ARGS_ASSERT_NOT_INCREMENTABLE;
1830 pv = sv_display(sv, tmpbuf, sizeof(tmpbuf));
1832 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1833 "Argument \"%s\" treated as 0 in increment (++)", pv);
1837 =for apidoc looks_like_number
1839 Test if the content of an SV looks like a number (or is a number).
1840 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1841 non-numeric warning), even if your atof() doesn't grok them. Get-magic is
1848 Perl_looks_like_number(pTHX_ SV *const sv)
1853 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1855 if (SvPOK(sv) || SvPOKp(sv)) {
1856 sbegin = SvPV_nomg_const(sv, len);
1859 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1860 return grok_number(sbegin, len, NULL);
1864 S_glob_2number(pTHX_ GV * const gv)
1866 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1868 /* We know that all GVs stringify to something that is not-a-number,
1869 so no need to test that. */
1870 if (ckWARN(WARN_NUMERIC))
1872 SV *const buffer = sv_newmortal();
1873 gv_efullname3(buffer, gv, "*");
1874 not_a_number(buffer);
1876 /* We just want something true to return, so that S_sv_2iuv_common
1877 can tail call us and return true. */
1881 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1882 until proven guilty, assume that things are not that bad... */
1887 As 64 bit platforms often have an NV that doesn't preserve all bits of
1888 an IV (an assumption perl has been based on to date) it becomes necessary
1889 to remove the assumption that the NV always carries enough precision to
1890 recreate the IV whenever needed, and that the NV is the canonical form.
1891 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1892 precision as a side effect of conversion (which would lead to insanity
1893 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1894 1) to distinguish between IV/UV/NV slots that have cached a valid
1895 conversion where precision was lost and IV/UV/NV slots that have a
1896 valid conversion which has lost no precision
1897 2) to ensure that if a numeric conversion to one form is requested that
1898 would lose precision, the precise conversion (or differently
1899 imprecise conversion) is also performed and cached, to prevent
1900 requests for different numeric formats on the same SV causing
1901 lossy conversion chains. (lossless conversion chains are perfectly
1906 SvIOKp is true if the IV slot contains a valid value
1907 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1908 SvNOKp is true if the NV slot contains a valid value
1909 SvNOK is true only if the NV value is accurate
1912 while converting from PV to NV, check to see if converting that NV to an
1913 IV(or UV) would lose accuracy over a direct conversion from PV to
1914 IV(or UV). If it would, cache both conversions, return NV, but mark
1915 SV as IOK NOKp (ie not NOK).
1917 While converting from PV to IV, check to see if converting that IV to an
1918 NV would lose accuracy over a direct conversion from PV to NV. If it
1919 would, cache both conversions, flag similarly.
1921 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1922 correctly because if IV & NV were set NV *always* overruled.
1923 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1924 changes - now IV and NV together means that the two are interchangeable:
1925 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1927 The benefit of this is that operations such as pp_add know that if
1928 SvIOK is true for both left and right operands, then integer addition
1929 can be used instead of floating point (for cases where the result won't
1930 overflow). Before, floating point was always used, which could lead to
1931 loss of precision compared with integer addition.
1933 * making IV and NV equal status should make maths accurate on 64 bit
1935 * may speed up maths somewhat if pp_add and friends start to use
1936 integers when possible instead of fp. (Hopefully the overhead in
1937 looking for SvIOK and checking for overflow will not outweigh the
1938 fp to integer speedup)
1939 * will slow down integer operations (callers of SvIV) on "inaccurate"
1940 values, as the change from SvIOK to SvIOKp will cause a call into
1941 sv_2iv each time rather than a macro access direct to the IV slot
1942 * should speed up number->string conversion on integers as IV is
1943 favoured when IV and NV are equally accurate
1945 ####################################################################
1946 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1947 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1948 On the other hand, SvUOK is true iff UV.
1949 ####################################################################
1951 Your mileage will vary depending your CPU's relative fp to integer
1955 #ifndef NV_PRESERVES_UV
1956 # define IS_NUMBER_UNDERFLOW_IV 1
1957 # define IS_NUMBER_UNDERFLOW_UV 2
1958 # define IS_NUMBER_IV_AND_UV 2
1959 # define IS_NUMBER_OVERFLOW_IV 4
1960 # define IS_NUMBER_OVERFLOW_UV 5
1962 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1964 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1966 S_sv_2iuv_non_preserve(pTHX_ SV *const sv
1974 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1976 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));
1977 if (SvNVX(sv) < (NV)IV_MIN) {
1978 (void)SvIOKp_on(sv);
1980 SvIV_set(sv, IV_MIN);
1981 return IS_NUMBER_UNDERFLOW_IV;
1983 if (SvNVX(sv) > (NV)UV_MAX) {
1984 (void)SvIOKp_on(sv);
1987 SvUV_set(sv, UV_MAX);
1988 return IS_NUMBER_OVERFLOW_UV;
1990 (void)SvIOKp_on(sv);
1992 /* Can't use strtol etc to convert this string. (See truth table in
1994 if (SvNVX(sv) <= (UV)IV_MAX) {
1995 SvIV_set(sv, I_V(SvNVX(sv)));
1996 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1997 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1999 /* Integer is imprecise. NOK, IOKp */
2001 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
2004 SvUV_set(sv, U_V(SvNVX(sv)));
2005 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2006 if (SvUVX(sv) == UV_MAX) {
2007 /* As we know that NVs don't preserve UVs, UV_MAX cannot
2008 possibly be preserved by NV. Hence, it must be overflow.
2010 return IS_NUMBER_OVERFLOW_UV;
2012 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
2014 /* Integer is imprecise. NOK, IOKp */
2016 return IS_NUMBER_OVERFLOW_IV;
2018 #endif /* !NV_PRESERVES_UV*/
2021 S_sv_2iuv_common(pTHX_ SV *const sv)
2025 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
2028 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
2029 * without also getting a cached IV/UV from it at the same time
2030 * (ie PV->NV conversion should detect loss of accuracy and cache
2031 * IV or UV at same time to avoid this. */
2032 /* IV-over-UV optimisation - choose to cache IV if possible */
2034 if (SvTYPE(sv) == SVt_NV)
2035 sv_upgrade(sv, SVt_PVNV);
2037 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
2038 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
2039 certainly cast into the IV range at IV_MAX, whereas the correct
2040 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
2042 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2043 if (Perl_isnan(SvNVX(sv))) {
2049 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2050 SvIV_set(sv, I_V(SvNVX(sv)));
2051 if (SvNVX(sv) == (NV) SvIVX(sv)
2052 #ifndef NV_PRESERVES_UV
2053 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2054 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2055 /* Don't flag it as "accurately an integer" if the number
2056 came from a (by definition imprecise) NV operation, and
2057 we're outside the range of NV integer precision */
2061 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2063 /* scalar has trailing garbage, eg "42a" */
2065 DEBUG_c(PerlIO_printf(Perl_debug_log,
2066 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2072 /* IV not precise. No need to convert from PV, as NV
2073 conversion would already have cached IV if it detected
2074 that PV->IV would be better than PV->NV->IV
2075 flags already correct - don't set public IOK. */
2076 DEBUG_c(PerlIO_printf(Perl_debug_log,
2077 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2082 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2083 but the cast (NV)IV_MIN rounds to a the value less (more
2084 negative) than IV_MIN which happens to be equal to SvNVX ??
2085 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2086 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2087 (NV)UVX == NVX are both true, but the values differ. :-(
2088 Hopefully for 2s complement IV_MIN is something like
2089 0x8000000000000000 which will be exact. NWC */
2092 SvUV_set(sv, U_V(SvNVX(sv)));
2094 (SvNVX(sv) == (NV) SvUVX(sv))
2095 #ifndef NV_PRESERVES_UV
2096 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2097 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2098 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2099 /* Don't flag it as "accurately an integer" if the number
2100 came from a (by definition imprecise) NV operation, and
2101 we're outside the range of NV integer precision */
2107 DEBUG_c(PerlIO_printf(Perl_debug_log,
2108 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2114 else if (SvPOKp(sv)) {
2116 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2117 /* We want to avoid a possible problem when we cache an IV/ a UV which
2118 may be later translated to an NV, and the resulting NV is not
2119 the same as the direct translation of the initial string
2120 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2121 be careful to ensure that the value with the .456 is around if the
2122 NV value is requested in the future).
2124 This means that if we cache such an IV/a UV, we need to cache the
2125 NV as well. Moreover, we trade speed for space, and do not
2126 cache the NV if we are sure it's not needed.
2129 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2130 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2131 == IS_NUMBER_IN_UV) {
2132 /* It's definitely an integer, only upgrade to PVIV */
2133 if (SvTYPE(sv) < SVt_PVIV)
2134 sv_upgrade(sv, SVt_PVIV);
2136 } else if (SvTYPE(sv) < SVt_PVNV)
2137 sv_upgrade(sv, SVt_PVNV);
2139 /* If NVs preserve UVs then we only use the UV value if we know that
2140 we aren't going to call atof() below. If NVs don't preserve UVs
2141 then the value returned may have more precision than atof() will
2142 return, even though value isn't perfectly accurate. */
2143 if ((numtype & (IS_NUMBER_IN_UV
2144 #ifdef NV_PRESERVES_UV
2147 )) == IS_NUMBER_IN_UV) {
2148 /* This won't turn off the public IOK flag if it was set above */
2149 (void)SvIOKp_on(sv);
2151 if (!(numtype & IS_NUMBER_NEG)) {
2153 if (value <= (UV)IV_MAX) {
2154 SvIV_set(sv, (IV)value);
2156 /* it didn't overflow, and it was positive. */
2157 SvUV_set(sv, value);
2161 /* 2s complement assumption */
2162 if (value <= (UV)IV_MIN) {
2163 SvIV_set(sv, -(IV)value);
2165 /* Too negative for an IV. This is a double upgrade, but
2166 I'm assuming it will be rare. */
2167 if (SvTYPE(sv) < SVt_PVNV)
2168 sv_upgrade(sv, SVt_PVNV);
2172 SvNV_set(sv, -(NV)value);
2173 SvIV_set(sv, IV_MIN);
2177 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2178 will be in the previous block to set the IV slot, and the next
2179 block to set the NV slot. So no else here. */
2181 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2182 != IS_NUMBER_IN_UV) {
2183 /* It wasn't an (integer that doesn't overflow the UV). */
2184 SvNV_set(sv, Atof(SvPVX_const(sv)));
2186 if (! numtype && ckWARN(WARN_NUMERIC))
2189 #if defined(USE_LONG_DOUBLE)
2190 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2191 PTR2UV(sv), SvNVX(sv)));
2193 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2194 PTR2UV(sv), SvNVX(sv)));
2197 #ifdef NV_PRESERVES_UV
2198 (void)SvIOKp_on(sv);
2200 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2201 SvIV_set(sv, I_V(SvNVX(sv)));
2202 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2205 NOOP; /* Integer is imprecise. NOK, IOKp */
2207 /* UV will not work better than IV */
2209 if (SvNVX(sv) > (NV)UV_MAX) {
2211 /* Integer is inaccurate. NOK, IOKp, is UV */
2212 SvUV_set(sv, UV_MAX);
2214 SvUV_set(sv, U_V(SvNVX(sv)));
2215 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2216 NV preservse UV so can do correct comparison. */
2217 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2220 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2225 #else /* NV_PRESERVES_UV */
2226 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2227 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2228 /* The IV/UV slot will have been set from value returned by
2229 grok_number above. The NV slot has just been set using
2232 assert (SvIOKp(sv));
2234 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2235 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2236 /* Small enough to preserve all bits. */
2237 (void)SvIOKp_on(sv);
2239 SvIV_set(sv, I_V(SvNVX(sv)));
2240 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2242 /* Assumption: first non-preserved integer is < IV_MAX,
2243 this NV is in the preserved range, therefore: */
2244 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2246 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);
2250 0 0 already failed to read UV.
2251 0 1 already failed to read UV.
2252 1 0 you won't get here in this case. IV/UV
2253 slot set, public IOK, Atof() unneeded.
2254 1 1 already read UV.
2255 so there's no point in sv_2iuv_non_preserve() attempting
2256 to use atol, strtol, strtoul etc. */
2258 sv_2iuv_non_preserve (sv, numtype);
2260 sv_2iuv_non_preserve (sv);
2264 #endif /* NV_PRESERVES_UV */
2265 /* It might be more code efficient to go through the entire logic above
2266 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2267 gets complex and potentially buggy, so more programmer efficient
2268 to do it this way, by turning off the public flags: */
2270 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2274 if (isGV_with_GP(sv))
2275 return glob_2number(MUTABLE_GV(sv));
2277 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2279 if (SvTYPE(sv) < SVt_IV)
2280 /* Typically the caller expects that sv_any is not NULL now. */
2281 sv_upgrade(sv, SVt_IV);
2282 /* Return 0 from the caller. */
2289 =for apidoc sv_2iv_flags
2291 Return the integer value of an SV, doing any necessary string
2292 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2293 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2299 Perl_sv_2iv_flags(pTHX_ SV *const sv, const I32 flags)
2303 PERL_ARGS_ASSERT_SV_2IV_FLAGS;
2305 assert (SvTYPE(sv) != SVt_PVAV && SvTYPE(sv) != SVt_PVHV
2306 && SvTYPE(sv) != SVt_PVFM);
2308 if (SvGMAGICAL(sv) && (flags & SV_GMAGIC))
2314 if (flags & SV_SKIP_OVERLOAD)
2316 tmpstr = AMG_CALLunary(sv, numer_amg);
2317 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2318 return SvIV(tmpstr);
2321 return PTR2IV(SvRV(sv));
2324 if (SvVALID(sv) || isREGEXP(sv)) {
2325 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2326 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2327 In practice they are extremely unlikely to actually get anywhere
2328 accessible by user Perl code - the only way that I'm aware of is when
2329 a constant subroutine which is used as the second argument to index.
2331 Regexps have no SvIVX and SvNVX fields.
2333 assert(isREGEXP(sv) || SvPOKp(sv));
2336 const char * const ptr =
2337 isREGEXP(sv) ? RX_WRAPPED((REGEXP*)sv) : SvPVX_const(sv);
2339 = grok_number(ptr, SvCUR(sv), &value);
2341 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2342 == IS_NUMBER_IN_UV) {
2343 /* It's definitely an integer */
2344 if (numtype & IS_NUMBER_NEG) {
2345 if (value < (UV)IV_MIN)
2348 if (value < (UV)IV_MAX)
2353 if (ckWARN(WARN_NUMERIC))
2356 return I_V(Atof(ptr));
2360 if (SvTHINKFIRST(sv)) {
2361 #ifdef PERL_OLD_COPY_ON_WRITE
2363 sv_force_normal_flags(sv, 0);
2366 if (SvREADONLY(sv) && !SvOK(sv)) {
2367 if (ckWARN(WARN_UNINITIALIZED))
2374 if (S_sv_2iuv_common(aTHX_ sv))
2378 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2379 PTR2UV(sv),SvIVX(sv)));
2380 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2384 =for apidoc sv_2uv_flags
2386 Return the unsigned integer value of an SV, doing any necessary string
2387 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2388 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2394 Perl_sv_2uv_flags(pTHX_ SV *const sv, const I32 flags)
2398 PERL_ARGS_ASSERT_SV_2UV_FLAGS;
2400 if (SvGMAGICAL(sv) && (flags & SV_GMAGIC))
2406 if (flags & SV_SKIP_OVERLOAD)
2408 tmpstr = AMG_CALLunary(sv, numer_amg);
2409 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2410 return SvUV(tmpstr);
2413 return PTR2UV(SvRV(sv));
2416 if (SvVALID(sv) || isREGEXP(sv)) {
2417 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2418 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2419 Regexps have no SvIVX and SvNVX fields. */
2420 assert(isREGEXP(sv) || SvPOKp(sv));
2423 const char * const ptr =
2424 isREGEXP(sv) ? RX_WRAPPED((REGEXP*)sv) : SvPVX_const(sv);
2426 = grok_number(ptr, SvCUR(sv), &value);
2428 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2429 == IS_NUMBER_IN_UV) {
2430 /* It's definitely an integer */
2431 if (!(numtype & IS_NUMBER_NEG))
2435 if (ckWARN(WARN_NUMERIC))
2438 return U_V(Atof(ptr));
2442 if (SvTHINKFIRST(sv)) {
2443 #ifdef PERL_OLD_COPY_ON_WRITE
2445 sv_force_normal_flags(sv, 0);
2448 if (SvREADONLY(sv) && !SvOK(sv)) {
2449 if (ckWARN(WARN_UNINITIALIZED))
2456 if (S_sv_2iuv_common(aTHX_ sv))
2460 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2461 PTR2UV(sv),SvUVX(sv)));
2462 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2466 =for apidoc sv_2nv_flags
2468 Return the num value of an SV, doing any necessary string or integer
2469 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2470 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2476 Perl_sv_2nv_flags(pTHX_ SV *const sv, const I32 flags)
2480 PERL_ARGS_ASSERT_SV_2NV_FLAGS;
2482 assert (SvTYPE(sv) != SVt_PVAV && SvTYPE(sv) != SVt_PVHV
2483 && SvTYPE(sv) != SVt_PVFM);
2484 if (SvGMAGICAL(sv) || SvVALID(sv) || isREGEXP(sv)) {
2485 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2486 the same flag bit as SVf_IVisUV, so must not let them cache NVs.
2487 Regexps have no SvIVX and SvNVX fields. */
2489 if (flags & SV_GMAGIC)
2493 if (SvPOKp(sv) && !SvIOKp(sv)) {
2494 ptr = SvPVX_const(sv);
2496 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2497 !grok_number(ptr, SvCUR(sv), NULL))
2503 return (NV)SvUVX(sv);
2505 return (NV)SvIVX(sv);
2511 ptr = RX_WRAPPED((REGEXP *)sv);
2514 assert(SvTYPE(sv) >= SVt_PVMG);
2515 /* This falls through to the report_uninit near the end of the
2517 } else if (SvTHINKFIRST(sv)) {
2522 if (flags & SV_SKIP_OVERLOAD)
2524 tmpstr = AMG_CALLunary(sv, numer_amg);
2525 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2526 return SvNV(tmpstr);
2529 return PTR2NV(SvRV(sv));
2531 #ifdef PERL_OLD_COPY_ON_WRITE
2533 sv_force_normal_flags(sv, 0);
2536 if (SvREADONLY(sv) && !SvOK(sv)) {
2537 if (ckWARN(WARN_UNINITIALIZED))
2542 if (SvTYPE(sv) < SVt_NV) {
2543 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2544 sv_upgrade(sv, SVt_NV);
2545 #ifdef USE_LONG_DOUBLE
2547 STORE_NUMERIC_LOCAL_SET_STANDARD();
2548 PerlIO_printf(Perl_debug_log,
2549 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2550 PTR2UV(sv), SvNVX(sv));
2551 RESTORE_NUMERIC_LOCAL();
2555 STORE_NUMERIC_LOCAL_SET_STANDARD();
2556 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2557 PTR2UV(sv), SvNVX(sv));
2558 RESTORE_NUMERIC_LOCAL();
2562 else if (SvTYPE(sv) < SVt_PVNV)
2563 sv_upgrade(sv, SVt_PVNV);
2568 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2569 #ifdef NV_PRESERVES_UV
2575 /* Only set the public NV OK flag if this NV preserves the IV */
2576 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2578 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2579 : (SvIVX(sv) == I_V(SvNVX(sv))))
2585 else if (SvPOKp(sv)) {
2587 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2588 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2590 #ifdef NV_PRESERVES_UV
2591 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2592 == IS_NUMBER_IN_UV) {
2593 /* It's definitely an integer */
2594 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2596 SvNV_set(sv, Atof(SvPVX_const(sv)));
2602 SvNV_set(sv, Atof(SvPVX_const(sv)));
2603 /* Only set the public NV OK flag if this NV preserves the value in
2604 the PV at least as well as an IV/UV would.
2605 Not sure how to do this 100% reliably. */
2606 /* if that shift count is out of range then Configure's test is
2607 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2609 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2610 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2611 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2612 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2613 /* Can't use strtol etc to convert this string, so don't try.
2614 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2617 /* value has been set. It may not be precise. */
2618 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2619 /* 2s complement assumption for (UV)IV_MIN */
2620 SvNOK_on(sv); /* Integer is too negative. */
2625 if (numtype & IS_NUMBER_NEG) {
2626 SvIV_set(sv, -(IV)value);
2627 } else if (value <= (UV)IV_MAX) {
2628 SvIV_set(sv, (IV)value);
2630 SvUV_set(sv, value);
2634 if (numtype & IS_NUMBER_NOT_INT) {
2635 /* I believe that even if the original PV had decimals,
2636 they are lost beyond the limit of the FP precision.
2637 However, neither is canonical, so both only get p
2638 flags. NWC, 2000/11/25 */
2639 /* Both already have p flags, so do nothing */
2641 const NV nv = SvNVX(sv);
2642 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2643 if (SvIVX(sv) == I_V(nv)) {
2646 /* It had no "." so it must be integer. */
2650 /* between IV_MAX and NV(UV_MAX).
2651 Could be slightly > UV_MAX */
2653 if (numtype & IS_NUMBER_NOT_INT) {
2654 /* UV and NV both imprecise. */
2656 const UV nv_as_uv = U_V(nv);
2658 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2667 /* It might be more code efficient to go through the entire logic above
2668 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2669 gets complex and potentially buggy, so more programmer efficient
2670 to do it this way, by turning off the public flags: */
2672 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2673 #endif /* NV_PRESERVES_UV */
2676 if (isGV_with_GP(sv)) {
2677 glob_2number(MUTABLE_GV(sv));
2681 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2683 assert (SvTYPE(sv) >= SVt_NV);
2684 /* Typically the caller expects that sv_any is not NULL now. */
2685 /* XXX Ilya implies that this is a bug in callers that assume this
2686 and ideally should be fixed. */
2689 #if defined(USE_LONG_DOUBLE)
2691 STORE_NUMERIC_LOCAL_SET_STANDARD();
2692 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2693 PTR2UV(sv), SvNVX(sv));
2694 RESTORE_NUMERIC_LOCAL();
2698 STORE_NUMERIC_LOCAL_SET_STANDARD();
2699 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2700 PTR2UV(sv), SvNVX(sv));
2701 RESTORE_NUMERIC_LOCAL();
2710 Return an SV with the numeric value of the source SV, doing any necessary
2711 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2712 access this function.
2718 Perl_sv_2num(pTHX_ SV *const sv)
2720 PERL_ARGS_ASSERT_SV_2NUM;
2725 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2726 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2727 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2728 return sv_2num(tmpsv);
2730 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2733 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2734 * UV as a string towards the end of buf, and return pointers to start and
2737 * We assume that buf is at least TYPE_CHARS(UV) long.
2741 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2743 char *ptr = buf + TYPE_CHARS(UV);
2744 char * const ebuf = ptr;
2747 PERL_ARGS_ASSERT_UIV_2BUF;
2759 *--ptr = '0' + (char)(uv % 10);
2768 =for apidoc sv_2pv_flags
2770 Returns a pointer to the string value of an SV, and sets *lp to its length.
2771 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a
2772 string if necessary. Normally invoked via the C<SvPV_flags> macro.
2773 C<sv_2pv()> and C<sv_2pv_nomg> usually end up here too.
2779 Perl_sv_2pv_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
2784 PERL_ARGS_ASSERT_SV_2PV_FLAGS;
2786 assert (SvTYPE(sv) != SVt_PVAV && SvTYPE(sv) != SVt_PVHV
2787 && SvTYPE(sv) != SVt_PVFM);
2788 if (SvGMAGICAL(sv) && (flags & SV_GMAGIC))
2793 if (flags & SV_SKIP_OVERLOAD)
2795 tmpstr = AMG_CALLunary(sv, string_amg);
2796 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2797 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2799 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2803 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2804 if (flags & SV_CONST_RETURN) {
2805 pv = (char *) SvPVX_const(tmpstr);
2807 pv = (flags & SV_MUTABLE_RETURN)
2808 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2811 *lp = SvCUR(tmpstr);
2813 pv = sv_2pv_flags(tmpstr, lp, flags);
2826 SV *const referent = SvRV(sv);
2830 retval = buffer = savepvn("NULLREF", len);
2831 } else if (SvTYPE(referent) == SVt_REGEXP &&
2832 (!(PL_curcop->cop_hints & HINT_NO_AMAGIC) ||
2833 amagic_is_enabled(string_amg))) {
2834 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2838 /* If the regex is UTF-8 we want the containing scalar to
2839 have an UTF-8 flag too */
2846 *lp = RX_WRAPLEN(re);
2848 return RX_WRAPPED(re);
2850 const char *const typestr = sv_reftype(referent, 0);
2851 const STRLEN typelen = strlen(typestr);
2852 UV addr = PTR2UV(referent);
2853 const char *stashname = NULL;
2854 STRLEN stashnamelen = 0; /* hush, gcc */
2855 const char *buffer_end;
2857 if (SvOBJECT(referent)) {
2858 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2861 stashname = HEK_KEY(name);
2862 stashnamelen = HEK_LEN(name);
2864 if (HEK_UTF8(name)) {
2870 stashname = "__ANON__";
2873 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2874 + 2 * sizeof(UV) + 2 /* )\0 */;
2876 len = typelen + 3 /* (0x */
2877 + 2 * sizeof(UV) + 2 /* )\0 */;
2880 Newx(buffer, len, char);
2881 buffer_end = retval = buffer + len;
2883 /* Working backwards */
2887 *--retval = PL_hexdigit[addr & 15];
2888 } while (addr >>= 4);
2894 memcpy(retval, typestr, typelen);
2898 retval -= stashnamelen;
2899 memcpy(retval, stashname, stashnamelen);
2901 /* retval may not necessarily have reached the start of the
2903 assert (retval >= buffer);
2905 len = buffer_end - retval - 1; /* -1 for that \0 */
2917 if (flags & SV_MUTABLE_RETURN)
2918 return SvPVX_mutable(sv);
2919 if (flags & SV_CONST_RETURN)
2920 return (char *)SvPVX_const(sv);
2925 /* I'm assuming that if both IV and NV are equally valid then
2926 converting the IV is going to be more efficient */
2927 const U32 isUIOK = SvIsUV(sv);
2928 char buf[TYPE_CHARS(UV)];
2932 if (SvTYPE(sv) < SVt_PVIV)
2933 sv_upgrade(sv, SVt_PVIV);
2934 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2936 /* inlined from sv_setpvn */
2937 s = SvGROW_mutable(sv, len + 1);
2938 Move(ptr, s, len, char);
2943 else if (SvNOK(sv)) {
2944 if (SvTYPE(sv) < SVt_PVNV)
2945 sv_upgrade(sv, SVt_PVNV);
2946 if (SvNVX(sv) == 0.0) {
2947 s = SvGROW_mutable(sv, 2);
2952 /* The +20 is pure guesswork. Configure test needed. --jhi */
2953 s = SvGROW_mutable(sv, NV_DIG + 20);
2954 /* some Xenix systems wipe out errno here */
2956 #ifndef USE_LOCALE_NUMERIC
2957 V_Gconvert(SvNVX(sv), NV_DIG, 0, s);
2960 /* Gconvert always uses the current locale. That's the right thing
2961 * to do if we're supposed to be using locales. But otherwise, we
2962 * want the result to be based on the C locale, so we need to
2963 * change to the C locale during the Gconvert and then change back.
2964 * But if we're already in the C locale (PL_numeric_standard is
2965 * TRUE in that case), no need to do any changing */
2966 if (PL_numeric_standard || IN_SOME_LOCALE_FORM_RUNTIME) {
2967 V_Gconvert(SvNVX(sv), NV_DIG, 0, s);
2969 /* If the radix character is UTF-8, and actually is in the
2970 * output, turn on the UTF-8 flag for the scalar */
2971 if (! PL_numeric_standard
2972 && PL_numeric_radix_sv && SvUTF8(PL_numeric_radix_sv)
2973 && instr(s, SvPVX_const(PL_numeric_radix_sv)))
2979 char *loc = savepv(setlocale(LC_NUMERIC, NULL));
2980 setlocale(LC_NUMERIC, "C");
2981 V_Gconvert(SvNVX(sv), NV_DIG, 0, s);
2982 setlocale(LC_NUMERIC, loc);
2987 /* We don't call SvPOK_on(), because it may come to pass that the
2988 * locale changes so that the stringification we just did is no
2989 * longer correct. We will have to re-stringify every time it is
2996 else if (isGV_with_GP(sv)) {
2997 GV *const gv = MUTABLE_GV(sv);
2998 SV *const buffer = sv_newmortal();
3000 gv_efullname3(buffer, gv, "*");
3002 assert(SvPOK(buffer));
3006 *lp = SvCUR(buffer);
3007 return SvPVX(buffer);
3009 else if (isREGEXP(sv)) {
3010 if (lp) *lp = RX_WRAPLEN((REGEXP *)sv);
3011 return RX_WRAPPED((REGEXP *)sv);
3016 if (flags & SV_UNDEF_RETURNS_NULL)
3018 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
3020 /* Typically the caller expects that sv_any is not NULL now. */
3021 if (!SvREADONLY(sv) && SvTYPE(sv) < SVt_PV)
3022 sv_upgrade(sv, SVt_PV);
3027 const STRLEN len = s - SvPVX_const(sv);
3032 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3033 PTR2UV(sv),SvPVX_const(sv)));
3034 if (flags & SV_CONST_RETURN)
3035 return (char *)SvPVX_const(sv);
3036 if (flags & SV_MUTABLE_RETURN)
3037 return SvPVX_mutable(sv);
3042 =for apidoc sv_copypv
3044 Copies a stringified representation of the source SV into the
3045 destination SV. Automatically performs any necessary mg_get and
3046 coercion of numeric values into strings. Guaranteed to preserve
3047 UTF8 flag even from overloaded objects. Similar in nature to
3048 sv_2pv[_flags] but operates directly on an SV instead of just the
3049 string. Mostly uses sv_2pv_flags to do its work, except when that
3050 would lose the UTF-8'ness of the PV.
3052 =for apidoc sv_copypv_nomg
3054 Like sv_copypv, but doesn't invoke get magic first.
3056 =for apidoc sv_copypv_flags
3058 Implementation of sv_copypv and sv_copypv_nomg. Calls get magic iff flags
3065 Perl_sv_copypv(pTHX_ SV *const dsv, SV *const ssv)
3067 PERL_ARGS_ASSERT_SV_COPYPV;
3069 sv_copypv_flags(dsv, ssv, 0);
3073 Perl_sv_copypv_flags(pTHX_ SV *const dsv, SV *const ssv, const I32 flags)
3078 PERL_ARGS_ASSERT_SV_COPYPV_FLAGS;
3080 if ((flags & SV_GMAGIC) && SvGMAGICAL(ssv))
3082 s = SvPV_nomg_const(ssv,len);
3083 sv_setpvn(dsv,s,len);
3091 =for apidoc sv_2pvbyte
3093 Return a pointer to the byte-encoded representation of the SV, and set *lp
3094 to its length. May cause the SV to be downgraded from UTF-8 as a
3097 Usually accessed via the C<SvPVbyte> macro.
3103 Perl_sv_2pvbyte(pTHX_ SV *sv, STRLEN *const lp)
3105 PERL_ARGS_ASSERT_SV_2PVBYTE;
3108 if (((SvREADONLY(sv) || SvFAKE(sv)) && !SvIsCOW(sv))
3109 || isGV_with_GP(sv) || SvROK(sv)) {
3110 SV *sv2 = sv_newmortal();
3111 sv_copypv_nomg(sv2,sv);
3114 sv_utf8_downgrade(sv,0);
3115 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3119 =for apidoc sv_2pvutf8
3121 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3122 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3124 Usually accessed via the C<SvPVutf8> macro.
3130 Perl_sv_2pvutf8(pTHX_ SV *sv, STRLEN *const lp)
3132 PERL_ARGS_ASSERT_SV_2PVUTF8;
3134 if (((SvREADONLY(sv) || SvFAKE(sv)) && !SvIsCOW(sv))
3135 || isGV_with_GP(sv) || SvROK(sv))
3136 sv = sv_mortalcopy(sv);
3139 sv_utf8_upgrade_nomg(sv);
3140 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3145 =for apidoc sv_2bool
3147 This macro is only used by sv_true() or its macro equivalent, and only if
3148 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3149 It calls sv_2bool_flags with the SV_GMAGIC flag.
3151 =for apidoc sv_2bool_flags
3153 This function is only used by sv_true() and friends, and only if
3154 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3155 contain SV_GMAGIC, then it does an mg_get() first.
3162 Perl_sv_2bool_flags(pTHX_ SV *sv, I32 flags)
3166 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3169 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3175 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3176 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv)))) {
3179 if(SvGMAGICAL(sv)) {
3181 goto restart; /* call sv_2bool */
3183 /* expanded SvTRUE_common(sv, (flags = 0, goto restart)) */
3184 else if(!SvOK(sv)) {
3187 else if(SvPOK(sv)) {
3188 svb = SvPVXtrue(sv);
3190 else if((SvFLAGS(sv) & (SVf_IOK|SVf_NOK))) {
3191 svb = (SvIOK(sv) && SvIVX(sv) != 0)
3192 || (SvNOK(sv) && SvNVX(sv) != 0.0);
3196 goto restart; /* call sv_2bool_nomg */
3201 return SvRV(sv) != 0;
3205 RX_WRAPLEN(sv) > 1 || (RX_WRAPLEN(sv) && *RX_WRAPPED(sv) != '0');
3206 return SvTRUE_common(sv, isGV_with_GP(sv) ? 1 : 0);
3210 =for apidoc sv_utf8_upgrade
3212 Converts the PV of an SV to its UTF-8-encoded form.
3213 Forces the SV to string form if it is not already.
3214 Will C<mg_get> on C<sv> if appropriate.
3215 Always sets the SvUTF8 flag to avoid future validity checks even
3216 if the whole string is the same in UTF-8 as not.
3217 Returns the number of bytes in the converted string
3219 This is not a general purpose byte encoding to Unicode interface:
3220 use the Encode extension for that.
3222 =for apidoc sv_utf8_upgrade_nomg
3224 Like sv_utf8_upgrade, but doesn't do magic on C<sv>.
3226 =for apidoc sv_utf8_upgrade_flags
3228 Converts the PV of an SV to its UTF-8-encoded form.
3229 Forces the SV to string form if it is not already.
3230 Always sets the SvUTF8 flag to avoid future validity checks even
3231 if all the bytes are invariant in UTF-8.
3232 If C<flags> has C<SV_GMAGIC> bit set,
3233 will C<mg_get> on C<sv> if appropriate, else not.
3235 If C<flags> has SV_FORCE_UTF8_UPGRADE set, this function assumes that the PV
3236 will expand when converted to UTF-8, and skips the extra work of checking for
3237 that. Typically this flag is used by a routine that has already parsed the
3238 string and found such characters, and passes this information on so that the
3239 work doesn't have to be repeated.
3241 Returns the number of bytes in the converted string.
3243 This is not a general purpose byte encoding to Unicode interface:
3244 use the Encode extension for that.
3246 =for apidoc sv_utf8_upgrade_flags_grow
3248 Like sv_utf8_upgrade_flags, but has an additional parameter C<extra>, which is
3249 the number of unused bytes the string of 'sv' is guaranteed to have free after
3250 it upon return. This allows the caller to reserve extra space that it intends
3251 to fill, to avoid extra grows.
3253 C<sv_utf8_upgrade>, C<sv_utf8_upgrade_nomg>, and C<sv_utf8_upgrade_flags>
3254 are implemented in terms of this function.
3256 Returns the number of bytes in the converted string (not including the spares).
3260 (One might think that the calling routine could pass in the position of the
3261 first variant character when it has set SV_FORCE_UTF8_UPGRADE, so it wouldn't
3262 have to be found again. But that is not the case, because typically when the
3263 caller is likely to use this flag, it won't be calling this routine unless it
3264 finds something that won't fit into a byte. Otherwise it tries to not upgrade
3265 and just use bytes. But some things that do fit into a byte are variants in
3266 utf8, and the caller may not have been keeping track of these.)
3268 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3269 isn't guaranteed due to having other routines do the work in some input cases,
3270 or if the input is already flagged as being in utf8.
3272 The speed of this could perhaps be improved for many cases if someone wanted to
3273 write a fast function that counts the number of variant characters in a string,
3274 especially if it could return the position of the first one.
3279 Perl_sv_utf8_upgrade_flags_grow(pTHX_ SV *const sv, const I32 flags, STRLEN extra)
3283 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3285 if (sv == &PL_sv_undef)
3287 if (!SvPOK_nog(sv)) {
3289 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3290 (void) sv_2pv_flags(sv,&len, flags);
3292 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3296 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3301 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3306 S_sv_uncow(aTHX_ sv, 0);
3309 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3310 sv_recode_to_utf8(sv, PL_encoding);
3311 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3315 if (SvCUR(sv) == 0) {
3316 if (extra) SvGROW(sv, extra);
3317 } else { /* Assume Latin-1/EBCDIC */
3318 /* This function could be much more efficient if we
3319 * had a FLAG in SVs to signal if there are any variant
3320 * chars in the PV. Given that there isn't such a flag
3321 * make the loop as fast as possible (although there are certainly ways
3322 * to speed this up, eg. through vectorization) */
3323 U8 * s = (U8 *) SvPVX_const(sv);
3324 U8 * e = (U8 *) SvEND(sv);
3326 STRLEN two_byte_count = 0;
3328 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3330 /* See if really will need to convert to utf8. We mustn't rely on our
3331 * incoming SV being well formed and having a trailing '\0', as certain
3332 * code in pp_formline can send us partially built SVs. */
3336 if (NATIVE_BYTE_IS_INVARIANT(ch)) continue;
3338 t--; /* t already incremented; re-point to first variant */
3343 /* utf8 conversion not needed because all are invariants. Mark as
3344 * UTF-8 even if no variant - saves scanning loop */
3346 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3351 /* Here, the string should be converted to utf8, either because of an
3352 * input flag (two_byte_count = 0), or because a character that
3353 * requires 2 bytes was found (two_byte_count = 1). t points either to
3354 * the beginning of the string (if we didn't examine anything), or to
3355 * the first variant. In either case, everything from s to t - 1 will
3356 * occupy only 1 byte each on output.
3358 * There are two main ways to convert. One is to create a new string
3359 * and go through the input starting from the beginning, appending each
3360 * converted value onto the new string as we go along. It's probably
3361 * best to allocate enough space in the string for the worst possible
3362 * case rather than possibly running out of space and having to
3363 * reallocate and then copy what we've done so far. Since everything
3364 * from s to t - 1 is invariant, the destination can be initialized
3365 * with these using a fast memory copy
3367 * The other way is to figure out exactly how big the string should be
3368 * by parsing the entire input. Then you don't have to make it big
3369 * enough to handle the worst possible case, and more importantly, if
3370 * the string you already have is large enough, you don't have to
3371 * allocate a new string, you can copy the last character in the input
3372 * string to the final position(s) that will be occupied by the
3373 * converted string and go backwards, stopping at t, since everything
3374 * before that is invariant.
3376 * There are advantages and disadvantages to each method.
3378 * In the first method, we can allocate a new string, do the memory
3379 * copy from the s to t - 1, and then proceed through the rest of the
3380 * string byte-by-byte.
3382 * In the second method, we proceed through the rest of the input
3383 * string just calculating how big the converted string will be. Then
3384 * there are two cases:
3385 * 1) if the string has enough extra space to handle the converted
3386 * value. We go backwards through the string, converting until we
3387 * get to the position we are at now, and then stop. If this
3388 * position is far enough along in the string, this method is
3389 * faster than the other method. If the memory copy were the same
3390 * speed as the byte-by-byte loop, that position would be about
3391 * half-way, as at the half-way mark, parsing to the end and back
3392 * is one complete string's parse, the same amount as starting
3393 * over and going all the way through. Actually, it would be
3394 * somewhat less than half-way, as it's faster to just count bytes
3395 * than to also copy, and we don't have the overhead of allocating
3396 * a new string, changing the scalar to use it, and freeing the
3397 * existing one. But if the memory copy is fast, the break-even
3398 * point is somewhere after half way. The counting loop could be
3399 * sped up by vectorization, etc, to move the break-even point
3400 * further towards the beginning.
3401 * 2) if the string doesn't have enough space to handle the converted
3402 * value. A new string will have to be allocated, and one might
3403 * as well, given that, start from the beginning doing the first
3404 * method. We've spent extra time parsing the string and in
3405 * exchange all we've gotten is that we know precisely how big to
3406 * make the new one. Perl is more optimized for time than space,
3407 * so this case is a loser.
3408 * So what I've decided to do is not use the 2nd method unless it is
3409 * guaranteed that a new string won't have to be allocated, assuming
3410 * the worst case. I also decided not to put any more conditions on it
3411 * than this, for now. It seems likely that, since the worst case is
3412 * twice as big as the unknown portion of the string (plus 1), we won't
3413 * be guaranteed enough space, causing us to go to the first method,
3414 * unless the string is short, or the first variant character is near
3415 * the end of it. In either of these cases, it seems best to use the
3416 * 2nd method. The only circumstance I can think of where this would
3417 * be really slower is if the string had once had much more data in it
3418 * than it does now, but there is still a substantial amount in it */
3421 STRLEN invariant_head = t - s;
3422 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3423 if (SvLEN(sv) < size) {
3425 /* Here, have decided to allocate a new string */
3430 Newx(dst, size, U8);
3432 /* If no known invariants at the beginning of the input string,
3433 * set so starts from there. Otherwise, can use memory copy to
3434 * get up to where we are now, and then start from here */
3436 if (invariant_head <= 0) {
3439 Copy(s, dst, invariant_head, char);
3440 d = dst + invariant_head;
3444 append_utf8_from_native_byte(*t, &d);
3448 SvPV_free(sv); /* No longer using pre-existing string */
3449 SvPV_set(sv, (char*)dst);
3450 SvCUR_set(sv, d - dst);
3451 SvLEN_set(sv, size);
3454 /* Here, have decided to get the exact size of the string.
3455 * Currently this happens only when we know that there is
3456 * guaranteed enough space to fit the converted string, so
3457 * don't have to worry about growing. If two_byte_count is 0,
3458 * then t points to the first byte of the string which hasn't
3459 * been examined yet. Otherwise two_byte_count is 1, and t
3460 * points to the first byte in the string that will expand to
3461 * two. Depending on this, start examining at t or 1 after t.
3464 U8 *d = t + two_byte_count;
3467 /* Count up the remaining bytes that expand to two */
3470 const U8 chr = *d++;
3471 if (! NATIVE_BYTE_IS_INVARIANT(chr)) two_byte_count++;
3474 /* The string will expand by just the number of bytes that
3475 * occupy two positions. But we are one afterwards because of
3476 * the increment just above. This is the place to put the
3477 * trailing NUL, and to set the length before we decrement */
3479 d += two_byte_count;
3480 SvCUR_set(sv, d - s);
3484 /* Having decremented d, it points to the position to put the
3485 * very last byte of the expanded string. Go backwards through
3486 * the string, copying and expanding as we go, stopping when we
3487 * get to the part that is invariant the rest of the way down */
3491 if (NATIVE_BYTE_IS_INVARIANT(*e)) {
3494 *d-- = UTF8_EIGHT_BIT_LO(*e);
3495 *d-- = UTF8_EIGHT_BIT_HI(*e);
3501 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3502 /* Update pos. We do it at the end rather than during
3503 * the upgrade, to avoid slowing down the common case
3504 * (upgrade without pos).
3505 * pos can be stored as either bytes or characters. Since
3506 * this was previously a byte string we can just turn off
3507 * the bytes flag. */
3508 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3510 mg->mg_flags &= ~MGf_BYTES;
3512 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3513 magic_setutf8(sv,mg); /* clear UTF8 cache */
3518 /* Mark as UTF-8 even if no variant - saves scanning loop */
3524 =for apidoc sv_utf8_downgrade
3526 Attempts to convert the PV of an SV from characters to bytes.
3527 If the PV contains a character that cannot fit
3528 in a byte, this conversion will fail;
3529 in this case, either returns false or, if C<fail_ok> is not
3532 This is not a general purpose Unicode to byte encoding interface:
3533 use the Encode extension for that.
3539 Perl_sv_utf8_downgrade(pTHX_ SV *const sv, const bool fail_ok)
3543 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3545 if (SvPOKp(sv) && SvUTF8(sv)) {
3549 int mg_flags = SV_GMAGIC;
3552 S_sv_uncow(aTHX_ sv, 0);
3554 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3556 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3557 if (mg && mg->mg_len > 0 && mg->mg_flags & MGf_BYTES) {
3558 mg->mg_len = sv_pos_b2u_flags(sv, mg->mg_len,
3559 SV_GMAGIC|SV_CONST_RETURN);
3560 mg_flags = 0; /* sv_pos_b2u does get magic */
3562 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3563 magic_setutf8(sv,mg); /* clear UTF8 cache */
3566 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3568 if (!utf8_to_bytes(s, &len)) {
3573 Perl_croak(aTHX_ "Wide character in %s",
3576 Perl_croak(aTHX_ "Wide character");
3587 =for apidoc sv_utf8_encode
3589 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3590 flag off so that it looks like octets again.
3596 Perl_sv_utf8_encode(pTHX_ SV *const sv)
3598 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3600 if (SvREADONLY(sv)) {
3601 sv_force_normal_flags(sv, 0);
3603 (void) sv_utf8_upgrade(sv);
3608 =for apidoc sv_utf8_decode
3610 If the PV of the SV is an octet sequence in UTF-8
3611 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3612 so that it looks like a character. If the PV contains only single-byte
3613 characters, the C<SvUTF8> flag stays off.
3614 Scans PV for validity and returns false if the PV is invalid UTF-8.
3620 Perl_sv_utf8_decode(pTHX_ SV *const sv)
3622 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3625 const U8 *start, *c;
3628 /* The octets may have got themselves encoded - get them back as
3631 if (!sv_utf8_downgrade(sv, TRUE))
3634 /* it is actually just a matter of turning the utf8 flag on, but
3635 * we want to make sure everything inside is valid utf8 first.
3637 c = start = (const U8 *) SvPVX_const(sv);
3638 if (!is_utf8_string(c, SvCUR(sv)))
3640 e = (const U8 *) SvEND(sv);
3643 if (!UTF8_IS_INVARIANT(ch)) {
3648 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3649 /* XXX Is this dead code? XS_utf8_decode calls SvSETMAGIC
3650 after this, clearing pos. Does anything on CPAN
3652 /* adjust pos to the start of a UTF8 char sequence */
3653 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3655 I32 pos = mg->mg_len;
3657 for (c = start + pos; c > start; c--) {
3658 if (UTF8_IS_START(*c))
3661 mg->mg_len = c - start;
3664 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3665 magic_setutf8(sv,mg); /* clear UTF8 cache */
3672 =for apidoc sv_setsv
3674 Copies the contents of the source SV C<ssv> into the destination SV
3675 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3676 function if the source SV needs to be reused. Does not handle 'set' magic on
3677 destination SV. Calls 'get' magic on source SV. Loosely speaking, it
3678 performs a copy-by-value, obliterating any previous content of the
3681 You probably want to use one of the assortment of wrappers, such as
3682 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3683 C<SvSetMagicSV_nosteal>.
3685 =for apidoc sv_setsv_flags
3687 Copies the contents of the source SV C<ssv> into the destination SV
3688 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3689 function if the source SV needs to be reused. Does not handle 'set' magic.
3690 Loosely speaking, it performs a copy-by-value, obliterating any previous
3691 content of the destination.
3692 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3693 C<ssv> if appropriate, else not. If the C<flags>
3694 parameter has the C<SV_NOSTEAL> bit set then the
3695 buffers of temps will not be stolen. <sv_setsv>
3696 and C<sv_setsv_nomg> are implemented in terms of this function.
3698 You probably want to use one of the assortment of wrappers, such as
3699 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3700 C<SvSetMagicSV_nosteal>.
3702 This is the primary function for copying scalars, and most other
3703 copy-ish functions and macros use this underneath.
3709 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3711 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3712 HV *old_stash = NULL;
3714 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3716 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3717 const char * const name = GvNAME(sstr);
3718 const STRLEN len = GvNAMELEN(sstr);
3720 if (dtype >= SVt_PV) {
3726 SvUPGRADE(dstr, SVt_PVGV);
3727 (void)SvOK_off(dstr);
3728 /* We have to turn this on here, even though we turn it off
3729 below, as GvSTASH will fail an assertion otherwise. */
3730 isGV_with_GP_on(dstr);
3732 GvSTASH(dstr) = GvSTASH(sstr);
3734 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3735 gv_name_set(MUTABLE_GV(dstr), name, len,
3736 GV_ADD | (GvNAMEUTF8(sstr) ? SVf_UTF8 : 0 ));
3737 SvFAKE_on(dstr); /* can coerce to non-glob */
3740 if(GvGP(MUTABLE_GV(sstr))) {
3741 /* If source has method cache entry, clear it */
3743 SvREFCNT_dec(GvCV(sstr));
3744 GvCV_set(sstr, NULL);
3747 /* If source has a real method, then a method is
3750 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3756 /* If dest already had a real method, that's a change as well */
3758 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3759 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3764 /* We don't need to check the name of the destination if it was not a
3765 glob to begin with. */
3766 if(dtype == SVt_PVGV) {
3767 const char * const name = GvNAME((const GV *)dstr);
3770 /* The stash may have been detached from the symbol table, so
3772 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3776 const STRLEN len = GvNAMELEN(dstr);
3777 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3778 || (len == 1 && name[0] == ':')) {
3781 /* Set aside the old stash, so we can reset isa caches on
3783 if((old_stash = GvHV(dstr)))
3784 /* Make sure we do not lose it early. */
3785 SvREFCNT_inc_simple_void_NN(
3786 sv_2mortal((SV *)old_stash)
3792 gp_free(MUTABLE_GV(dstr));
3793 isGV_with_GP_off(dstr); /* SvOK_off does not like globs. */
3794 (void)SvOK_off(dstr);
3795 isGV_with_GP_on(dstr);
3796 GvINTRO_off(dstr); /* one-shot flag */
3797 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3798 if (SvTAINTED(sstr))
3800 if (GvIMPORTED(dstr) != GVf_IMPORTED
3801 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3803 GvIMPORTED_on(dstr);
3806 if(mro_changes == 2) {
3807 if (GvAV((const GV *)sstr)) {
3809 SV * const sref = (SV *)GvAV((const GV *)dstr);
3810 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3811 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3812 AV * const ary = newAV();
3813 av_push(ary, mg->mg_obj); /* takes the refcount */
3814 mg->mg_obj = (SV *)ary;
3816 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3818 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3820 mro_isa_changed_in(GvSTASH(dstr));
3822 else if(mro_changes == 3) {
3823 HV * const stash = GvHV(dstr);
3824 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3830 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3831 if (GvIO(dstr) && dtype == SVt_PVGV) {
3832 DEBUG_o(Perl_deb(aTHX_
3833 "glob_assign_glob clearing PL_stashcache\n"));
3834 /* It's a cache. It will rebuild itself quite happily.
3835 It's a lot of effort to work out exactly which key (or keys)
3836 might be invalidated by the creation of the this file handle.
3838 hv_clear(PL_stashcache);
3844 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3846 SV * const sref = SvRV(sstr);
3848 const int intro = GvINTRO(dstr);
3851 const U32 stype = SvTYPE(sref);
3853 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3856 GvINTRO_off(dstr); /* one-shot flag */
3857 GvLINE(dstr) = CopLINE(PL_curcop);
3858 GvEGV(dstr) = MUTABLE_GV(dstr);
3863 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3864 import_flag = GVf_IMPORTED_CV;
3867 location = (SV **) &GvHV(dstr);
3868 import_flag = GVf_IMPORTED_HV;
3871 location = (SV **) &GvAV(dstr);
3872 import_flag = GVf_IMPORTED_AV;
3875 location = (SV **) &GvIOp(dstr);
3878 location = (SV **) &GvFORM(dstr);
3881 location = &GvSV(dstr);
3882 import_flag = GVf_IMPORTED_SV;
3885 if (stype == SVt_PVCV) {
3886 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3887 if (GvCVGEN(dstr)) {
3888 SvREFCNT_dec(GvCV(dstr));
3889 GvCV_set(dstr, NULL);
3890 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3893 /* SAVEt_GVSLOT takes more room on the savestack and has more
3894 overhead in leave_scope than SAVEt_GENERIC_SV. But for CVs
3895 leave_scope needs access to the GV so it can reset method
3896 caches. We must use SAVEt_GVSLOT whenever the type is
3897 SVt_PVCV, even if the stash is anonymous, as the stash may
3898 gain a name somehow before leave_scope. */
3899 if (stype == SVt_PVCV) {
3900 /* There is no save_pushptrptrptr. Creating it for this
3901 one call site would be overkill. So inline the ss add
3905 SS_ADD_PTR(location);
3906 SS_ADD_PTR(SvREFCNT_inc(*location));
3907 SS_ADD_UV(SAVEt_GVSLOT);
3910 else SAVEGENERICSV(*location);
3913 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3914 CV* const cv = MUTABLE_CV(*location);
3916 if (!GvCVGEN((const GV *)dstr) &&
3917 (CvROOT(cv) || CvXSUB(cv)) &&
3918 /* redundant check that avoids creating the extra SV
3919 most of the time: */
3920 (CvCONST(cv) || ckWARN(WARN_REDEFINE)))
3922 SV * const new_const_sv =
3923 CvCONST((const CV *)sref)
3924 ? cv_const_sv((const CV *)sref)
3926 report_redefined_cv(
3927 sv_2mortal(Perl_newSVpvf(aTHX_
3930 HvNAME_HEK(GvSTASH((const GV *)dstr))
3932 HEKfARG(GvENAME_HEK(MUTABLE_GV(dstr)))
3935 CvCONST((const CV *)sref) ? &new_const_sv : NULL
3939 cv_ckproto_len_flags(cv, (const GV *)dstr,
3940 SvPOK(sref) ? CvPROTO(sref) : NULL,
3941 SvPOK(sref) ? CvPROTOLEN(sref) : 0,
3942 SvPOK(sref) ? SvUTF8(sref) : 0);
3944 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3945 GvASSUMECV_on(dstr);
3946 if(GvSTASH(dstr)) gv_method_changed(dstr); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3948 *location = SvREFCNT_inc_simple_NN(sref);
3949 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3950 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3951 GvFLAGS(dstr) |= import_flag;
3953 if (stype == SVt_PVHV) {
3954 const char * const name = GvNAME((GV*)dstr);
3955 const STRLEN len = GvNAMELEN(dstr);
3958 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3959 || (len == 1 && name[0] == ':')
3961 && (!dref || HvENAME_get(dref))
3964 (HV *)sref, (HV *)dref,
3970 stype == SVt_PVAV && sref != dref
3971 && strEQ(GvNAME((GV*)dstr), "ISA")
3972 /* The stash may have been detached from the symbol table, so
3973 check its name before doing anything. */
3974 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3977 MAGIC * const omg = dref && SvSMAGICAL(dref)
3978 ? mg_find(dref, PERL_MAGIC_isa)
3980 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3981 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3982 AV * const ary = newAV();
3983 av_push(ary, mg->mg_obj); /* takes the refcount */
3984 mg->mg_obj = (SV *)ary;
3987 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3988 SV **svp = AvARRAY((AV *)omg->mg_obj);
3989 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3993 SvREFCNT_inc_simple_NN(*svp++)
3999 SvREFCNT_inc_simple_NN(omg->mg_obj)
4003 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
4008 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
4010 mg = mg_find(sref, PERL_MAGIC_isa);
4012 /* Since the *ISA assignment could have affected more than
4013 one stash, don't call mro_isa_changed_in directly, but let
4014 magic_clearisa do it for us, as it already has the logic for
4015 dealing with globs vs arrays of globs. */
4017 Perl_magic_clearisa(aTHX_ NULL, mg);
4019 else if (stype == SVt_PVIO) {
4020 DEBUG_o(Perl_deb(aTHX_ "glob_assign_ref clearing PL_stashcache\n"));
4021 /* It's a cache. It will rebuild itself quite happily.
4022 It's a lot of effort to work out exactly which key (or keys)
4023 might be invalidated by the creation of the this file handle.
4025 hv_clear(PL_stashcache);
4029 if (!intro) SvREFCNT_dec(dref);
4030 if (SvTAINTED(sstr))
4035 /* Work around compiler warnings about unsigned >= THRESHOLD when thres-
4037 #if SV_COW_THRESHOLD
4038 # define GE_COW_THRESHOLD(len) ((len) >= SV_COW_THRESHOLD)
4040 # define GE_COW_THRESHOLD(len) 1
4042 #if SV_COWBUF_THRESHOLD
4043 # define GE_COWBUF_THRESHOLD(len) ((len) >= SV_COWBUF_THRESHOLD)
4045 # define GE_COWBUF_THRESHOLD(len) 1
4049 Perl_sv_setsv_flags(pTHX_ SV *dstr, SV* sstr, const I32 flags)
4056 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
4061 if (SvIS_FREED(dstr)) {
4062 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
4063 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
4065 SV_CHECK_THINKFIRST_COW_DROP(dstr);
4067 sstr = &PL_sv_undef;
4068 if (SvIS_FREED(sstr)) {
4069 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
4070 (void*)sstr, (void*)dstr);
4072 stype = SvTYPE(sstr);
4073 dtype = SvTYPE(dstr);
4075 /* There's a lot of redundancy below but we're going for speed here */
4080 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
4081 (void)SvOK_off(dstr);
4089 sv_upgrade(dstr, SVt_IV);
4093 sv_upgrade(dstr, SVt_PVIV);
4097 goto end_of_first_switch;
4099 (void)SvIOK_only(dstr);
4100 SvIV_set(dstr, SvIVX(sstr));
4103 /* SvTAINTED can only be true if the SV has taint magic, which in
4104 turn means that the SV type is PVMG (or greater). This is the
4105 case statement for SVt_IV, so this cannot be true (whatever gcov
4107 assert(!SvTAINTED(sstr));
4112 if (dtype < SVt_PV && dtype != SVt_IV)
4113 sv_upgrade(dstr, SVt_IV);
4121 sv_upgrade(dstr, SVt_NV);
4125 sv_upgrade(dstr, SVt_PVNV);
4129 goto end_of_first_switch;
4131 SvNV_set(dstr, SvNVX(sstr));
4132 (void)SvNOK_only(dstr);
4133 /* SvTAINTED can only be true if the SV has taint magic, which in
4134 turn means that the SV type is PVMG (or greater). This is the
4135 case statement for SVt_NV, so this cannot be true (whatever gcov
4137 assert(!SvTAINTED(sstr));
4144 sv_upgrade(dstr, SVt_PV);
4147 if (dtype < SVt_PVIV)
4148 sv_upgrade(dstr, SVt_PVIV);
4151 if (dtype < SVt_PVNV)
4152 sv_upgrade(dstr, SVt_PVNV);
4156 const char * const type = sv_reftype(sstr,0);
4158 /* diag_listed_as: Bizarre copy of %s */
4159 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4161 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4167 if (dtype < SVt_REGEXP)
4169 if (dtype >= SVt_PV) {
4175 sv_upgrade(dstr, SVt_REGEXP);
4183 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4185 if (SvTYPE(sstr) != stype)
4186 stype = SvTYPE(sstr);
4188 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4189 glob_assign_glob(dstr, sstr, dtype);
4192 if (stype == SVt_PVLV)
4194 if (isREGEXP(sstr)) goto upgregexp;
4195 SvUPGRADE(dstr, SVt_PVNV);
4198 SvUPGRADE(dstr, (svtype)stype);
4200 end_of_first_switch:
4202 /* dstr may have been upgraded. */
4203 dtype = SvTYPE(dstr);
4204 sflags = SvFLAGS(sstr);
4206 if (dtype == SVt_PVCV) {
4207 /* Assigning to a subroutine sets the prototype. */
4210 const char *const ptr = SvPV_const(sstr, len);
4212 SvGROW(dstr, len + 1);
4213 Copy(ptr, SvPVX(dstr), len + 1, char);
4214 SvCUR_set(dstr, len);
4216 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4217 CvAUTOLOAD_off(dstr);
4222 else if (dtype == SVt_PVAV || dtype == SVt_PVHV || dtype == SVt_PVFM) {
4223 const char * const type = sv_reftype(dstr,0);
4225 /* diag_listed_as: Cannot copy to %s */
4226 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4228 Perl_croak(aTHX_ "Cannot copy to %s", type);
4229 } else if (sflags & SVf_ROK) {
4230 if (isGV_with_GP(dstr)
4231 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4234 if (GvIMPORTED(dstr) != GVf_IMPORTED
4235 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4237 GvIMPORTED_on(dstr);
4242 glob_assign_glob(dstr, sstr, dtype);
4246 if (dtype >= SVt_PV) {
4247 if (isGV_with_GP(dstr)) {
4248 glob_assign_ref(dstr, sstr);
4251 if (SvPVX_const(dstr)) {
4257 (void)SvOK_off(dstr);
4258 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4259 SvFLAGS(dstr) |= sflags & SVf_ROK;
4260 assert(!(sflags & SVp_NOK));
4261 assert(!(sflags & SVp_IOK));
4262 assert(!(sflags & SVf_NOK));
4263 assert(!(sflags & SVf_IOK));
4265 else if (isGV_with_GP(dstr)) {
4266 if (!(sflags & SVf_OK)) {
4267 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4268 "Undefined value assigned to typeglob");
4271 GV *gv = gv_fetchsv_nomg(sstr, GV_ADD, SVt_PVGV);
4272 if (dstr != (const SV *)gv) {
4273 const char * const name = GvNAME((const GV *)dstr);
4274 const STRLEN len = GvNAMELEN(dstr);
4275 HV *old_stash = NULL;
4276 bool reset_isa = FALSE;
4277 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4278 || (len == 1 && name[0] == ':')) {
4279 /* Set aside the old stash, so we can reset isa caches
4280 on its subclasses. */
4281 if((old_stash = GvHV(dstr))) {
4282 /* Make sure we do not lose it early. */
4283 SvREFCNT_inc_simple_void_NN(
4284 sv_2mortal((SV *)old_stash)
4291 gp_free(MUTABLE_GV(dstr));
4292 GvGP_set(dstr, gp_ref(GvGP(gv)));
4295 HV * const stash = GvHV(dstr);
4297 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4307 else if ((dtype == SVt_REGEXP || dtype == SVt_PVLV)
4308 && (stype == SVt_REGEXP || isREGEXP(sstr))) {
4309 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4311 else if (sflags & SVp_POK) {
4312 const STRLEN cur = SvCUR(sstr);
4313 const STRLEN len = SvLEN(sstr);
4316 * We have three basic ways to copy the string:
4322 * Which we choose is based on various factors. The following
4323 * things are listed in order of speed, fastest to slowest:
4325 * - Copying a short string
4326 * - Copy-on-write bookkeeping
4328 * - Copying a long string
4330 * We swipe the string (steal the string buffer) if the SV on the
4331 * rhs is about to be freed anyway (TEMP and refcnt==1). This is a
4332 * big win on long strings. It should be a win on short strings if
4333 * SvPVX_const(dstr) has to be allocated. If not, it should not
4334 * slow things down, as SvPVX_const(sstr) would have been freed
4337 * We also steal the buffer from a PADTMP (operator target) if it
4338 * is ‘long enough’. For short strings, a swipe does not help
4339 * here, as it causes more malloc calls the next time the target
4340 * is used. Benchmarks show that even if SvPVX_const(dstr) has to
4341 * be allocated it is still not worth swiping PADTMPs for short
4342 * strings, as the savings here are small.
4344 * If the rhs is already flagged as a copy-on-write string and COW
4345 * is possible here, we use copy-on-write and make both SVs share
4346 * the string buffer.
4348 * If the rhs is not flagged as copy-on-write, then we see whether
4349 * it is worth upgrading it to such. If the lhs already has a buf-
4350 * fer big enough and the string is short, we skip it and fall back
4351 * to method 3, since memcpy is faster for short strings than the
4352 * later bookkeeping overhead that copy-on-write entails.
4354 * If there is no buffer on the left, or the buffer is too small,
4355 * then we use copy-on-write.
4358 /* Whichever path we take through the next code, we want this true,
4359 and doing it now facilitates the COW check. */
4360 (void)SvPOK_only(dstr);
4364 /* slated for free anyway (and not COW)? */
4365 (sflags & (SVs_TEMP|SVf_IsCOW)) == SVs_TEMP
4366 /* or a swipable TARG */
4367 || ((sflags & (SVs_PADTMP|SVf_READONLY|SVf_IsCOW))
4369 /* whose buffer is worth stealing */
4370 && GE_COWBUF_THRESHOLD(cur)
4373 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4374 (!(flags & SV_NOSTEAL)) &&
4375 /* and we're allowed to steal temps */
4376 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4377 len) /* and really is a string */
4378 { /* Passes the swipe test. */
4379 if (SvPVX_const(dstr)) /* we know that dtype >= SVt_PV */
4381 SvPV_set(dstr, SvPVX_mutable(sstr));
4382 SvLEN_set(dstr, SvLEN(sstr));
4383 SvCUR_set(dstr, SvCUR(sstr));
4386 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4387 SvPV_set(sstr, NULL);
4392 else if (flags & SV_COW_SHARED_HASH_KEYS
4394 #ifdef PERL_OLD_COPY_ON_WRITE
4395 ( sflags & SVf_IsCOW
4396 || ( (sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4397 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4398 && SvTYPE(sstr) >= SVt_PVIV && len
4401 #elif defined(PERL_NEW_COPY_ON_WRITE)
4404 ( (GE_COWBUF_THRESHOLD(cur) || SvLEN(dstr) < cur+1)
4405 /* If this is a regular (non-hek) COW, only so
4406 many COW "copies" are possible. */
4407 && CowREFCNT(sstr) != SV_COW_REFCNT_MAX ))
4408 : ( (sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4409 && !(SvFLAGS(dstr) & SVf_BREAK)
4410 && GE_COW_THRESHOLD(cur) && cur+1 < len
4411 && (GE_COWBUF_THRESHOLD(cur) || SvLEN(dstr) < cur+1)
4415 && !(SvFLAGS(dstr) & SVf_BREAK)
4418 /* Either it's a shared hash key, or it's suitable for
4421 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4426 if (!(sflags & SVf_IsCOW)) {
4428 # ifdef PERL_OLD_COPY_ON_WRITE
4429 /* Make the source SV into a loop of 1.
4430 (about to become 2) */
4431 SV_COW_NEXT_SV_SET(sstr, sstr);
4433 CowREFCNT(sstr) = 0;
4437 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4443 # ifdef PERL_OLD_COPY_ON_WRITE
4444 assert (SvTYPE(dstr) >= SVt_PVIV);
4445 /* SvIsCOW_normal */
4446 /* splice us in between source and next-after-source. */
4447 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4448 SV_COW_NEXT_SV_SET(sstr, dstr);
4452 SvPV_set(dstr, SvPVX_mutable(sstr));
4456 /* SvIsCOW_shared_hash */
4457 DEBUG_C(PerlIO_printf(Perl_debug_log,
4458 "Copy on write: Sharing hash\n"));
4460 assert (SvTYPE(dstr) >= SVt_PV);
4462 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4464 SvLEN_set(dstr, len);
4465 SvCUR_set(dstr, cur);
4468 /* Failed the swipe test, and we cannot do copy-on-write either.
4469 Have to copy the string. */
4470 SvGROW(dstr, cur + 1); /* inlined from sv_setpvn */
4471 Move(SvPVX_const(sstr),SvPVX(dstr),cur,char);
4472 SvCUR_set(dstr, cur);
4473 *SvEND(dstr) = '\0';
4475 if (sflags & SVp_NOK) {
4476 SvNV_set(dstr, SvNVX(sstr));
4478 if (sflags & SVp_IOK) {
4479 SvIV_set(dstr, SvIVX(sstr));
4480 /* Must do this otherwise some other overloaded use of 0x80000000
4481 gets confused. I guess SVpbm_VALID */
4482 if (sflags & SVf_IVisUV)
4485 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4487 const MAGIC * const smg = SvVSTRING_mg(sstr);
4489 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4490 smg->mg_ptr, smg->mg_len);
4491 SvRMAGICAL_on(dstr);
4495 else if (sflags & (SVp_IOK|SVp_NOK)) {
4496 (void)SvOK_off(dstr);
4497 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4498 if (sflags & SVp_IOK) {
4499 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4500 SvIV_set(dstr, SvIVX(sstr));
4502 if (sflags & SVp_NOK) {
4503 SvNV_set(dstr, SvNVX(sstr));
4507 if (isGV_with_GP(sstr)) {
4508 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4511 (void)SvOK_off(dstr);
4513 if (SvTAINTED(sstr))
4518 =for apidoc sv_setsv_mg
4520 Like C<sv_setsv>, but also handles 'set' magic.
4526 Perl_sv_setsv_mg(pTHX_ SV *const dstr, SV *const sstr)
4528 PERL_ARGS_ASSERT_SV_SETSV_MG;
4530 sv_setsv(dstr,sstr);
4535 # ifdef PERL_OLD_COPY_ON_WRITE
4536 # define SVt_COW SVt_PVIV
4538 # define SVt_COW SVt_PV
4541 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4543 STRLEN cur = SvCUR(sstr);
4544 STRLEN len = SvLEN(sstr);
4547 PERL_ARGS_ASSERT_SV_SETSV_COW;
4550 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4551 (void*)sstr, (void*)dstr);
4558 if (SvTHINKFIRST(dstr))
4559 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4560 else if (SvPVX_const(dstr))
4561 Safefree(SvPVX_mutable(dstr));
4565 SvUPGRADE(dstr, SVt_COW);
4567 assert (SvPOK(sstr));
4568 assert (SvPOKp(sstr));
4569 # ifdef PERL_OLD_COPY_ON_WRITE
4570 assert (!SvIOK(sstr));
4571 assert (!SvIOKp(sstr));
4572 assert (!SvNOK(sstr));
4573 assert (!SvNOKp(sstr));
4576 if (SvIsCOW(sstr)) {
4578 if (SvLEN(sstr) == 0) {
4579 /* source is a COW shared hash key. */
4580 DEBUG_C(PerlIO_printf(Perl_debug_log,
4581 "Fast copy on write: Sharing hash\n"));
4582 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4585 # ifdef PERL_OLD_COPY_ON_WRITE
4586 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4588 assert(SvCUR(sstr)+1 < SvLEN(sstr));
4589 assert(CowREFCNT(sstr) < SV_COW_REFCNT_MAX);
4592 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4593 SvUPGRADE(sstr, SVt_COW);
4595 DEBUG_C(PerlIO_printf(Perl_debug_log,
4596 "Fast copy on write: Converting sstr to COW\n"));
4597 # ifdef PERL_OLD_COPY_ON_WRITE
4598 SV_COW_NEXT_SV_SET(dstr, sstr);
4600 CowREFCNT(sstr) = 0;
4603 # ifdef PERL_OLD_COPY_ON_WRITE
4604 SV_COW_NEXT_SV_SET(sstr, dstr);
4608 new_pv = SvPVX_mutable(sstr);
4611 SvPV_set(dstr, new_pv);
4612 SvFLAGS(dstr) = (SVt_COW|SVf_POK|SVp_POK|SVf_IsCOW);
4615 SvLEN_set(dstr, len);
4616 SvCUR_set(dstr, cur);
4625 =for apidoc sv_setpvn
4627 Copies a string into an SV. The C<len> parameter indicates the number of
4628 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4629 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4635 Perl_sv_setpvn(pTHX_ SV *const sv, const char *const ptr, const STRLEN len)
4640 PERL_ARGS_ASSERT_SV_SETPVN;
4642 SV_CHECK_THINKFIRST_COW_DROP(sv);
4648 /* len is STRLEN which is unsigned, need to copy to signed */
4651 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen %"
4654 SvUPGRADE(sv, SVt_PV);
4656 dptr = SvGROW(sv, len + 1);
4657 Move(ptr,dptr,len,char);
4660 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4662 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4666 =for apidoc sv_setpvn_mg
4668 Like C<sv_setpvn>, but also handles 'set' magic.
4674 Perl_sv_setpvn_mg(pTHX_ SV *const sv, const char *const ptr, const STRLEN len)
4676 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4678 sv_setpvn(sv,ptr,len);
4683 =for apidoc sv_setpv
4685 Copies a string into an SV. The string must be null-terminated. Does not
4686 handle 'set' magic. See C<sv_setpv_mg>.
4692 Perl_sv_setpv(pTHX_ SV *const sv, const char *const ptr)
4697 PERL_ARGS_ASSERT_SV_SETPV;
4699 SV_CHECK_THINKFIRST_COW_DROP(sv);
4705 SvUPGRADE(sv, SVt_PV);
4707 SvGROW(sv, len + 1);
4708 Move(ptr,SvPVX(sv),len+1,char);
4710 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4712 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4716 =for apidoc sv_setpv_mg
4718 Like C<sv_setpv>, but also handles 'set' magic.
4724 Perl_sv_setpv_mg(pTHX_ SV *const sv, const char *const ptr)
4726 PERL_ARGS_ASSERT_SV_SETPV_MG;
4733 Perl_sv_sethek(pTHX_ SV *const sv, const HEK *const hek)
4737 PERL_ARGS_ASSERT_SV_SETHEK;
4743 if (HEK_LEN(hek) == HEf_SVKEY) {
4744 sv_setsv(sv, *(SV**)HEK_KEY(hek));
4747 const int flags = HEK_FLAGS(hek);
4748 if (flags & HVhek_WASUTF8) {
4749 STRLEN utf8_len = HEK_LEN(hek);
4750 char *as_utf8 = (char *)bytes_to_utf8((U8*)HEK_KEY(hek), &utf8_len);
4751 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
4754 } else if (flags & HVhek_UNSHARED) {
4755 sv_setpvn(sv, HEK_KEY(hek), HEK_LEN(hek));
4758 else SvUTF8_off(sv);
4762 SV_CHECK_THINKFIRST_COW_DROP(sv);
4763 SvUPGRADE(sv, SVt_PV);
4765 SvPV_set(sv,(char *)HEK_KEY(share_hek_hek(hek)));
4766 SvCUR_set(sv, HEK_LEN(hek));
4772 else SvUTF8_off(sv);
4780 =for apidoc sv_usepvn_flags
4782 Tells an SV to use C<ptr> to find its string value. Normally the
4783 string is stored inside the SV but sv_usepvn allows the SV to use an
4784 outside string. The C<ptr> should point to memory that was allocated
4785 by C<malloc>. It must be the start of a mallocked block
4786 of memory, and not a pointer to the middle of it. The
4787 string length, C<len>, must be supplied. By default
4788 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4789 so that pointer should not be freed or used by the programmer after
4790 giving it to sv_usepvn, and neither should any pointers from "behind"
4791 that pointer (e.g. ptr + 1) be used.
4793 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4794 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4795 will be skipped (i.e. the buffer is actually at least 1 byte longer than
4796 C<len>, and already meets the requirements for storing in C<SvPVX>).
4802 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4807 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4809 SV_CHECK_THINKFIRST_COW_DROP(sv);
4810 SvUPGRADE(sv, SVt_PV);
4813 if (flags & SV_SMAGIC)
4817 if (SvPVX_const(sv))
4821 if (flags & SV_HAS_TRAILING_NUL)
4822 assert(ptr[len] == '\0');
4825 allocate = (flags & SV_HAS_TRAILING_NUL)
4827 #ifdef Perl_safesysmalloc_size
4830 PERL_STRLEN_ROUNDUP(len + 1);
4832 if (flags & SV_HAS_TRAILING_NUL) {
4833 /* It's long enough - do nothing.
4834 Specifically Perl_newCONSTSUB is relying on this. */
4837 /* Force a move to shake out bugs in callers. */
4838 char *new_ptr = (char*)safemalloc(allocate);
4839 Copy(ptr, new_ptr, len, char);
4840 PoisonFree(ptr,len,char);
4844 ptr = (char*) saferealloc (ptr, allocate);
4847 #ifdef Perl_safesysmalloc_size
4848 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4850 SvLEN_set(sv, allocate);
4854 if (!(flags & SV_HAS_TRAILING_NUL)) {
4857 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4859 if (flags & SV_SMAGIC)
4863 #ifdef PERL_OLD_COPY_ON_WRITE
4864 /* Need to do this *after* making the SV normal, as we need the buffer
4865 pointer to remain valid until after we've copied it. If we let go too early,
4866 another thread could invalidate it by unsharing last of the same hash key
4867 (which it can do by means other than releasing copy-on-write Svs)
4868 or by changing the other copy-on-write SVs in the loop. */
4870 S_sv_release_COW(pTHX_ SV *sv, const char *pvx, SV *after)
4872 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4874 { /* this SV was SvIsCOW_normal(sv) */
4875 /* we need to find the SV pointing to us. */
4876 SV *current = SV_COW_NEXT_SV(after);
4878 if (current == sv) {
4879 /* The SV we point to points back to us (there were only two of us
4881 Hence other SV is no longer copy on write either. */
4884 /* We need to follow the pointers around the loop. */
4886 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4889 /* don't loop forever if the structure is bust, and we have
4890 a pointer into a closed loop. */
4891 assert (current != after);
4892 assert (SvPVX_const(current) == pvx);
4894 /* Make the SV before us point to the SV after us. */
4895 SV_COW_NEXT_SV_SET(current, after);
4901 =for apidoc sv_force_normal_flags
4903 Undo various types of fakery on an SV, where fakery means
4904 "more than" a string: if the PV is a shared string, make
4905 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4906 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4907 we do the copy, and is also used locally; if this is a
4908 vstring, drop the vstring magic. If C<SV_COW_DROP_PV> is set
4909 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4910 SvPOK_off rather than making a copy. (Used where this
4911 scalar is about to be set to some other value.) In addition,
4912 the C<flags> parameter gets passed to C<sv_unref_flags()>
4913 when unreffing. C<sv_force_normal> calls this function
4914 with flags set to 0.
4920 S_sv_uncow(pTHX_ SV * const sv, const U32 flags)
4924 assert(SvIsCOW(sv));
4927 const char * const pvx = SvPVX_const(sv);
4928 const STRLEN len = SvLEN(sv);
4929 const STRLEN cur = SvCUR(sv);
4930 # ifdef PERL_OLD_COPY_ON_WRITE
4931 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4932 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4933 we'll fail an assertion. */
4934 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4938 PerlIO_printf(Perl_debug_log,
4939 "Copy on write: Force normal %ld\n",
4944 # ifdef PERL_NEW_COPY_ON_WRITE
4945 if (len && CowREFCNT(sv) == 0)
4946 /* We own the buffer ourselves. */
4952 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4953 # ifdef PERL_NEW_COPY_ON_WRITE
4954 /* Must do this first, since the macro uses SvPVX. */
4955 if (len) CowREFCNT(sv)--;
4959 if (flags & SV_COW_DROP_PV) {
4960 /* OK, so we don't need to copy our buffer. */
4963 SvGROW(sv, cur + 1);
4964 Move(pvx,SvPVX(sv),cur,char);
4969 # ifdef PERL_OLD_COPY_ON_WRITE
4970 sv_release_COW(sv, pvx, next);
4973 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4980 const char * const pvx = SvPVX_const(sv);
4981 const STRLEN len = SvCUR(sv);
4985 if (flags & SV_COW_DROP_PV) {
4986 /* OK, so we don't need to copy our buffer. */
4989 SvGROW(sv, len + 1);
4990 Move(pvx,SvPVX(sv),len,char);
4993 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4999 Perl_sv_force_normal_flags(pTHX_ SV *const sv, const U32 flags)
5001 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
5004 Perl_croak_no_modify();
5005 else if (SvIsCOW(sv))
5006 S_sv_uncow(aTHX_ sv, flags);
5008 sv_unref_flags(sv, flags);
5009 else if (SvFAKE(sv) && isGV_with_GP(sv))
5010 sv_unglob(sv, flags);
5011 else if (SvFAKE(sv) && isREGEXP(sv)) {
5012 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
5013 to sv_unglob. We only need it here, so inline it. */
5014 const bool islv = SvTYPE(sv) == SVt_PVLV;
5015 const svtype new_type =
5016 islv ? SVt_NULL : SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
5017 SV *const temp = newSV_type(new_type);
5018 regexp *const temp_p = ReANY((REGEXP *)sv);
5020 if (new_type == SVt_PVMG) {
5021 SvMAGIC_set(temp, SvMAGIC(sv));
5022 SvMAGIC_set(sv, NULL);
5023 SvSTASH_set(temp, SvSTASH(sv));
5024 SvSTASH_set(sv, NULL);
5026 if (!islv) SvCUR_set(temp, SvCUR(sv));
5027 /* Remember that SvPVX is in the head, not the body. But
5028 RX_WRAPPED is in the body. */
5029 assert(ReANY((REGEXP *)sv)->mother_re);
5030 /* Their buffer is already owned by someone else. */
5031 if (flags & SV_COW_DROP_PV) {
5032 /* SvLEN is already 0. For SVt_REGEXP, we have a brand new
5033 zeroed body. For SVt_PVLV, it should have been set to 0
5034 before turning into a regexp. */
5035 assert(!SvLEN(islv ? sv : temp));
5036 sv->sv_u.svu_pv = 0;
5039 sv->sv_u.svu_pv = savepvn(RX_WRAPPED((REGEXP *)sv), SvCUR(sv));
5040 SvLEN_set(islv ? sv : temp, SvCUR(sv)+1);
5044 /* Now swap the rest of the bodies. */
5048 SvFLAGS(sv) &= ~SVTYPEMASK;
5049 SvFLAGS(sv) |= new_type;
5050 SvANY(sv) = SvANY(temp);
5053 SvFLAGS(temp) &= ~(SVTYPEMASK);
5054 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
5055 SvANY(temp) = temp_p;
5056 temp->sv_u.svu_rx = (regexp *)temp_p;
5058 SvREFCNT_dec_NN(temp);
5060 else if (SvVOK(sv)) sv_unmagic(sv, PERL_MAGIC_vstring);
5066 Efficient removal of characters from the beginning of the string buffer.
5067 SvPOK(sv), or at least SvPOKp(sv), must be true and the C<ptr> must be a
5068 pointer to somewhere inside the string buffer. The C<ptr> becomes the first
5069 character of the adjusted string. Uses the "OOK hack". On return, only
5070 SvPOK(sv) and SvPOKp(sv) among the OK flags will be true.
5072 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
5073 refer to the same chunk of data.
5075 The unfortunate similarity of this function's name to that of Perl's C<chop>
5076 operator is strictly coincidental. This function works from the left;
5077 C<chop> works from the right.
5083 Perl_sv_chop(pTHX_ SV *const sv, const char *const ptr)
5094 PERL_ARGS_ASSERT_SV_CHOP;
5096 if (!ptr || !SvPOKp(sv))
5098 delta = ptr - SvPVX_const(sv);
5100 /* Nothing to do. */
5103 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
5104 if (delta > max_delta)
5105 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
5106 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
5107 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), so don't use ptr any more */
5108 SV_CHECK_THINKFIRST(sv);
5109 SvPOK_only_UTF8(sv);
5112 if (!SvLEN(sv)) { /* make copy of shared string */
5113 const char *pvx = SvPVX_const(sv);
5114 const STRLEN len = SvCUR(sv);
5115 SvGROW(sv, len + 1);
5116 Move(pvx,SvPVX(sv),len,char);
5122 SvOOK_offset(sv, old_delta);
5124 SvLEN_set(sv, SvLEN(sv) - delta);
5125 SvCUR_set(sv, SvCUR(sv) - delta);
5126 SvPV_set(sv, SvPVX(sv) + delta);
5128 p = (U8 *)SvPVX_const(sv);
5131 /* how many bytes were evacuated? we will fill them with sentinel
5132 bytes, except for the part holding the new offset of course. */
5135 evacn += (old_delta < 0x100 ? 1 : 1 + sizeof(STRLEN));
5137 assert(evacn <= delta + old_delta);
5141 /* This sets 'delta' to the accumulated value of all deltas so far */
5145 /* If 'delta' fits in a byte, store it just prior to the new beginning of
5146 * the string; otherwise store a 0 byte there and store 'delta' just prior
5147 * to that, using as many bytes as a STRLEN occupies. Thus it overwrites a
5148 * portion of the chopped part of the string */
5149 if (delta < 0x100) {
5153 p -= sizeof(STRLEN);
5154 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
5158 /* Fill the preceding buffer with sentinals to verify that no-one is
5168 =for apidoc sv_catpvn
5170 Concatenates the string onto the end of the string which is in the SV. The
5171 C<len> indicates number of bytes to copy. If the SV has the UTF-8
5172 status set, then the bytes appended should be valid UTF-8.
5173 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
5175 =for apidoc sv_catpvn_flags
5177 Concatenates the string onto the end of the string which is in the SV. The
5178 C<len> indicates number of bytes to copy. If the SV has the UTF-8
5179 status set, then the bytes appended should be valid UTF-8.
5180 If C<flags> has the C<SV_SMAGIC> bit set, will
5181 C<mg_set> on C<dsv> afterwards if appropriate.
5182 C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
5183 in terms of this function.
5189 Perl_sv_catpvn_flags(pTHX_ SV *const dsv, const char *sstr, const STRLEN slen, const I32 flags)
5193 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
5195 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
5196 assert((flags & (SV_CATBYTES|SV_CATUTF8)) != (SV_CATBYTES|SV_CATUTF8));
5198 if (!(flags & SV_CATBYTES) || !SvUTF8(dsv)) {
5199 if (flags & SV_CATUTF8 && !SvUTF8(dsv)) {
5200 sv_utf8_upgrade_flags_grow(dsv, 0, slen + 1);
5203 else SvGROW(dsv, dlen + slen + 1);
5205 sstr = SvPVX_const(dsv);
5206 Move(sstr, SvPVX(dsv) + dlen, slen, char);
5207 SvCUR_set(dsv, SvCUR(dsv) + slen);
5210 /* We inline bytes_to_utf8, to avoid an extra malloc. */
5211 const char * const send = sstr + slen;
5214 /* Something this code does not account for, which I think is
5215 impossible; it would require the same pv to be treated as
5216 bytes *and* utf8, which would indicate a bug elsewhere. */
5217 assert(sstr != dstr);
5219 SvGROW(dsv, dlen + slen * 2 + 1);
5220 d = (U8 *)SvPVX(dsv) + dlen;
5222 while (sstr < send) {
5223 append_utf8_from_native_byte(*sstr, &d);
5226 SvCUR_set(dsv, d-(const U8 *)SvPVX(dsv));
5229 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
5231 if (flags & SV_SMAGIC)
5236 =for apidoc sv_catsv
5238 Concatenates the string from SV C<ssv> onto the end of the string in SV
5239 C<dsv>. If C<ssv> is null, does nothing; otherwise modifies only C<dsv>.
5240 Handles 'get' magic on both SVs, but no 'set' magic. See C<sv_catsv_mg> and
5243 =for apidoc sv_catsv_flags
5245 Concatenates the string from SV C<ssv> onto the end of the string in SV
5246 C<dsv>. If C<ssv> is null, does nothing; otherwise modifies only C<dsv>.
5247 If C<flags> include C<SV_GMAGIC> bit set, will call C<mg_get> on both SVs if
5248 appropriate. If C<flags> include C<SV_SMAGIC>, C<mg_set> will be called on
5249 the modified SV afterward, if appropriate. C<sv_catsv>, C<sv_catsv_nomg>,
5250 and C<sv_catsv_mg> are implemented in terms of this function.
5255 Perl_sv_catsv_flags(pTHX_ SV *const dsv, SV *const ssv, const I32 flags)
5259 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
5263 const char *spv = SvPV_flags_const(ssv, slen, flags);
5265 if (flags & SV_GMAGIC)
5267 sv_catpvn_flags(dsv, spv, slen,
5268 DO_UTF8(ssv) ? SV_CATUTF8 : SV_CATBYTES);
5269 if (flags & SV_SMAGIC)
5276 =for apidoc sv_catpv
5278 Concatenates the string onto the end of the string which is in the SV.
5279 If the SV has the UTF-8 status set, then the bytes appended should be
5280 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5285 Perl_sv_catpv(pTHX_ SV *const sv, const char *ptr)
5292 PERL_ARGS_ASSERT_SV_CATPV;
5296 junk = SvPV_force(sv, tlen);
5298 SvGROW(sv, tlen + len + 1);
5300 ptr = SvPVX_const(sv);
5301 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5302 SvCUR_set(sv, SvCUR(sv) + len);
5303 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5308 =for apidoc sv_catpv_flags
5310 Concatenates the string onto the end of the string which is in the SV.
5311 If the SV has the UTF-8 status set, then the bytes appended should
5312 be valid UTF-8. If C<flags> has the C<SV_SMAGIC> bit set, will C<mg_set>
5313 on the modified SV if appropriate.
5319 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5321 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5322 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5326 =for apidoc sv_catpv_mg
5328 Like C<sv_catpv>, but also handles 'set' magic.
5334 Perl_sv_catpv_mg(pTHX_ SV *const sv, const char *const ptr)
5336 PERL_ARGS_ASSERT_SV_CATPV_MG;
5345 Creates a new SV. A non-zero C<len> parameter indicates the number of
5346 bytes of preallocated string space the SV should have. An extra byte for a
5347 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5348 space is allocated.) The reference count for the new SV is set to 1.
5350 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5351 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5352 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5353 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5354 modules supporting older perls.
5360 Perl_newSV(pTHX_ const STRLEN len)
5367 sv_upgrade(sv, SVt_PV);
5368 SvGROW(sv, len + 1);
5373 =for apidoc sv_magicext
5375 Adds magic to an SV, upgrading it if necessary. Applies the
5376 supplied vtable and returns a pointer to the magic added.
5378 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5379 In particular, you can add magic to SvREADONLY SVs, and add more than
5380 one instance of the same 'how'.
5382 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5383 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5384 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5385 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5387 (This is now used as a subroutine by C<sv_magic>.)
5392 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5393 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5398 PERL_ARGS_ASSERT_SV_MAGICEXT;
5400 if (SvTYPE(sv)==SVt_PVAV) { assert (!AvPAD_NAMELIST(sv)); }
5402 SvUPGRADE(sv, SVt_PVMG);
5403 Newxz(mg, 1, MAGIC);
5404 mg->mg_moremagic = SvMAGIC(sv);
5405 SvMAGIC_set(sv, mg);
5407 /* Sometimes a magic contains a reference loop, where the sv and
5408 object refer to each other. To prevent a reference loop that
5409 would prevent such objects being freed, we look for such loops
5410 and if we find one we avoid incrementing the object refcount.
5412 Note we cannot do this to avoid self-tie loops as intervening RV must
5413 have its REFCNT incremented to keep it in existence.
5416 if (!obj || obj == sv ||
5417 how == PERL_MAGIC_arylen ||
5418 how == PERL_MAGIC_symtab ||
5419 (SvTYPE(obj) == SVt_PVGV &&
5420 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5421 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5422 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5427 mg->mg_obj = SvREFCNT_inc_simple(obj);
5428 mg->mg_flags |= MGf_REFCOUNTED;
5431 /* Normal self-ties simply pass a null object, and instead of
5432 using mg_obj directly, use the SvTIED_obj macro to produce a
5433 new RV as needed. For glob "self-ties", we are tieing the PVIO
5434 with an RV obj pointing to the glob containing the PVIO. In
5435 this case, to avoid a reference loop, we need to weaken the
5439 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5440 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5446 mg->mg_len = namlen;
5449 mg->mg_ptr = savepvn(name, namlen);
5450 else if (namlen == HEf_SVKEY) {
5451 /* Yes, this is casting away const. This is only for the case of
5452 HEf_SVKEY. I think we need to document this aberation of the
5453 constness of the API, rather than making name non-const, as
5454 that change propagating outwards a long way. */
5455 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5457 mg->mg_ptr = (char *) name;
5459 mg->mg_virtual = (MGVTBL *) vtable;
5466 Perl_sv_magicext_mglob(pTHX_ SV *sv)
5468 PERL_ARGS_ASSERT_SV_MAGICEXT_MGLOB;
5469 if (SvTYPE(sv) == SVt_PVLV && LvTYPE(sv) == 'y') {
5470 /* This sv is only a delegate. //g magic must be attached to
5475 #ifdef PERL_OLD_COPY_ON_WRITE
5477 sv_force_normal_flags(sv, 0);
5479 return sv_magicext(sv, NULL, PERL_MAGIC_regex_global,
5480 &PL_vtbl_mglob, 0, 0);
5484 =for apidoc sv_magic
5486 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if
5487 necessary, then adds a new magic item of type C<how> to the head of the
5490 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5491 handling of the C<name> and C<namlen> arguments.
5493 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5494 to add more than one instance of the same 'how'.
5500 Perl_sv_magic(pTHX_ SV *const sv, SV *const obj, const int how,
5501 const char *const name, const I32 namlen)
5504 const MGVTBL *vtable;
5507 unsigned int vtable_index;
5509 PERL_ARGS_ASSERT_SV_MAGIC;
5511 if (how < 0 || (unsigned)how > C_ARRAY_LENGTH(PL_magic_data)
5512 || ((flags = PL_magic_data[how]),
5513 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5514 > magic_vtable_max))
5515 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5517 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5518 Useful for attaching extension internal data to perl vars.
5519 Note that multiple extensions may clash if magical scalars
5520 etc holding private data from one are passed to another. */
5522 vtable = (vtable_index == magic_vtable_max)
5523 ? NULL : PL_magic_vtables + vtable_index;
5525 #ifdef PERL_OLD_COPY_ON_WRITE
5527 sv_force_normal_flags(sv, 0);
5529 if (SvREADONLY(sv)) {
5531 !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5534 Perl_croak_no_modify();
5537 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5538 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5539 /* sv_magic() refuses to add a magic of the same 'how' as an
5542 if (how == PERL_MAGIC_taint)
5548 /* Force pos to be stored as characters, not bytes. */
5549 if (SvMAGICAL(sv) && DO_UTF8(sv)
5550 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
5552 && mg->mg_flags & MGf_BYTES) {
5553 mg->mg_len = (SSize_t)sv_pos_b2u_flags(sv, (STRLEN)mg->mg_len,
5555 mg->mg_flags &= ~MGf_BYTES;
5558 /* Rest of work is done else where */
5559 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5562 case PERL_MAGIC_taint:
5565 case PERL_MAGIC_ext:
5566 case PERL_MAGIC_dbfile:
5573 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5580 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5582 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5583 for (mg = *mgp; mg; mg = *mgp) {
5584 const MGVTBL* const virt = mg->mg_virtual;
5585 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5586 *mgp = mg->mg_moremagic;
5587 if (virt && virt->svt_free)
5588 virt->svt_free(aTHX_ sv, mg);
5589 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5591 Safefree(mg->mg_ptr);
5592 else if (mg->mg_len == HEf_SVKEY)
5593 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5594 else if (mg->mg_type == PERL_MAGIC_utf8)
5595 Safefree(mg->mg_ptr);
5597 if (mg->mg_flags & MGf_REFCOUNTED)
5598 SvREFCNT_dec(mg->mg_obj);
5602 mgp = &mg->mg_moremagic;
5605 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5606 mg_magical(sv); /* else fix the flags now */
5610 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5616 =for apidoc sv_unmagic
5618 Removes all magic of type C<type> from an SV.
5624 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5626 PERL_ARGS_ASSERT_SV_UNMAGIC;
5627 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5631 =for apidoc sv_unmagicext
5633 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5639 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5641 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5642 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5646 =for apidoc sv_rvweaken
5648 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5649 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5650 push a back-reference to this RV onto the array of backreferences
5651 associated with that magic. If the RV is magical, set magic will be
5652 called after the RV is cleared.
5658 Perl_sv_rvweaken(pTHX_ SV *const sv)
5662 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5664 if (!SvOK(sv)) /* let undefs pass */
5667 Perl_croak(aTHX_ "Can't weaken a nonreference");
5668 else if (SvWEAKREF(sv)) {
5669 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5672 else if (SvREADONLY(sv)) croak_no_modify();
5674 Perl_sv_add_backref(aTHX_ tsv, sv);
5676 SvREFCNT_dec_NN(tsv);
5680 /* Give tsv backref magic if it hasn't already got it, then push a
5681 * back-reference to sv onto the array associated with the backref magic.
5683 * As an optimisation, if there's only one backref and it's not an AV,
5684 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5685 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5689 /* A discussion about the backreferences array and its refcount:
5691 * The AV holding the backreferences is pointed to either as the mg_obj of
5692 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5693 * xhv_backreferences field. The array is created with a refcount
5694 * of 2. This means that if during global destruction the array gets
5695 * picked on before its parent to have its refcount decremented by the
5696 * random zapper, it won't actually be freed, meaning it's still there for
5697 * when its parent gets freed.
5699 * When the parent SV is freed, the extra ref is killed by
5700 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5701 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5703 * When a single backref SV is stored directly, it is not reference
5708 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5715 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5717 /* find slot to store array or singleton backref */
5719 if (SvTYPE(tsv) == SVt_PVHV) {
5720 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5723 mg = mg_find(tsv, PERL_MAGIC_backref);
5725 mg = sv_magicext(tsv, NULL, PERL_MAGIC_backref, &PL_vtbl_backref, NULL, 0);
5726 svp = &(mg->mg_obj);
5729 /* create or retrieve the array */
5731 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5732 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5736 mg->mg_flags |= MGf_REFCOUNTED;
5739 SvREFCNT_inc_simple_void_NN(av);
5740 /* av now has a refcnt of 2; see discussion above */
5741 av_extend(av, *svp ? 2 : 1);
5743 /* move single existing backref to the array */
5744 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5749 av = MUTABLE_AV(*svp);
5751 /* optimisation: store single backref directly in HvAUX or mg_obj */
5755 assert(SvTYPE(av) == SVt_PVAV);
5756 if (AvFILLp(av) >= AvMAX(av)) {
5757 av_extend(av, AvFILLp(av)+1);
5760 /* push new backref */
5761 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5764 /* delete a back-reference to ourselves from the backref magic associated
5765 * with the SV we point to.
5769 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5774 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5776 if (SvTYPE(tsv) == SVt_PVHV) {
5778 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5780 else if (SvIS_FREED(tsv) && PL_phase == PERL_PHASE_DESTRUCT) {
5781 /* It's possible for the the last (strong) reference to tsv to have
5782 become freed *before* the last thing holding a weak reference.
5783 If both survive longer than the backreferences array, then when
5784 the referent's reference count drops to 0 and it is freed, it's
5785 not able to chase the backreferences, so they aren't NULLed.
5787 For example, a CV holds a weak reference to its stash. If both the
5788 CV and the stash survive longer than the backreferences array,
5789 and the CV gets picked for the SvBREAK() treatment first,
5790 *and* it turns out that the stash is only being kept alive because
5791 of an our variable in the pad of the CV, then midway during CV
5792 destruction the stash gets freed, but CvSTASH() isn't set to NULL.
5793 It ends up pointing to the freed HV. Hence it's chased in here, and
5794 if this block wasn't here, it would hit the !svp panic just below.
5796 I don't believe that "better" destruction ordering is going to help
5797 here - during global destruction there's always going to be the
5798 chance that something goes out of order. We've tried to make it
5799 foolproof before, and it only resulted in evolutionary pressure on
5800 fools. Which made us look foolish for our hubris. :-(
5806 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5807 svp = mg ? &(mg->mg_obj) : NULL;
5811 Perl_croak(aTHX_ "panic: del_backref, svp=0");
5813 /* It's possible that sv is being freed recursively part way through the
5814 freeing of tsv. If this happens, the backreferences array of tsv has
5815 already been freed, and so svp will be NULL. If this is the case,
5816 we should not panic. Instead, nothing needs doing, so return. */
5817 if (PL_phase == PERL_PHASE_DESTRUCT && SvREFCNT(tsv) == 0)
5819 Perl_croak(aTHX_ "panic: del_backref, *svp=%p phase=%s refcnt=%" UVuf,
5820 *svp, PL_phase_names[PL_phase], (UV)SvREFCNT(tsv));
5823 if (SvTYPE(*svp) == SVt_PVAV) {
5827 AV * const av = (AV*)*svp;
5829 assert(!SvIS_FREED(av));
5833 /* for an SV with N weak references to it, if all those
5834 * weak refs are deleted, then sv_del_backref will be called
5835 * N times and O(N^2) compares will be done within the backref
5836 * array. To ameliorate this potential slowness, we:
5837 * 1) make sure this code is as tight as possible;
5838 * 2) when looking for SV, look for it at both the head and tail of the
5839 * array first before searching the rest, since some create/destroy
5840 * patterns will cause the backrefs to be freed in order.
5847 SV **p = &svp[fill];
5848 SV *const topsv = *p;
5855 /* We weren't the last entry.
5856 An unordered list has this property that you
5857 can take the last element off the end to fill
5858 the hole, and it's still an unordered list :-)
5864 break; /* should only be one */
5871 AvFILLp(av) = fill-1;
5873 else if (SvIS_FREED(*svp) && PL_phase == PERL_PHASE_DESTRUCT) {
5874 /* freed AV; skip */
5877 /* optimisation: only a single backref, stored directly */
5879 Perl_croak(aTHX_ "panic: del_backref, *svp=%p, sv=%p", *svp, sv);
5886 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5892 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5897 /* after multiple passes through Perl_sv_clean_all() for a thingy
5898 * that has badly leaked, the backref array may have gotten freed,
5899 * since we only protect it against 1 round of cleanup */
5900 if (SvIS_FREED(av)) {
5901 if (PL_in_clean_all) /* All is fair */
5904 "panic: magic_killbackrefs (freed backref AV/SV)");
5908 is_array = (SvTYPE(av) == SVt_PVAV);
5910 assert(!SvIS_FREED(av));
5913 last = svp + AvFILLp(av);
5916 /* optimisation: only a single backref, stored directly */
5922 while (svp <= last) {
5924 SV *const referrer = *svp;
5925 if (SvWEAKREF(referrer)) {
5926 /* XXX Should we check that it hasn't changed? */
5927 assert(SvROK(referrer));
5928 SvRV_set(referrer, 0);
5930 SvWEAKREF_off(referrer);
5931 SvSETMAGIC(referrer);
5932 } else if (SvTYPE(referrer) == SVt_PVGV ||
5933 SvTYPE(referrer) == SVt_PVLV) {
5934 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5935 /* You lookin' at me? */
5936 assert(GvSTASH(referrer));
5937 assert(GvSTASH(referrer) == (const HV *)sv);
5938 GvSTASH(referrer) = 0;
5939 } else if (SvTYPE(referrer) == SVt_PVCV ||
5940 SvTYPE(referrer) == SVt_PVFM) {
5941 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5942 /* You lookin' at me? */
5943 assert(CvSTASH(referrer));
5944 assert(CvSTASH(referrer) == (const HV *)sv);
5945 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5948 assert(SvTYPE(sv) == SVt_PVGV);
5949 /* You lookin' at me? */
5950 assert(CvGV(referrer));
5951 assert(CvGV(referrer) == (const GV *)sv);
5952 anonymise_cv_maybe(MUTABLE_GV(sv),
5953 MUTABLE_CV(referrer));
5958 "panic: magic_killbackrefs (flags=%"UVxf")",
5959 (UV)SvFLAGS(referrer));
5970 SvREFCNT_dec_NN(av); /* remove extra count added by sv_add_backref() */
5976 =for apidoc sv_insert
5978 Inserts a string at the specified offset/length within the SV. Similar to
5979 the Perl substr() function. Handles get magic.
5981 =for apidoc sv_insert_flags
5983 Same as C<sv_insert>, but the extra C<flags> are passed to the
5984 C<SvPV_force_flags> that applies to C<bigstr>.
5990 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5997 SSize_t i; /* better be sizeof(STRLEN) or bad things happen */
6000 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
6003 Perl_croak(aTHX_ "Can't modify nonexistent substring");
6004 SvPV_force_flags(bigstr, curlen, flags);
6005 (void)SvPOK_only_UTF8(bigstr);
6006 if (offset + len > curlen) {
6007 SvGROW(bigstr, offset+len+1);
6008 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
6009 SvCUR_set(bigstr, offset+len);
6013 i = littlelen - len;
6014 if (i > 0) { /* string might grow */
6015 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
6016 mid = big + offset + len;
6017 midend = bigend = big + SvCUR(bigstr);
6020 while (midend > mid) /* shove everything down */
6021 *--bigend = *--midend;
6022 Move(little,big+offset,littlelen,char);
6023 SvCUR_set(bigstr, SvCUR(bigstr) + i);
6028 Move(little,SvPVX(bigstr)+offset,len,char);
6033 big = SvPVX(bigstr);
6036 bigend = big + SvCUR(bigstr);
6038 if (midend > bigend)
6039 Perl_croak(aTHX_ "panic: sv_insert, midend=%p, bigend=%p",
6042 if (mid - big > bigend - midend) { /* faster to shorten from end */
6044 Move(little, mid, littlelen,char);
6047 i = bigend - midend;
6049 Move(midend, mid, i,char);
6053 SvCUR_set(bigstr, mid - big);
6055 else if ((i = mid - big)) { /* faster from front */
6056 midend -= littlelen;
6058 Move(big, midend - i, i, char);
6059 sv_chop(bigstr,midend-i);
6061 Move(little, mid, littlelen,char);
6063 else if (littlelen) {
6064 midend -= littlelen;
6065 sv_chop(bigstr,midend);
6066 Move(little,midend,littlelen,char);
6069 sv_chop(bigstr,midend);
6075 =for apidoc sv_replace
6077 Make the first argument a copy of the second, then delete the original.
6078 The target SV physically takes over ownership of the body of the source SV
6079 and inherits its flags; however, the target keeps any magic it owns,
6080 and any magic in the source is discarded.
6081 Note that this is a rather specialist SV copying operation; most of the
6082 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
6088 Perl_sv_replace(pTHX_ SV *const sv, SV *const nsv)
6091 const U32 refcnt = SvREFCNT(sv);
6093 PERL_ARGS_ASSERT_SV_REPLACE;
6095 SV_CHECK_THINKFIRST_COW_DROP(sv);
6096 if (SvREFCNT(nsv) != 1) {
6097 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
6098 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
6100 if (SvMAGICAL(sv)) {
6104 sv_upgrade(nsv, SVt_PVMG);
6105 SvMAGIC_set(nsv, SvMAGIC(sv));
6106 SvFLAGS(nsv) |= SvMAGICAL(sv);
6108 SvMAGIC_set(sv, NULL);
6112 assert(!SvREFCNT(sv));
6113 #ifdef DEBUG_LEAKING_SCALARS
6114 sv->sv_flags = nsv->sv_flags;
6115 sv->sv_any = nsv->sv_any;
6116 sv->sv_refcnt = nsv->sv_refcnt;
6117 sv->sv_u = nsv->sv_u;
6119 StructCopy(nsv,sv,SV);
6121 if(SvTYPE(sv) == SVt_IV) {
6123 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
6127 #ifdef PERL_OLD_COPY_ON_WRITE
6128 if (SvIsCOW_normal(nsv)) {
6129 /* We need to follow the pointers around the loop to make the
6130 previous SV point to sv, rather than nsv. */
6133 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
6136 assert(SvPVX_const(current) == SvPVX_const(nsv));
6138 /* Make the SV before us point to the SV after us. */
6140 PerlIO_printf(Perl_debug_log, "previous is\n");
6142 PerlIO_printf(Perl_debug_log,
6143 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
6144 (UV) SV_COW_NEXT_SV(current), (UV) sv);
6146 SV_COW_NEXT_SV_SET(current, sv);
6149 SvREFCNT(sv) = refcnt;
6150 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
6155 /* We're about to free a GV which has a CV that refers back to us.
6156 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
6160 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
6165 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
6168 assert(SvREFCNT(gv) == 0);
6169 assert(isGV(gv) && isGV_with_GP(gv));
6171 assert(!CvANON(cv));
6172 assert(CvGV(cv) == gv);
6173 assert(!CvNAMED(cv));
6175 /* will the CV shortly be freed by gp_free() ? */
6176 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
6177 SvANY(cv)->xcv_gv_u.xcv_gv = NULL;
6181 /* if not, anonymise: */
6182 gvname = (GvSTASH(gv) && HvNAME(GvSTASH(gv)) && HvENAME(GvSTASH(gv)))
6183 ? newSVhek(HvENAME_HEK(GvSTASH(gv)))
6184 : newSVpvn_flags( "__ANON__", 8, 0 );
6185 sv_catpvs(gvname, "::__ANON__");
6186 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
6187 SvREFCNT_dec_NN(gvname);
6191 SvANY(cv)->xcv_gv_u.xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
6196 =for apidoc sv_clear
6198 Clear an SV: call any destructors, free up any memory used by the body,
6199 and free the body itself. The SV's head is I<not> freed, although
6200 its type is set to all 1's so that it won't inadvertently be assumed
6201 to be live during global destruction etc.
6202 This function should only be called when REFCNT is zero. Most of the time
6203 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
6210 Perl_sv_clear(pTHX_ SV *const orig_sv)
6215 const struct body_details *sv_type_details;
6221 PERL_ARGS_ASSERT_SV_CLEAR;
6223 /* within this loop, sv is the SV currently being freed, and
6224 * iter_sv is the most recent AV or whatever that's being iterated
6225 * over to provide more SVs */
6231 assert(SvREFCNT(sv) == 0);
6232 assert(SvTYPE(sv) != (svtype)SVTYPEMASK);
6234 if (type <= SVt_IV) {
6235 /* See the comment in sv.h about the collusion between this
6236 * early return and the overloading of the NULL slots in the
6240 SvFLAGS(sv) &= SVf_BREAK;
6241 SvFLAGS(sv) |= SVTYPEMASK;
6245 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
6247 if (type >= SVt_PVMG) {
6249 if (!curse(sv, 1)) goto get_next_sv;
6250 type = SvTYPE(sv); /* destructor may have changed it */
6252 /* Free back-references before magic, in case the magic calls
6253 * Perl code that has weak references to sv. */
6254 if (type == SVt_PVHV) {
6255 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
6259 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
6260 SvREFCNT_dec(SvOURSTASH(sv));
6262 else if (type == SVt_PVAV && AvPAD_NAMELIST(sv)) {
6263 assert(!SvMAGICAL(sv));
6264 } else if (SvMAGIC(sv)) {
6265 /* Free back-references before other types of magic. */
6266 sv_unmagic(sv, PERL_MAGIC_backref);
6270 if (type == SVt_PVMG && SvPAD_TYPED(sv))
6271 SvREFCNT_dec(SvSTASH(sv));
6274 /* case SVt_INVLIST: */
6277 IoIFP(sv) != PerlIO_stdin() &&
6278 IoIFP(sv) != PerlIO_stdout() &&
6279 IoIFP(sv) != PerlIO_stderr() &&
6280 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6282 io_close(MUTABLE_IO(sv), FALSE);
6284 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6285 PerlDir_close(IoDIRP(sv));
6286 IoDIRP(sv) = (DIR*)NULL;
6287 Safefree(IoTOP_NAME(sv));
6288 Safefree(IoFMT_NAME(sv));
6289 Safefree(IoBOTTOM_NAME(sv));
6290 if ((const GV *)sv == PL_statgv)
6294 /* FIXME for plugins */
6296 pregfree2((REGEXP*) sv);
6300 cv_undef(MUTABLE_CV(sv));
6301 /* If we're in a stash, we don't own a reference to it.
6302 * However it does have a back reference to us, which needs to
6304 if ((stash = CvSTASH(sv)))
6305 sv_del_backref(MUTABLE_SV(stash), sv);
6308 if (PL_last_swash_hv == (const HV *)sv) {
6309 PL_last_swash_hv = NULL;
6311 if (HvTOTALKEYS((HV*)sv) > 0) {
6313 /* this statement should match the one at the beginning of
6314 * hv_undef_flags() */
6315 if ( PL_phase != PERL_PHASE_DESTRUCT
6316 && (name = HvNAME((HV*)sv)))
6318 if (PL_stashcache) {
6319 DEBUG_o(Perl_deb(aTHX_ "sv_clear clearing PL_stashcache for '%"SVf"'\n",
6321 (void)hv_deletehek(PL_stashcache,
6322 HvNAME_HEK((HV*)sv), G_DISCARD);
6324 hv_name_set((HV*)sv, NULL, 0, 0);
6327 /* save old iter_sv in unused SvSTASH field */
6328 assert(!SvOBJECT(sv));
6329 SvSTASH(sv) = (HV*)iter_sv;
6332 /* save old hash_index in unused SvMAGIC field */
6333 assert(!SvMAGICAL(sv));
6334 assert(!SvMAGIC(sv));
6335 ((XPVMG*) SvANY(sv))->xmg_u.xmg_hash_index = hash_index;
6338 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6339 goto get_next_sv; /* process this new sv */
6341 /* free empty hash */
6342 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6343 assert(!HvARRAY((HV*)sv));
6347 AV* av = MUTABLE_AV(sv);
6348 if (PL_comppad == av) {
6352 if (AvREAL(av) && AvFILLp(av) > -1) {
6353 next_sv = AvARRAY(av)[AvFILLp(av)--];
6354 /* save old iter_sv in top-most slot of AV,
6355 * and pray that it doesn't get wiped in the meantime */
6356 AvARRAY(av)[AvMAX(av)] = iter_sv;
6358 goto get_next_sv; /* process this new sv */
6360 Safefree(AvALLOC(av));
6365 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6366 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6367 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6368 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6370 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6371 SvREFCNT_dec(LvTARG(sv));
6372 if (isREGEXP(sv)) goto freeregexp;
6374 if (isGV_with_GP(sv)) {
6375 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6376 && HvENAME_get(stash))
6377 mro_method_changed_in(stash);
6378 gp_free(MUTABLE_GV(sv));
6380 unshare_hek(GvNAME_HEK(sv));
6381 /* If we're in a stash, we don't own a reference to it.
6382 * However it does have a back reference to us, which
6383 * needs to be cleared. */
6384 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6385 sv_del_backref(MUTABLE_SV(stash), sv);
6387 /* FIXME. There are probably more unreferenced pointers to SVs
6388 * in the interpreter struct that we should check and tidy in
6389 * a similar fashion to this: */
6390 /* See also S_sv_unglob, which does the same thing. */
6391 if ((const GV *)sv == PL_last_in_gv)
6392 PL_last_in_gv = NULL;
6393 else if ((const GV *)sv == PL_statgv)
6395 else if ((const GV *)sv == PL_stderrgv)
6403 /* Don't bother with SvOOK_off(sv); as we're only going to
6407 SvOOK_offset(sv, offset);
6408 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6409 /* Don't even bother with turning off the OOK flag. */
6414 SV * const target = SvRV(sv);
6416 sv_del_backref(target, sv);
6422 else if (SvPVX_const(sv)
6423 && !(SvTYPE(sv) == SVt_PVIO
6424 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6428 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6432 # ifdef PERL_OLD_COPY_ON_WRITE
6433 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6435 if (CowREFCNT(sv)) {
6441 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6445 # ifdef PERL_OLD_COPY_ON_WRITE
6449 Safefree(SvPVX_mutable(sv));
6453 else if (SvPVX_const(sv) && SvLEN(sv)
6454 && !(SvTYPE(sv) == SVt_PVIO
6455 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6456 Safefree(SvPVX_mutable(sv));
6457 else if (SvPVX_const(sv) && SvIsCOW(sv)) {
6458 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6468 SvFLAGS(sv) &= SVf_BREAK;
6469 SvFLAGS(sv) |= SVTYPEMASK;
6471 sv_type_details = bodies_by_type + type;
6472 if (sv_type_details->arena) {
6473 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6474 &PL_body_roots[type]);
6476 else if (sv_type_details->body_size) {
6477 safefree(SvANY(sv));
6481 /* caller is responsible for freeing the head of the original sv */
6482 if (sv != orig_sv && !SvREFCNT(sv))
6485 /* grab and free next sv, if any */
6493 else if (!iter_sv) {
6495 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6496 AV *const av = (AV*)iter_sv;
6497 if (AvFILLp(av) > -1) {
6498 sv = AvARRAY(av)[AvFILLp(av)--];
6500 else { /* no more elements of current AV to free */
6503 /* restore previous value, squirrelled away */
6504 iter_sv = AvARRAY(av)[AvMAX(av)];
6505 Safefree(AvALLOC(av));
6508 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6509 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6510 if (!sv && !HvTOTALKEYS((HV *)iter_sv)) {
6511 /* no more elements of current HV to free */
6514 /* Restore previous values of iter_sv and hash_index,
6515 * squirrelled away */
6516 assert(!SvOBJECT(sv));
6517 iter_sv = (SV*)SvSTASH(sv);
6518 assert(!SvMAGICAL(sv));
6519 hash_index = ((XPVMG*) SvANY(sv))->xmg_u.xmg_hash_index;
6521 /* perl -DA does not like rubbish in SvMAGIC. */
6525 /* free any remaining detritus from the hash struct */
6526 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6527 assert(!HvARRAY((HV*)sv));
6532 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6536 if (!SvREFCNT(sv)) {
6540 if (--(SvREFCNT(sv)))
6544 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6545 "Attempt to free temp prematurely: SV 0x%"UVxf
6546 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6550 if (SvIMMORTAL(sv)) {
6551 /* make sure SvREFCNT(sv)==0 happens very seldom */
6552 SvREFCNT(sv) = SvREFCNT_IMMORTAL;
6561 /* This routine curses the sv itself, not the object referenced by sv. So
6562 sv does not have to be ROK. */
6565 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6568 PERL_ARGS_ASSERT_CURSE;
6569 assert(SvOBJECT(sv));
6571 if (PL_defstash && /* Still have a symbol table? */
6577 stash = SvSTASH(sv);
6578 assert(SvTYPE(stash) == SVt_PVHV);
6579 if (HvNAME(stash)) {
6580 CV* destructor = NULL;
6581 assert (SvOOK(stash));
6582 if (!SvOBJECT(stash)) destructor = (CV *)SvSTASH(stash);
6583 if (!destructor || HvMROMETA(stash)->destroy_gen
6584 != PL_sub_generation)
6587 gv_fetchmeth_autoload(stash, "DESTROY", 7, 0);
6588 if (gv) destructor = GvCV(gv);
6589 if (!SvOBJECT(stash))
6592 destructor ? (HV *)destructor : ((HV *)0)+1;
6593 HvAUX(stash)->xhv_mro_meta->destroy_gen =
6597 assert(!destructor || destructor == ((CV *)0)+1
6598 || SvTYPE(destructor) == SVt_PVCV);
6599 if (destructor && destructor != ((CV *)0)+1
6600 /* A constant subroutine can have no side effects, so
6601 don't bother calling it. */
6602 && !CvCONST(destructor)
6603 /* Don't bother calling an empty destructor or one that
6604 returns immediately. */
6605 && (CvISXSUB(destructor)
6606 || (CvSTART(destructor)
6607 && (CvSTART(destructor)->op_next->op_type
6609 && (CvSTART(destructor)->op_next->op_type
6611 || CvSTART(destructor)->op_next->op_next->op_type
6617 SV* const tmpref = newRV(sv);
6618 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6620 PUSHSTACKi(PERLSI_DESTROY);
6625 call_sv(MUTABLE_SV(destructor),
6626 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6630 if(SvREFCNT(tmpref) < 2) {
6631 /* tmpref is not kept alive! */
6633 SvRV_set(tmpref, NULL);
6636 SvREFCNT_dec_NN(tmpref);
6639 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6642 if (check_refcnt && SvREFCNT(sv)) {
6643 if (PL_in_clean_objs)
6645 "DESTROY created new reference to dead object '%"HEKf"'",
6646 HEKfARG(HvNAME_HEK(stash)));
6647 /* DESTROY gave object new lease on life */
6653 HV * const stash = SvSTASH(sv);
6654 /* Curse before freeing the stash, as freeing the stash could cause
6655 a recursive call into S_curse. */
6656 SvOBJECT_off(sv); /* Curse the object. */
6657 SvSTASH_set(sv,0); /* SvREFCNT_dec may try to read this */
6658 SvREFCNT_dec(stash); /* possibly of changed persuasion */
6664 =for apidoc sv_newref
6666 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6673 Perl_sv_newref(pTHX_ SV *const sv)
6675 PERL_UNUSED_CONTEXT;
6684 Decrement an SV's reference count, and if it drops to zero, call
6685 C<sv_clear> to invoke destructors and free up any memory used by
6686 the body; finally, deallocate the SV's head itself.
6687 Normally called via a wrapper macro C<SvREFCNT_dec>.
6693 Perl_sv_free(pTHX_ SV *const sv)
6699 /* Private helper function for SvREFCNT_dec().
6700 * Called with rc set to original SvREFCNT(sv), where rc == 0 or 1 */
6703 Perl_sv_free2(pTHX_ SV *const sv, const U32 rc)
6707 PERL_ARGS_ASSERT_SV_FREE2;
6709 if (LIKELY( rc == 1 )) {
6715 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6716 "Attempt to free temp prematurely: SV 0x%"UVxf
6717 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6721 if (SvIMMORTAL(sv)) {
6722 /* make sure SvREFCNT(sv)==0 happens very seldom */
6723 SvREFCNT(sv) = SvREFCNT_IMMORTAL;
6727 if (! SvREFCNT(sv)) /* may have have been resurrected */
6732 /* handle exceptional cases */
6736 if (SvFLAGS(sv) & SVf_BREAK)
6737 /* this SV's refcnt has been artificially decremented to
6738 * trigger cleanup */
6740 if (PL_in_clean_all) /* All is fair */
6742 if (SvIMMORTAL(sv)) {
6743 /* make sure SvREFCNT(sv)==0 happens very seldom */
6744 SvREFCNT(sv) = SvREFCNT_IMMORTAL;
6747 if (ckWARN_d(WARN_INTERNAL)) {
6748 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6749 Perl_dump_sv_child(aTHX_ sv);
6751 #ifdef DEBUG_LEAKING_SCALARS
6754 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6755 if (PL_warnhook == PERL_WARNHOOK_FATAL
6756 || ckDEAD(packWARN(WARN_INTERNAL))) {
6757 /* Don't let Perl_warner cause us to escape our fate: */
6761 /* This may not return: */
6762 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6763 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6764 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6767 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6777 Returns the length of the string in the SV. Handles magic and type
6778 coercion and sets the UTF8 flag appropriately. See also C<SvCUR>, which
6779 gives raw access to the xpv_cur slot.
6785 Perl_sv_len(pTHX_ SV *const sv)
6792 (void)SvPV_const(sv, len);
6797 =for apidoc sv_len_utf8
6799 Returns the number of characters in the string in an SV, counting wide
6800 UTF-8 bytes as a single character. Handles magic and type coercion.
6806 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6807 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6808 * (Note that the mg_len is not the length of the mg_ptr field.
6809 * This allows the cache to store the character length of the string without
6810 * needing to malloc() extra storage to attach to the mg_ptr.)
6815 Perl_sv_len_utf8(pTHX_ SV *const sv)
6821 return sv_len_utf8_nomg(sv);
6825 Perl_sv_len_utf8_nomg(pTHX_ SV * const sv)
6829 const U8 *s = (U8*)SvPV_nomg_const(sv, len);
6831 PERL_ARGS_ASSERT_SV_LEN_UTF8_NOMG;
6833 if (PL_utf8cache && SvUTF8(sv)) {
6835 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6837 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6838 if (mg->mg_len != -1)
6841 /* We can use the offset cache for a headstart.
6842 The longer value is stored in the first pair. */
6843 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6845 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6849 if (PL_utf8cache < 0) {
6850 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6851 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6855 ulen = Perl_utf8_length(aTHX_ s, s + len);
6856 utf8_mg_len_cache_update(sv, &mg, ulen);
6860 return SvUTF8(sv) ? Perl_utf8_length(aTHX_ s, s + len) : len;
6863 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6866 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6867 STRLEN *const uoffset_p, bool *const at_end)
6869 const U8 *s = start;
6870 STRLEN uoffset = *uoffset_p;
6872 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6874 while (s < send && uoffset) {
6881 else if (s > send) {
6883 /* This is the existing behaviour. Possibly it should be a croak, as
6884 it's actually a bounds error */
6887 *uoffset_p -= uoffset;
6891 /* Given the length of the string in both bytes and UTF-8 characters, decide
6892 whether to walk forwards or backwards to find the byte corresponding to
6893 the passed in UTF-8 offset. */
6895 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6896 STRLEN uoffset, const STRLEN uend)
6898 STRLEN backw = uend - uoffset;
6900 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6902 if (uoffset < 2 * backw) {
6903 /* The assumption is that going forwards is twice the speed of going
6904 forward (that's where the 2 * backw comes from).
6905 (The real figure of course depends on the UTF-8 data.) */
6906 const U8 *s = start;
6908 while (s < send && uoffset--)
6918 while (UTF8_IS_CONTINUATION(*send))
6921 return send - start;
6924 /* For the string representation of the given scalar, find the byte
6925 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6926 give another position in the string, *before* the sought offset, which
6927 (which is always true, as 0, 0 is a valid pair of positions), which should
6928 help reduce the amount of linear searching.
6929 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6930 will be used to reduce the amount of linear searching. The cache will be
6931 created if necessary, and the found value offered to it for update. */
6933 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6934 const U8 *const send, STRLEN uoffset,
6935 STRLEN uoffset0, STRLEN boffset0)
6937 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6939 bool at_end = FALSE;
6941 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6943 assert (uoffset >= uoffset0);
6948 if (!SvREADONLY(sv) && !SvGMAGICAL(sv) && SvPOK(sv)
6950 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6951 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6952 if ((*mgp)->mg_ptr) {
6953 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6954 if (cache[0] == uoffset) {
6955 /* An exact match. */
6958 if (cache[2] == uoffset) {
6959 /* An exact match. */
6963 if (cache[0] < uoffset) {
6964 /* The cache already knows part of the way. */
6965 if (cache[0] > uoffset0) {
6966 /* The cache knows more than the passed in pair */
6967 uoffset0 = cache[0];
6968 boffset0 = cache[1];
6970 if ((*mgp)->mg_len != -1) {
6971 /* And we know the end too. */
6973 + sv_pos_u2b_midway(start + boffset0, send,
6975 (*mgp)->mg_len - uoffset0);
6977 uoffset -= uoffset0;
6979 + sv_pos_u2b_forwards(start + boffset0,
6980 send, &uoffset, &at_end);
6981 uoffset += uoffset0;
6984 else if (cache[2] < uoffset) {
6985 /* We're between the two cache entries. */
6986 if (cache[2] > uoffset0) {
6987 /* and the cache knows more than the passed in pair */
6988 uoffset0 = cache[2];
6989 boffset0 = cache[3];
6993 + sv_pos_u2b_midway(start + boffset0,
6996 cache[0] - uoffset0);
6999 + sv_pos_u2b_midway(start + boffset0,
7002 cache[2] - uoffset0);
7006 else if ((*mgp)->mg_len != -1) {
7007 /* If we can take advantage of a passed in offset, do so. */
7008 /* In fact, offset0 is either 0, or less than offset, so don't
7009 need to worry about the other possibility. */
7011 + sv_pos_u2b_midway(start + boffset0, send,
7013 (*mgp)->mg_len - uoffset0);
7018 if (!found || PL_utf8cache < 0) {
7019 STRLEN real_boffset;
7020 uoffset -= uoffset0;
7021 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
7022 send, &uoffset, &at_end);
7023 uoffset += uoffset0;
7025 if (found && PL_utf8cache < 0)
7026 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
7028 boffset = real_boffset;
7031 if (PL_utf8cache && !SvGMAGICAL(sv) && SvPOK(sv)) {
7033 utf8_mg_len_cache_update(sv, mgp, uoffset);
7035 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
7042 =for apidoc sv_pos_u2b_flags
7044 Converts the offset from a count of UTF-8 chars from
7045 the start of the string, to a count of the equivalent number of bytes; if
7046 lenp is non-zero, it does the same to lenp, but this time starting from
7047 the offset, rather than from the start
7048 of the string. Handles type coercion.
7049 I<flags> is passed to C<SvPV_flags>, and usually should be
7050 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
7056 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
7057 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7058 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
7063 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
7070 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
7072 start = (U8*)SvPV_flags(sv, len, flags);
7074 const U8 * const send = start + len;
7076 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
7079 && *lenp /* don't bother doing work for 0, as its bytes equivalent
7080 is 0, and *lenp is already set to that. */) {
7081 /* Convert the relative offset to absolute. */
7082 const STRLEN uoffset2 = uoffset + *lenp;
7083 const STRLEN boffset2
7084 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
7085 uoffset, boffset) - boffset;
7099 =for apidoc sv_pos_u2b
7101 Converts the value pointed to by offsetp from a count of UTF-8 chars from
7102 the start of the string, to a count of the equivalent number of bytes; if
7103 lenp is non-zero, it does the same to lenp, but this time starting from
7104 the offset, rather than from the start of the string. Handles magic and
7107 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
7114 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
7115 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7116 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
7120 /* This function is subject to size and sign problems */
7123 Perl_sv_pos_u2b(pTHX_ SV *const sv, I32 *const offsetp, I32 *const lenp)
7125 PERL_ARGS_ASSERT_SV_POS_U2B;
7128 STRLEN ulen = (STRLEN)*lenp;
7129 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
7130 SV_GMAGIC|SV_CONST_RETURN);
7133 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
7134 SV_GMAGIC|SV_CONST_RETURN);
7139 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
7142 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
7143 if (SvREADONLY(sv) || SvGMAGICAL(sv) || !SvPOK(sv))
7146 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
7147 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
7148 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
7152 (*mgp)->mg_len = ulen;
7155 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
7156 byte length pairing. The (byte) length of the total SV is passed in too,
7157 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
7158 may not have updated SvCUR, so we can't rely on reading it directly.
7160 The proffered utf8/byte length pairing isn't used if the cache already has
7161 two pairs, and swapping either for the proffered pair would increase the
7162 RMS of the intervals between known byte offsets.
7164 The cache itself consists of 4 STRLEN values
7165 0: larger UTF-8 offset
7166 1: corresponding byte offset
7167 2: smaller UTF-8 offset
7168 3: corresponding byte offset
7170 Unused cache pairs have the value 0, 0.
7171 Keeping the cache "backwards" means that the invariant of
7172 cache[0] >= cache[2] is maintained even with empty slots, which means that
7173 the code that uses it doesn't need to worry if only 1 entry has actually
7174 been set to non-zero. It also makes the "position beyond the end of the
7175 cache" logic much simpler, as the first slot is always the one to start
7179 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
7180 const STRLEN utf8, const STRLEN blen)
7184 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
7189 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
7190 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
7191 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
7193 (*mgp)->mg_len = -1;
7197 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
7198 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
7199 (*mgp)->mg_ptr = (char *) cache;
7203 if (PL_utf8cache < 0 && SvPOKp(sv)) {
7204 /* SvPOKp() because it's possible that sv has string overloading, and
7205 therefore is a reference, hence SvPVX() is actually a pointer.
7206 This cures the (very real) symptoms of RT 69422, but I'm not actually
7207 sure whether we should even be caching the results of UTF-8
7208 operations on overloading, given that nothing stops overloading
7209 returning a different value every time it's called. */
7210 const U8 *start = (const U8 *) SvPVX_const(sv);
7211 const STRLEN realutf8 = utf8_length(start, start + byte);
7213 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
7217 /* Cache is held with the later position first, to simplify the code
7218 that deals with unbounded ends. */
7220 ASSERT_UTF8_CACHE(cache);
7221 if (cache[1] == 0) {
7222 /* Cache is totally empty */
7225 } else if (cache[3] == 0) {
7226 if (byte > cache[1]) {
7227 /* New one is larger, so goes first. */
7228 cache[2] = cache[0];
7229 cache[3] = cache[1];
7237 #define THREEWAY_SQUARE(a,b,c,d) \
7238 ((float)((d) - (c))) * ((float)((d) - (c))) \
7239 + ((float)((c) - (b))) * ((float)((c) - (b))) \
7240 + ((float)((b) - (a))) * ((float)((b) - (a)))
7242 /* Cache has 2 slots in use, and we know three potential pairs.
7243 Keep the two that give the lowest RMS distance. Do the
7244 calculation in bytes simply because we always know the byte
7245 length. squareroot has the same ordering as the positive value,
7246 so don't bother with the actual square root. */
7247 if (byte > cache[1]) {
7248 /* New position is after the existing pair of pairs. */
7249 const float keep_earlier
7250 = THREEWAY_SQUARE(0, cache[3], byte, blen);
7251 const float keep_later
7252 = THREEWAY_SQUARE(0, cache[1], byte, blen);
7254 if (keep_later < keep_earlier) {
7255 cache[2] = cache[0];
7256 cache[3] = cache[1];
7265 else if (byte > cache[3]) {
7266 /* New position is between the existing pair of pairs. */
7267 const float keep_earlier
7268 = THREEWAY_SQUARE(0, cache[3], byte, blen);
7269 const float keep_later
7270 = THREEWAY_SQUARE(0, byte, cache[1], blen);
7272 if (keep_later < keep_earlier) {
7282 /* New position is before the existing pair of pairs. */
7283 const float keep_earlier
7284 = THREEWAY_SQUARE(0, byte, cache[3], blen);
7285 const float keep_later
7286 = THREEWAY_SQUARE(0, byte, cache[1], blen);
7288 if (keep_later < keep_earlier) {
7293 cache[0] = cache[2];
7294 cache[1] = cache[3];
7300 ASSERT_UTF8_CACHE(cache);
7303 /* We already know all of the way, now we may be able to walk back. The same
7304 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
7305 backward is half the speed of walking forward. */
7307 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
7308 const U8 *end, STRLEN endu)
7310 const STRLEN forw = target - s;
7311 STRLEN backw = end - target;
7313 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
7315 if (forw < 2 * backw) {
7316 return utf8_length(s, target);
7319 while (end > target) {
7321 while (UTF8_IS_CONTINUATION(*end)) {
7330 =for apidoc sv_pos_b2u_flags
7332 Converts the offset from a count of bytes from the start of the string, to
7333 a count of the equivalent number of UTF-8 chars. Handles type coercion.
7334 I<flags> is passed to C<SvPV_flags>, and usually should be
7335 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
7341 * sv_pos_b2u_flags() uses, like sv_pos_u2b_flags(), the mg_ptr of the
7342 * potential PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8
7347 Perl_sv_pos_b2u_flags(pTHX_ SV *const sv, STRLEN const offset, U32 flags)
7350 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
7356 PERL_ARGS_ASSERT_SV_POS_B2U_FLAGS;
7358 s = (const U8*)SvPV_flags(sv, blen, flags);
7361 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset, blen=%"UVuf
7362 ", byte=%"UVuf, (UV)blen, (UV)offset);
7368 && SvTYPE(sv) >= SVt_PVMG
7369 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7372 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7373 if (cache[1] == offset) {
7374 /* An exact match. */
7377 if (cache[3] == offset) {
7378 /* An exact match. */
7382 if (cache[1] < offset) {
7383 /* We already know part of the way. */
7384 if (mg->mg_len != -1) {
7385 /* Actually, we know the end too. */
7387 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7388 s + blen, mg->mg_len - cache[0]);
7390 len = cache[0] + utf8_length(s + cache[1], send);
7393 else if (cache[3] < offset) {
7394 /* We're between the two cached pairs, so we do the calculation
7395 offset by the byte/utf-8 positions for the earlier pair,
7396 then add the utf-8 characters from the string start to
7398 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7399 s + cache[1], cache[0] - cache[2])
7403 else { /* cache[3] > offset */
7404 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7408 ASSERT_UTF8_CACHE(cache);
7410 } else if (mg->mg_len != -1) {
7411 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7415 if (!found || PL_utf8cache < 0) {
7416 const STRLEN real_len = utf8_length(s, send);
7418 if (found && PL_utf8cache < 0)
7419 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7425 utf8_mg_len_cache_update(sv, &mg, len);
7427 utf8_mg_pos_cache_update(sv, &mg, offset, len, blen);
7434 =for apidoc sv_pos_b2u
7436 Converts the value pointed to by offsetp from a count of bytes from the
7437 start of the string, to a count of the equivalent number of UTF-8 chars.
7438 Handles magic and type coercion.
7440 Use C<sv_pos_b2u_flags> in preference, which correctly handles strings
7447 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
7448 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7453 Perl_sv_pos_b2u(pTHX_ SV *const sv, I32 *const offsetp)
7455 PERL_ARGS_ASSERT_SV_POS_B2U;
7460 *offsetp = (I32)sv_pos_b2u_flags(sv, (STRLEN)*offsetp,
7461 SV_GMAGIC|SV_CONST_RETURN);
7465 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7466 STRLEN real, SV *const sv)
7468 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7470 /* As this is debugging only code, save space by keeping this test here,
7471 rather than inlining it in all the callers. */
7472 if (from_cache == real)
7475 /* Need to turn the assertions off otherwise we may recurse infinitely
7476 while printing error messages. */
7477 SAVEI8(PL_utf8cache);
7479 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7480 func, (UV) from_cache, (UV) real, SVfARG(sv));
7486 Returns a boolean indicating whether the strings in the two SVs are
7487 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7488 coerce its args to strings if necessary.
7490 =for apidoc sv_eq_flags
7492 Returns a boolean indicating whether the strings in the two SVs are
7493 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7494 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7500 Perl_sv_eq_flags(pTHX_ SV *sv1, SV *sv2, const U32 flags)
7508 SV* svrecode = NULL;
7515 /* if pv1 and pv2 are the same, second SvPV_const call may
7516 * invalidate pv1 (if we are handling magic), so we may need to
7518 if (sv1 == sv2 && flags & SV_GMAGIC
7519 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7520 pv1 = SvPV_const(sv1, cur1);
7521 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7523 pv1 = SvPV_flags_const(sv1, cur1, flags);
7531 pv2 = SvPV_flags_const(sv2, cur2, flags);
7533 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7534 /* Differing utf8ness.
7535 * Do not UTF8size the comparands as a side-effect. */
7538 svrecode = newSVpvn(pv2, cur2);
7539 sv_recode_to_utf8(svrecode, PL_encoding);
7540 pv2 = SvPV_const(svrecode, cur2);
7543 svrecode = newSVpvn(pv1, cur1);
7544 sv_recode_to_utf8(svrecode, PL_encoding);
7545 pv1 = SvPV_const(svrecode, cur1);
7547 /* Now both are in UTF-8. */
7549 SvREFCNT_dec_NN(svrecode);
7555 /* sv1 is the UTF-8 one */
7556 return bytes_cmp_utf8((const U8*)pv2, cur2,
7557 (const U8*)pv1, cur1) == 0;
7560 /* sv2 is the UTF-8 one */
7561 return bytes_cmp_utf8((const U8*)pv1, cur1,
7562 (const U8*)pv2, cur2) == 0;
7568 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7570 SvREFCNT_dec(svrecode);
7578 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7579 string in C<sv1> is less than, equal to, or greater than the string in
7580 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7581 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7583 =for apidoc sv_cmp_flags
7585 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7586 string in C<sv1> is less than, equal to, or greater than the string in
7587 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7588 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7589 also C<sv_cmp_locale_flags>.
7595 Perl_sv_cmp(pTHX_ SV *const sv1, SV *const sv2)
7597 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7601 Perl_sv_cmp_flags(pTHX_ SV *const sv1, SV *const sv2,
7606 const char *pv1, *pv2;
7608 SV *svrecode = NULL;
7615 pv1 = SvPV_flags_const(sv1, cur1, flags);
7622 pv2 = SvPV_flags_const(sv2, cur2, flags);
7624 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7625 /* Differing utf8ness.
7626 * Do not UTF8size the comparands as a side-effect. */
7629 svrecode = newSVpvn(pv2, cur2);
7630 sv_recode_to_utf8(svrecode, PL_encoding);
7631 pv2 = SvPV_const(svrecode, cur2);
7634 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7635 (const U8*)pv1, cur1);
7636 return retval ? retval < 0 ? -1 : +1 : 0;
7641 svrecode = newSVpvn(pv1, cur1);
7642 sv_recode_to_utf8(svrecode, PL_encoding);
7643 pv1 = SvPV_const(svrecode, cur1);
7646 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7647 (const U8*)pv2, cur2);
7648 return retval ? retval < 0 ? -1 : +1 : 0;
7654 cmp = cur2 ? -1 : 0;
7658 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7661 cmp = retval < 0 ? -1 : 1;
7662 } else if (cur1 == cur2) {
7665 cmp = cur1 < cur2 ? -1 : 1;
7669 SvREFCNT_dec(svrecode);
7675 =for apidoc sv_cmp_locale
7677 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7678 'use bytes' aware, handles get magic, and will coerce its args to strings
7679 if necessary. See also C<sv_cmp>.
7681 =for apidoc sv_cmp_locale_flags
7683 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7684 'use bytes' aware and will coerce its args to strings if necessary. If the
7685 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7691 Perl_sv_cmp_locale(pTHX_ SV *const sv1, SV *const sv2)
7693 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7697 Perl_sv_cmp_locale_flags(pTHX_ SV *const sv1, SV *const sv2,
7701 #ifdef USE_LOCALE_COLLATE
7707 if (PL_collation_standard)
7711 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7713 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7715 if (!pv1 || !len1) {
7726 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7729 return retval < 0 ? -1 : 1;
7732 * When the result of collation is equality, that doesn't mean
7733 * that there are no differences -- some locales exclude some
7734 * characters from consideration. So to avoid false equalities,
7735 * we use the raw string as a tiebreaker.
7741 #endif /* USE_LOCALE_COLLATE */
7743 return sv_cmp(sv1, sv2);
7747 #ifdef USE_LOCALE_COLLATE
7750 =for apidoc sv_collxfrm
7752 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7753 C<sv_collxfrm_flags>.
7755 =for apidoc sv_collxfrm_flags
7757 Add Collate Transform magic to an SV if it doesn't already have it. If the
7758 flags contain SV_GMAGIC, it handles get-magic.
7760 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7761 scalar data of the variable, but transformed to such a format that a normal
7762 memory comparison can be used to compare the data according to the locale
7769 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7774 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7776 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7777 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7783 Safefree(mg->mg_ptr);
7784 s = SvPV_flags_const(sv, len, flags);
7785 if ((xf = mem_collxfrm(s, len, &xlen))) {
7787 #ifdef PERL_OLD_COPY_ON_WRITE
7789 sv_force_normal_flags(sv, 0);
7791 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7805 if (mg && mg->mg_ptr) {
7807 return mg->mg_ptr + sizeof(PL_collation_ix);
7815 #endif /* USE_LOCALE_COLLATE */
7818 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7820 SV * const tsv = newSV(0);
7823 sv_gets(tsv, fp, 0);
7824 sv_utf8_upgrade_nomg(tsv);
7825 SvCUR_set(sv,append);
7828 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7832 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7835 const STRLEN recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7836 /* Grab the size of the record we're getting */
7837 char *buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7845 /* With a true, record-oriented file on VMS, we need to use read directly
7846 * to ensure that we respect RMS record boundaries. The user is responsible
7847 * for providing a PL_rs value that corresponds to the FAB$W_MRS (maximum
7848 * record size) field. N.B. This is likely to produce invalid results on
7849 * varying-width character data when a record ends mid-character.
7851 fd = PerlIO_fileno(fp);
7853 && PerlLIO_fstat(fd, &st) == 0
7854 && (st.st_fab_rfm == FAB$C_VAR
7855 || st.st_fab_rfm == FAB$C_VFC
7856 || st.st_fab_rfm == FAB$C_FIX)) {
7858 bytesread = PerlLIO_read(fd, buffer, recsize);
7860 else /* in-memory file from PerlIO::Scalar
7861 * or not a record-oriented file
7865 bytesread = PerlIO_read(fp, buffer, recsize);
7867 /* At this point, the logic in sv_get() means that sv will
7868 be treated as utf-8 if the handle is utf8.
7870 if (PerlIO_isutf8(fp) && bytesread > 0) {
7871 char *bend = buffer + bytesread;
7872 char *bufp = buffer;
7873 size_t charcount = 0;
7874 bool charstart = TRUE;
7877 while (charcount < recsize) {
7878 /* count accumulated characters */
7879 while (bufp < bend) {
7881 skip = UTF8SKIP(bufp);
7883 if (bufp + skip > bend) {
7884 /* partial at the end */
7895 if (charcount < recsize) {
7897 STRLEN bufp_offset = bufp - buffer;
7898 SSize_t morebytesread;
7900 /* originally I read enough to fill any incomplete
7901 character and the first byte of the next
7902 character if needed, but if there's many
7903 multi-byte encoded characters we're going to be
7904 making a read call for every character beyond
7905 the original read size.
7907 So instead, read the rest of the character if
7908 any, and enough bytes to match at least the
7909 start bytes for each character we're going to
7913 readsize = recsize - charcount;
7915 readsize = skip - (bend - bufp) + recsize - charcount - 1;
7916 buffer = SvGROW(sv, append + bytesread + readsize + 1) + append;
7917 bend = buffer + bytesread;
7918 morebytesread = PerlIO_read(fp, bend, readsize);
7919 if (morebytesread <= 0) {
7920 /* we're done, if we still have incomplete
7921 characters the check code in sv_gets() will
7924 I'd originally considered doing
7925 PerlIO_ungetc() on all but the lead
7926 character of the incomplete character, but
7927 read() doesn't do that, so I don't.
7932 /* prepare to scan some more */
7933 bytesread += morebytesread;
7934 bend = buffer + bytesread;
7935 bufp = buffer + bufp_offset;
7943 SvCUR_set(sv, bytesread + append);
7944 buffer[bytesread] = '\0';
7945 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7951 Get a line from the filehandle and store it into the SV, optionally
7952 appending to the currently-stored string. If C<append> is not 0, the
7953 line is appended to the SV instead of overwriting it. C<append> should
7954 be set to the byte offset that the appended string should start at
7955 in the SV (typically, C<SvCUR(sv)> is a suitable choice).
7961 Perl_sv_gets(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7972 PERL_ARGS_ASSERT_SV_GETS;
7974 if (SvTHINKFIRST(sv))
7975 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7976 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7978 However, perlbench says it's slower, because the existing swipe code
7979 is faster than copy on write.
7980 Swings and roundabouts. */
7981 SvUPGRADE(sv, SVt_PV);
7984 if (PerlIO_isutf8(fp)) {
7986 sv_utf8_upgrade_nomg(sv);
7987 sv_pos_u2b(sv,&append,0);
7989 } else if (SvUTF8(sv)) {
7990 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7998 if (PerlIO_isutf8(fp))
8001 if (IN_PERL_COMPILETIME) {
8002 /* we always read code in line mode */
8006 else if (RsSNARF(PL_rs)) {
8007 /* If it is a regular disk file use size from stat() as estimate
8008 of amount we are going to read -- may result in mallocing
8009 more memory than we really need if the layers below reduce
8010 the size we read (e.g. CRLF or a gzip layer).
8013 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
8014 const Off_t offset = PerlIO_tell(fp);
8015 if (offset != (Off_t) -1 && st.st_size + append > offset) {
8016 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
8022 else if (RsRECORD(PL_rs)) {
8023 return S_sv_gets_read_record(aTHX_ sv, fp, append);
8025 else if (RsPARA(PL_rs)) {
8031 /* Get $/ i.e. PL_rs into same encoding as stream wants */
8032 if (PerlIO_isutf8(fp)) {
8033 rsptr = SvPVutf8(PL_rs, rslen);
8036 if (SvUTF8(PL_rs)) {
8037 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
8038 Perl_croak(aTHX_ "Wide character in $/");
8041 rsptr = SvPV_const(PL_rs, rslen);
8045 rslast = rslen ? rsptr[rslen - 1] : '\0';
8047 if (rspara) { /* have to do this both before and after */
8048 do { /* to make sure file boundaries work right */
8051 i = PerlIO_getc(fp);
8055 PerlIO_ungetc(fp,i);
8061 /* See if we know enough about I/O mechanism to cheat it ! */
8063 /* This used to be #ifdef test - it is made run-time test for ease
8064 of abstracting out stdio interface. One call should be cheap
8065 enough here - and may even be a macro allowing compile
8069 if (PerlIO_fast_gets(fp)) {
8072 * We're going to steal some values from the stdio struct
8073 * and put EVERYTHING in the innermost loop into registers.
8079 #if defined(VMS) && defined(PERLIO_IS_STDIO)
8080 /* An ungetc()d char is handled separately from the regular
8081 * buffer, so we getc() it back out and stuff it in the buffer.
8083 i = PerlIO_getc(fp);
8084 if (i == EOF) return 0;
8085 *(--((*fp)->_ptr)) = (unsigned char) i;
8089 /* Here is some breathtakingly efficient cheating */
8091 cnt = PerlIO_get_cnt(fp); /* get count into register */
8092 /* make sure we have the room */
8093 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
8094 /* Not room for all of it
8095 if we are looking for a separator and room for some
8097 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
8098 /* just process what we have room for */
8099 shortbuffered = cnt - SvLEN(sv) + append + 1;
8100 cnt -= shortbuffered;
8104 /* remember that cnt can be negative */
8105 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
8110 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
8111 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
8112 DEBUG_P(PerlIO_printf(Perl_debug_log,
8113 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
8114 DEBUG_P(PerlIO_printf(Perl_debug_log,
8115 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%zd, base=%"
8117 PTR2UV(PerlIO_get_ptr(fp)), PerlIO_get_cnt(fp),
8118 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
8123 while (cnt > 0) { /* this | eat */
8125 if ((*bp++ = *ptr++) == rslast) /* really | dust */
8126 goto thats_all_folks; /* screams | sed :-) */
8130 Copy(ptr, bp, cnt, char); /* this | eat */
8131 bp += cnt; /* screams | dust */
8132 ptr += cnt; /* louder | sed :-) */
8134 assert (!shortbuffered);
8135 goto cannot_be_shortbuffered;
8139 if (shortbuffered) { /* oh well, must extend */
8140 cnt = shortbuffered;
8142 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
8144 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
8145 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
8149 cannot_be_shortbuffered:
8150 DEBUG_P(PerlIO_printf(Perl_debug_log,
8151 "Screamer: going to getc, ptr=%"UVuf", cnt=%zd\n",
8153 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
8155 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
8156 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%zd, base=%"UVuf"\n",
8157 PTR2UV(PerlIO_get_ptr(fp)), PerlIO_get_cnt(fp),
8158 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
8160 /* This used to call 'filbuf' in stdio form, but as that behaves like
8161 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
8162 another abstraction. */
8163 i = PerlIO_getc(fp); /* get more characters */
8165 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
8166 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%zd, base=%"UVuf"\n",
8167 PTR2UV(PerlIO_get_ptr(fp)), PerlIO_get_cnt(fp),
8168 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
8170 cnt = PerlIO_get_cnt(fp);
8171 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
8172 DEBUG_P(PerlIO_printf(Perl_debug_log,
8173 "Screamer: after getc, ptr=%"UVuf", cnt=%zd\n",
8176 if (i == EOF) /* all done for ever? */
8177 goto thats_really_all_folks;
8179 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
8181 SvGROW(sv, bpx + cnt + 2);
8182 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
8184 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
8186 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
8187 goto thats_all_folks;
8191 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
8192 memNE((char*)bp - rslen, rsptr, rslen))
8193 goto screamer; /* go back to the fray */
8194 thats_really_all_folks:
8196 cnt += shortbuffered;
8197 DEBUG_P(PerlIO_printf(Perl_debug_log,
8198 "Screamer: quitting, ptr=%"UVuf", cnt=%zd\n",PTR2UV(ptr),cnt));
8199 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
8200 DEBUG_P(PerlIO_printf(Perl_debug_log,
8201 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%zd, base=%"UVuf
8203 PTR2UV(PerlIO_get_ptr(fp)), PerlIO_get_cnt(fp),
8204 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
8206 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
8207 DEBUG_P(PerlIO_printf(Perl_debug_log,
8208 "Screamer: done, len=%ld, string=|%.*s|\n",
8209 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
8213 /*The big, slow, and stupid way. */
8214 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
8215 STDCHAR *buf = NULL;
8216 Newx(buf, 8192, STDCHAR);
8224 const STDCHAR * const bpe = buf + sizeof(buf);
8226 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
8227 ; /* keep reading */
8231 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
8232 /* Accommodate broken VAXC compiler, which applies U8 cast to
8233 * both args of ?: operator, causing EOF to change into 255
8236 i = (U8)buf[cnt - 1];
8242 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
8244 sv_catpvn_nomg(sv, (char *) buf, cnt);
8246 sv_setpvn(sv, (char *) buf, cnt); /* "nomg" is implied */
8248 if (i != EOF && /* joy */
8250 SvCUR(sv) < rslen ||
8251 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
8255 * If we're reading from a TTY and we get a short read,
8256 * indicating that the user hit his EOF character, we need
8257 * to notice it now, because if we try to read from the TTY
8258 * again, the EOF condition will disappear.
8260 * The comparison of cnt to sizeof(buf) is an optimization
8261 * that prevents unnecessary calls to feof().
8265 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
8269 #ifdef USE_HEAP_INSTEAD_OF_STACK
8274 if (rspara) { /* have to do this both before and after */
8275 while (i != EOF) { /* to make sure file boundaries work right */
8276 i = PerlIO_getc(fp);
8278 PerlIO_ungetc(fp,i);
8284 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
8290 Auto-increment of the value in the SV, doing string to numeric conversion
8291 if necessary. Handles 'get' magic and operator overloading.
8297 Perl_sv_inc(pTHX_ SV *const sv)
8306 =for apidoc sv_inc_nomg
8308 Auto-increment of the value in the SV, doing string to numeric conversion
8309 if necessary. Handles operator overloading. Skips handling 'get' magic.
8315 Perl_sv_inc_nomg(pTHX_ SV *const sv)
8323 if (SvTHINKFIRST(sv)) {
8324 if (SvREADONLY(sv)) {
8325 Perl_croak_no_modify();
8329 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
8331 i = PTR2IV(SvRV(sv));
8335 else sv_force_normal_flags(sv, 0);
8337 flags = SvFLAGS(sv);
8338 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
8339 /* It's (privately or publicly) a float, but not tested as an
8340 integer, so test it to see. */
8342 flags = SvFLAGS(sv);
8344 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8345 /* It's publicly an integer, or privately an integer-not-float */
8346 #ifdef PERL_PRESERVE_IVUV
8350 if (SvUVX(sv) == UV_MAX)
8351 sv_setnv(sv, UV_MAX_P1);
8353 (void)SvIOK_only_UV(sv);
8354 SvUV_set(sv, SvUVX(sv) + 1);
8356 if (SvIVX(sv) == IV_MAX)
8357 sv_setuv(sv, (UV)IV_MAX + 1);
8359 (void)SvIOK_only(sv);
8360 SvIV_set(sv, SvIVX(sv) + 1);
8365 if (flags & SVp_NOK) {
8366 const NV was = SvNVX(sv);
8367 if (NV_OVERFLOWS_INTEGERS_AT &&
8368 was >= NV_OVERFLOWS_INTEGERS_AT) {
8369 /* diag_listed_as: Lost precision when %s %f by 1 */
8370 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8371 "Lost precision when incrementing %" NVff " by 1",
8374 (void)SvNOK_only(sv);
8375 SvNV_set(sv, was + 1.0);
8379 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
8380 if ((flags & SVTYPEMASK) < SVt_PVIV)
8381 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
8382 (void)SvIOK_only(sv);
8387 while (isALPHA(*d)) d++;
8388 while (isDIGIT(*d)) d++;
8389 if (d < SvEND(sv)) {
8390 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8391 #ifdef PERL_PRESERVE_IVUV
8392 /* Got to punt this as an integer if needs be, but we don't issue
8393 warnings. Probably ought to make the sv_iv_please() that does
8394 the conversion if possible, and silently. */
8395 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8396 /* Need to try really hard to see if it's an integer.
8397 9.22337203685478e+18 is an integer.
8398 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8399 so $a="9.22337203685478e+18"; $a+0; $a++
8400 needs to be the same as $a="9.22337203685478e+18"; $a++
8407 /* sv_2iv *should* have made this an NV */
8408 if (flags & SVp_NOK) {
8409 (void)SvNOK_only(sv);
8410 SvNV_set(sv, SvNVX(sv) + 1.0);
8413 /* I don't think we can get here. Maybe I should assert this
8414 And if we do get here I suspect that sv_setnv will croak. NWC
8416 #if defined(USE_LONG_DOUBLE)
8417 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",
8418 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8420 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8421 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8424 #endif /* PERL_PRESERVE_IVUV */
8425 if (!numtype && ckWARN(WARN_NUMERIC))
8426 not_incrementable(sv);
8427 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
8431 while (d >= SvPVX_const(sv)) {
8439 /* MKS: The original code here died if letters weren't consecutive.
8440 * at least it didn't have to worry about non-C locales. The
8441 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
8442 * arranged in order (although not consecutively) and that only
8443 * [A-Za-z] are accepted by isALPHA in the C locale.
8445 if (*d != 'z' && *d != 'Z') {
8446 do { ++*d; } while (!isALPHA(*d));
8449 *(d--) -= 'z' - 'a';
8454 *(d--) -= 'z' - 'a' + 1;
8458 /* oh,oh, the number grew */
8459 SvGROW(sv, SvCUR(sv) + 2);
8460 SvCUR_set(sv, SvCUR(sv) + 1);
8461 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
8472 Auto-decrement of the value in the SV, doing string to numeric conversion
8473 if necessary. Handles 'get' magic and operator overloading.
8479 Perl_sv_dec(pTHX_ SV *const sv)
8489 =for apidoc sv_dec_nomg
8491 Auto-decrement of the value in the SV, doing string to numeric conversion
8492 if necessary. Handles operator overloading. Skips handling 'get' magic.
8498 Perl_sv_dec_nomg(pTHX_ SV *const sv)
8505 if (SvTHINKFIRST(sv)) {
8506 if (SvREADONLY(sv)) {
8507 Perl_croak_no_modify();
8511 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8513 i = PTR2IV(SvRV(sv));
8517 else sv_force_normal_flags(sv, 0);
8519 /* Unlike sv_inc we don't have to worry about string-never-numbers
8520 and keeping them magic. But we mustn't warn on punting */
8521 flags = SvFLAGS(sv);
8522 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8523 /* It's publicly an integer, or privately an integer-not-float */
8524 #ifdef PERL_PRESERVE_IVUV
8528 if (SvUVX(sv) == 0) {
8529 (void)SvIOK_only(sv);
8533 (void)SvIOK_only_UV(sv);
8534 SvUV_set(sv, SvUVX(sv) - 1);
8537 if (SvIVX(sv) == IV_MIN) {
8538 sv_setnv(sv, (NV)IV_MIN);
8542 (void)SvIOK_only(sv);
8543 SvIV_set(sv, SvIVX(sv) - 1);
8548 if (flags & SVp_NOK) {
8551 const NV was = SvNVX(sv);
8552 if (NV_OVERFLOWS_INTEGERS_AT &&
8553 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8554 /* diag_listed_as: Lost precision when %s %f by 1 */
8555 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8556 "Lost precision when decrementing %" NVff " by 1",
8559 (void)SvNOK_only(sv);
8560 SvNV_set(sv, was - 1.0);
8564 if (!(flags & SVp_POK)) {
8565 if ((flags & SVTYPEMASK) < SVt_PVIV)
8566 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8568 (void)SvIOK_only(sv);
8571 #ifdef PERL_PRESERVE_IVUV
8573 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8574 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8575 /* Need to try really hard to see if it's an integer.
8576 9.22337203685478e+18 is an integer.
8577 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8578 so $a="9.22337203685478e+18"; $a+0; $a--
8579 needs to be the same as $a="9.22337203685478e+18"; $a--
8586 /* sv_2iv *should* have made this an NV */
8587 if (flags & SVp_NOK) {
8588 (void)SvNOK_only(sv);
8589 SvNV_set(sv, SvNVX(sv) - 1.0);
8592 /* I don't think we can get here. Maybe I should assert this
8593 And if we do get here I suspect that sv_setnv will croak. NWC
8595 #if defined(USE_LONG_DOUBLE)
8596 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",
8597 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8599 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8600 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8604 #endif /* PERL_PRESERVE_IVUV */
8605 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8608 /* this define is used to eliminate a chunk of duplicated but shared logic
8609 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8610 * used anywhere but here - yves
8612 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8615 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8619 =for apidoc sv_mortalcopy
8621 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8622 The new SV is marked as mortal. It will be destroyed "soon", either by an
8623 explicit call to FREETMPS, or by an implicit call at places such as
8624 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8629 /* Make a string that will exist for the duration of the expression
8630 * evaluation. Actually, it may have to last longer than that, but
8631 * hopefully we won't free it until it has been assigned to a
8632 * permanent location. */
8635 Perl_sv_mortalcopy_flags(pTHX_ SV *const oldstr, U32 flags)
8640 if (flags & SV_GMAGIC)
8641 SvGETMAGIC(oldstr); /* before new_SV, in case it dies */
8643 sv_setsv_flags(sv,oldstr,flags & ~SV_GMAGIC);
8644 PUSH_EXTEND_MORTAL__SV_C(sv);
8650 =for apidoc sv_newmortal
8652 Creates a new null SV which is mortal. The reference count of the SV is
8653 set to 1. It will be destroyed "soon", either by an explicit call to
8654 FREETMPS, or by an implicit call at places such as statement boundaries.
8655 See also C<sv_mortalcopy> and C<sv_2mortal>.
8661 Perl_sv_newmortal(pTHX)
8667 SvFLAGS(sv) = SVs_TEMP;
8668 PUSH_EXTEND_MORTAL__SV_C(sv);
8674 =for apidoc newSVpvn_flags
8676 Creates a new SV and copies a string into it. The reference count for the
8677 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8678 string. You are responsible for ensuring that the source string is at least
8679 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8680 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8681 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8682 returning. If C<SVf_UTF8> is set, C<s>
8683 is considered to be in UTF-8 and the
8684 C<SVf_UTF8> flag will be set on the new SV.
8685 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8687 #define newSVpvn_utf8(s, len, u) \
8688 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8694 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8699 /* All the flags we don't support must be zero.
8700 And we're new code so I'm going to assert this from the start. */
8701 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8703 sv_setpvn(sv,s,len);
8705 /* This code used to do a sv_2mortal(), however we now unroll the call to
8706 * sv_2mortal() and do what it does ourselves here. Since we have asserted
8707 * that flags can only have the SVf_UTF8 and/or SVs_TEMP flags set above we
8708 * can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8709 * in turn means we dont need to mask out the SVf_UTF8 flag below, which
8710 * means that we eliminate quite a few steps than it looks - Yves
8711 * (explaining patch by gfx) */
8713 SvFLAGS(sv) |= flags;
8715 if(flags & SVs_TEMP){
8716 PUSH_EXTEND_MORTAL__SV_C(sv);
8723 =for apidoc sv_2mortal
8725 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8726 by an explicit call to FREETMPS, or by an implicit call at places such as
8727 statement boundaries. SvTEMP() is turned on which means that the SV's
8728 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8729 and C<sv_mortalcopy>.
8735 Perl_sv_2mortal(pTHX_ SV *const sv)
8742 PUSH_EXTEND_MORTAL__SV_C(sv);
8750 Creates a new SV and copies a string into it. The reference count for the
8751 SV is set to 1. If C<len> is zero, Perl will compute the length using
8752 strlen(). For efficiency, consider using C<newSVpvn> instead.
8758 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8764 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8769 =for apidoc newSVpvn
8771 Creates a new SV and copies a buffer into it, which may contain NUL characters
8772 (C<\0>) and other binary data. The reference count for the SV is set to 1.
8773 Note that if C<len> is zero, Perl will create a zero length (Perl) string. You
8774 are responsible for ensuring that the source buffer is at least
8775 C<len> bytes long. If the C<buffer> argument is NULL the new SV will be
8782 Perl_newSVpvn(pTHX_ const char *const buffer, const STRLEN len)
8788 sv_setpvn(sv,buffer,len);
8793 =for apidoc newSVhek
8795 Creates a new SV from the hash key structure. It will generate scalars that
8796 point to the shared string table where possible. Returns a new (undefined)
8797 SV if the hek is NULL.
8803 Perl_newSVhek(pTHX_ const HEK *const hek)
8813 if (HEK_LEN(hek) == HEf_SVKEY) {
8814 return newSVsv(*(SV**)HEK_KEY(hek));
8816 const int flags = HEK_FLAGS(hek);
8817 if (flags & HVhek_WASUTF8) {
8819 Andreas would like keys he put in as utf8 to come back as utf8
8821 STRLEN utf8_len = HEK_LEN(hek);
8822 SV * const sv = newSV_type(SVt_PV);
8823 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8824 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8825 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8828 } else if (flags & HVhek_UNSHARED) {
8829 /* A hash that isn't using shared hash keys has to have
8830 the flag in every key so that we know not to try to call
8831 share_hek_hek on it. */
8833 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8838 /* This will be overwhelminly the most common case. */
8840 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8841 more efficient than sharepvn(). */
8845 sv_upgrade(sv, SVt_PV);
8846 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8847 SvCUR_set(sv, HEK_LEN(hek));
8859 =for apidoc newSVpvn_share
8861 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8862 table. If the string does not already exist in the table, it is
8863 created first. Turns on the SvIsCOW flag (or READONLY
8864 and FAKE in 5.16 and earlier). If the C<hash> parameter
8865 is non-zero, that value is used; otherwise the hash is computed.
8866 The string's hash can later be retrieved from the SV
8867 with the C<SvSHARED_HASH()> macro. The idea here is
8868 that as the string table is used for shared hash keys these strings will have
8869 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8875 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8879 bool is_utf8 = FALSE;
8880 const char *const orig_src = src;
8883 STRLEN tmplen = -len;
8885 /* See the note in hv.c:hv_fetch() --jhi */
8886 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8890 PERL_HASH(hash, src, len);
8892 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8893 changes here, update it there too. */
8894 sv_upgrade(sv, SVt_PV);
8895 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8902 if (src != orig_src)
8908 =for apidoc newSVpv_share
8910 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8917 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8919 return newSVpvn_share(src, strlen(src), hash);
8922 #if defined(PERL_IMPLICIT_CONTEXT)
8924 /* pTHX_ magic can't cope with varargs, so this is a no-context
8925 * version of the main function, (which may itself be aliased to us).
8926 * Don't access this version directly.
8930 Perl_newSVpvf_nocontext(const char *const pat, ...)
8936 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8938 va_start(args, pat);
8939 sv = vnewSVpvf(pat, &args);
8946 =for apidoc newSVpvf
8948 Creates a new SV and initializes it with the string formatted like
8955 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8960 PERL_ARGS_ASSERT_NEWSVPVF;
8962 va_start(args, pat);
8963 sv = vnewSVpvf(pat, &args);
8968 /* backend for newSVpvf() and newSVpvf_nocontext() */
8971 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8976 PERL_ARGS_ASSERT_VNEWSVPVF;
8979 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8986 Creates a new SV and copies a floating point value into it.
8987 The reference count for the SV is set to 1.
8993 Perl_newSVnv(pTHX_ const NV n)
9006 Creates a new SV and copies an integer into it. The reference count for the
9013 Perl_newSViv(pTHX_ const IV i)
9026 Creates a new SV and copies an unsigned integer into it.
9027 The reference count for the SV is set to 1.
9033 Perl_newSVuv(pTHX_ const UV u)
9044 =for apidoc newSV_type
9046 Creates a new SV, of the type specified. The reference count for the new SV
9053 Perl_newSV_type(pTHX_ const svtype type)
9058 sv_upgrade(sv, type);
9063 =for apidoc newRV_noinc
9065 Creates an RV wrapper for an SV. The reference count for the original
9066 SV is B<not> incremented.
9072 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
9075 SV *sv = newSV_type(SVt_IV);
9077 PERL_ARGS_ASSERT_NEWRV_NOINC;
9080 SvRV_set(sv, tmpRef);
9085 /* newRV_inc is the official function name to use now.
9086 * newRV_inc is in fact #defined to newRV in sv.h
9090 Perl_newRV(pTHX_ SV *const sv)
9094 PERL_ARGS_ASSERT_NEWRV;
9096 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
9102 Creates a new SV which is an exact duplicate of the original SV.
9109 Perl_newSVsv(pTHX_ SV *const old)
9116 if (SvTYPE(old) == (svtype)SVTYPEMASK) {
9117 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
9120 /* Do this here, otherwise we leak the new SV if this croaks. */
9123 /* SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
9124 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
9125 sv_setsv_flags(sv, old, SV_NOSTEAL);
9130 =for apidoc sv_reset
9132 Underlying implementation for the C<reset> Perl function.
9133 Note that the perl-level function is vaguely deprecated.
9139 Perl_sv_reset(pTHX_ const char *s, HV *const stash)
9141 PERL_ARGS_ASSERT_SV_RESET;
9143 sv_resetpvn(*s ? s : NULL, strlen(s), stash);
9147 Perl_sv_resetpvn(pTHX_ const char *s, STRLEN len, HV * const stash)
9150 char todo[PERL_UCHAR_MAX+1];
9153 if (!stash || SvTYPE(stash) != SVt_PVHV)
9156 if (!s) { /* reset ?? searches */
9157 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
9159 const U32 count = mg->mg_len / sizeof(PMOP**);
9160 PMOP **pmp = (PMOP**) mg->mg_ptr;
9161 PMOP *const *const end = pmp + count;
9165 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
9167 (*pmp)->op_pmflags &= ~PMf_USED;
9175 /* reset variables */
9177 if (!HvARRAY(stash))
9180 Zero(todo, 256, char);
9184 I32 i = (unsigned char)*s;
9188 max = (unsigned char)*s++;
9189 for ( ; i <= max; i++) {
9192 for (i = 0; i <= (I32) HvMAX(stash); i++) {
9194 for (entry = HvARRAY(stash)[i];
9196 entry = HeNEXT(entry))
9201 if (!todo[(U8)*HeKEY(entry)])
9203 gv = MUTABLE_GV(HeVAL(entry));
9205 if (sv && !SvREADONLY(sv)) {
9206 SV_CHECK_THINKFIRST_COW_DROP(sv);
9207 if (!isGV(sv)) SvOK_off(sv);
9212 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
9223 Using various gambits, try to get an IO from an SV: the IO slot if its a
9224 GV; or the recursive result if we're an RV; or the IO slot of the symbol
9225 named after the PV if we're a string.
9227 'Get' magic is ignored on the sv passed in, but will be called on
9228 C<SvRV(sv)> if sv is an RV.
9234 Perl_sv_2io(pTHX_ SV *const sv)
9239 PERL_ARGS_ASSERT_SV_2IO;
9241 switch (SvTYPE(sv)) {
9243 io = MUTABLE_IO(sv);
9247 if (isGV_with_GP(sv)) {
9248 gv = MUTABLE_GV(sv);
9251 Perl_croak(aTHX_ "Bad filehandle: %"HEKf,
9252 HEKfARG(GvNAME_HEK(gv)));
9258 Perl_croak(aTHX_ PL_no_usym, "filehandle");
9260 SvGETMAGIC(SvRV(sv));
9261 return sv_2io(SvRV(sv));
9263 gv = gv_fetchsv_nomg(sv, 0, SVt_PVIO);
9270 if (SvGMAGICAL(sv)) {
9271 newsv = sv_newmortal();
9272 sv_setsv_nomg(newsv, sv);
9274 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(newsv));
9284 Using various gambits, try to get a CV from an SV; in addition, try if
9285 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
9286 The flags in C<lref> are passed to gv_fetchsv.
9292 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
9298 PERL_ARGS_ASSERT_SV_2CV;
9305 switch (SvTYPE(sv)) {
9309 return MUTABLE_CV(sv);
9319 sv = amagic_deref_call(sv, to_cv_amg);
9322 if (SvTYPE(sv) == SVt_PVCV) {
9323 cv = MUTABLE_CV(sv);
9328 else if(SvGETMAGIC(sv), isGV_with_GP(sv))
9329 gv = MUTABLE_GV(sv);
9331 Perl_croak(aTHX_ "Not a subroutine reference");
9333 else if (isGV_with_GP(sv)) {
9334 gv = MUTABLE_GV(sv);
9337 gv = gv_fetchsv_nomg(sv, lref, SVt_PVCV);
9344 /* Some flags to gv_fetchsv mean don't really create the GV */
9345 if (!isGV_with_GP(gv)) {
9350 if (lref & ~GV_ADDMG && !GvCVu(gv)) {
9351 /* XXX this is probably not what they think they're getting.
9352 * It has the same effect as "sub name;", i.e. just a forward
9363 Returns true if the SV has a true value by Perl's rules.
9364 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
9365 instead use an in-line version.
9371 Perl_sv_true(pTHX_ SV *const sv)
9376 const XPV* const tXpv = (XPV*)SvANY(sv);
9378 (tXpv->xpv_cur > 1 ||
9379 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
9386 return SvIVX(sv) != 0;
9389 return SvNVX(sv) != 0.0;
9391 return sv_2bool(sv);
9397 =for apidoc sv_pvn_force
9399 Get a sensible string out of the SV somehow.
9400 A private implementation of the C<SvPV_force> macro for compilers which
9401 can't cope with complex macro expressions. Always use the macro instead.
9403 =for apidoc sv_pvn_force_flags
9405 Get a sensible string out of the SV somehow.
9406 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
9407 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
9408 implemented in terms of this function.
9409 You normally want to use the various wrapper macros instead: see
9410 C<SvPV_force> and C<SvPV_force_nomg>
9416 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
9420 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
9422 if (flags & SV_GMAGIC) SvGETMAGIC(sv);
9423 if (SvTHINKFIRST(sv) && (!SvROK(sv) || SvREADONLY(sv)))
9424 sv_force_normal_flags(sv, 0);
9434 if (SvTYPE(sv) > SVt_PVLV
9435 || isGV_with_GP(sv))
9436 /* diag_listed_as: Can't coerce %s to %s in %s */
9437 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
9439 s = sv_2pv_flags(sv, &len, flags &~ SV_GMAGIC);
9446 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
9449 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
9450 SvGROW(sv, len + 1);
9451 Move(s,SvPVX(sv),len,char);
9453 SvPVX(sv)[len] = '\0';
9456 SvPOK_on(sv); /* validate pointer */
9458 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9459 PTR2UV(sv),SvPVX_const(sv)));
9462 (void)SvPOK_only_UTF8(sv);
9463 return SvPVX_mutable(sv);
9467 =for apidoc sv_pvbyten_force
9469 The backend for the C<SvPVbytex_force> macro. Always use the macro
9476 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9478 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9480 sv_pvn_force(sv,lp);
9481 sv_utf8_downgrade(sv,0);
9487 =for apidoc sv_pvutf8n_force
9489 The backend for the C<SvPVutf8x_force> macro. Always use the macro
9496 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9498 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9501 sv_utf8_upgrade_nomg(sv);
9507 =for apidoc sv_reftype
9509 Returns a string describing what the SV is a reference to.
9515 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9517 PERL_ARGS_ASSERT_SV_REFTYPE;
9518 if (ob && SvOBJECT(sv)) {
9519 return SvPV_nolen_const(sv_ref(NULL, sv, ob));
9522 switch (SvTYPE(sv)) {
9537 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9538 /* tied lvalues should appear to be
9539 * scalars for backwards compatibility */
9540 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9541 ? "SCALAR" : "LVALUE");
9542 case SVt_PVAV: return "ARRAY";
9543 case SVt_PVHV: return "HASH";
9544 case SVt_PVCV: return "CODE";
9545 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9546 ? "GLOB" : "SCALAR");
9547 case SVt_PVFM: return "FORMAT";
9548 case SVt_PVIO: return "IO";
9549 case SVt_INVLIST: return "INVLIST";
9550 case SVt_REGEXP: return "REGEXP";
9551 default: return "UNKNOWN";
9559 Returns a SV describing what the SV passed in is a reference to.
9565 Perl_sv_ref(pTHX_ SV *dst, const SV *const sv, const int ob)
9567 PERL_ARGS_ASSERT_SV_REF;
9570 dst = sv_newmortal();
9572 if (ob && SvOBJECT(sv)) {
9573 HvNAME_get(SvSTASH(sv))
9574 ? sv_sethek(dst, HvNAME_HEK(SvSTASH(sv)))
9575 : sv_setpvn(dst, "__ANON__", 8);
9578 const char * reftype = sv_reftype(sv, 0);
9579 sv_setpv(dst, reftype);
9585 =for apidoc sv_isobject
9587 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9588 object. If the SV is not an RV, or if the object is not blessed, then this
9595 Perl_sv_isobject(pTHX_ SV *sv)
9611 Returns a boolean indicating whether the SV is blessed into the specified
9612 class. This does not check for subtypes; use C<sv_derived_from> to verify
9613 an inheritance relationship.
9619 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9623 PERL_ARGS_ASSERT_SV_ISA;
9633 hvname = HvNAME_get(SvSTASH(sv));
9637 return strEQ(hvname, name);
9643 Creates a new SV for the existing RV, C<rv>, to point to. If C<rv> is not an
9644 RV then it will be upgraded to one. If C<classname> is non-null then the new
9645 SV will be blessed in the specified package. The new SV is returned and its
9646 reference count is 1. The reference count 1 is owned by C<rv>.
9652 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9657 PERL_ARGS_ASSERT_NEWSVRV;
9661 SV_CHECK_THINKFIRST_COW_DROP(rv);
9663 if (SvTYPE(rv) >= SVt_PVMG) {
9664 const U32 refcnt = SvREFCNT(rv);
9668 SvREFCNT(rv) = refcnt;
9670 sv_upgrade(rv, SVt_IV);
9671 } else if (SvROK(rv)) {
9672 SvREFCNT_dec(SvRV(rv));
9674 prepare_SV_for_RV(rv);
9682 HV* const stash = gv_stashpv(classname, GV_ADD);
9683 (void)sv_bless(rv, stash);
9689 Perl_newSVavdefelem(pTHX_ AV *av, SSize_t ix, bool extendible)
9691 SV * const lv = newSV_type(SVt_PVLV);
9692 PERL_ARGS_ASSERT_NEWSVAVDEFELEM;
9694 sv_magic(lv, NULL, PERL_MAGIC_defelem, NULL, 0);
9695 LvTARG(lv) = SvREFCNT_inc_simple_NN(av);
9696 LvSTARGOFF(lv) = ix;
9697 LvTARGLEN(lv) = extendible ? 1 : (STRLEN)UV_MAX;
9702 =for apidoc sv_setref_pv
9704 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9705 argument will be upgraded to an RV. That RV will be modified to point to
9706 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9707 into the SV. The C<classname> argument indicates the package for the
9708 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9709 will have a reference count of 1, and the RV will be returned.
9711 Do not use with other Perl types such as HV, AV, SV, CV, because those
9712 objects will become corrupted by the pointer copy process.
9714 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9720 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9724 PERL_ARGS_ASSERT_SV_SETREF_PV;
9727 sv_setsv(rv, &PL_sv_undef);
9731 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9736 =for apidoc sv_setref_iv
9738 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9739 argument will be upgraded to an RV. That RV will be modified to point to
9740 the new SV. The C<classname> argument indicates the package for the
9741 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9742 will have a reference count of 1, and the RV will be returned.
9748 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9750 PERL_ARGS_ASSERT_SV_SETREF_IV;
9752 sv_setiv(newSVrv(rv,classname), iv);
9757 =for apidoc sv_setref_uv
9759 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9760 argument will be upgraded to an RV. That RV will be modified to point to
9761 the new SV. The C<classname> argument indicates the package for the
9762 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9763 will have a reference count of 1, and the RV will be returned.
9769 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9771 PERL_ARGS_ASSERT_SV_SETREF_UV;
9773 sv_setuv(newSVrv(rv,classname), uv);
9778 =for apidoc sv_setref_nv
9780 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9781 argument will be upgraded to an RV. That RV will be modified to point to
9782 the new SV. The C<classname> argument indicates the package for the
9783 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9784 will have a reference count of 1, and the RV will be returned.
9790 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9792 PERL_ARGS_ASSERT_SV_SETREF_NV;
9794 sv_setnv(newSVrv(rv,classname), nv);
9799 =for apidoc sv_setref_pvn
9801 Copies a string into a new SV, optionally blessing the SV. The length of the
9802 string must be specified with C<n>. The C<rv> argument will be upgraded to
9803 an RV. That RV will be modified to point to the new SV. The C<classname>
9804 argument indicates the package for the blessing. Set C<classname> to
9805 C<NULL> to avoid the blessing. The new SV will have a reference count
9806 of 1, and the RV will be returned.
9808 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9814 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9815 const char *const pv, const STRLEN n)
9817 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9819 sv_setpvn(newSVrv(rv,classname), pv, n);
9824 =for apidoc sv_bless
9826 Blesses an SV into a specified package. The SV must be an RV. The package
9827 must be designated by its stash (see C<gv_stashpv()>). The reference count
9828 of the SV is unaffected.
9834 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9838 HV *oldstash = NULL;
9840 PERL_ARGS_ASSERT_SV_BLESS;
9844 Perl_croak(aTHX_ "Can't bless non-reference value");
9846 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9847 if (SvREADONLY(tmpRef))
9848 Perl_croak_no_modify();
9849 if (SvOBJECT(tmpRef)) {
9850 oldstash = SvSTASH(tmpRef);
9853 SvOBJECT_on(tmpRef);
9854 SvUPGRADE(tmpRef, SVt_PVMG);
9855 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9856 SvREFCNT_dec(oldstash);
9858 if(SvSMAGICAL(tmpRef))
9859 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9867 /* Downgrades a PVGV to a PVMG. If it's actually a PVLV, we leave the type
9868 * as it is after unglobbing it.
9871 PERL_STATIC_INLINE void
9872 S_sv_unglob(pTHX_ SV *const sv, U32 flags)
9877 SV * const temp = flags & SV_COW_DROP_PV ? NULL : sv_newmortal();
9879 PERL_ARGS_ASSERT_SV_UNGLOB;
9881 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9883 if (!(flags & SV_COW_DROP_PV))
9884 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9887 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9888 && HvNAME_get(stash))
9889 mro_method_changed_in(stash);
9890 gp_free(MUTABLE_GV(sv));
9893 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9897 if (GvNAME_HEK(sv)) {
9898 unshare_hek(GvNAME_HEK(sv));
9900 isGV_with_GP_off(sv);
9902 if(SvTYPE(sv) == SVt_PVGV) {
9903 /* need to keep SvANY(sv) in the right arena */
9904 xpvmg = new_XPVMG();
9905 StructCopy(SvANY(sv), xpvmg, XPVMG);
9906 del_XPVGV(SvANY(sv));
9909 SvFLAGS(sv) &= ~SVTYPEMASK;
9910 SvFLAGS(sv) |= SVt_PVMG;
9913 /* Intentionally not calling any local SET magic, as this isn't so much a
9914 set operation as merely an internal storage change. */
9915 if (flags & SV_COW_DROP_PV) SvOK_off(sv);
9916 else sv_setsv_flags(sv, temp, 0);
9918 if ((const GV *)sv == PL_last_in_gv)
9919 PL_last_in_gv = NULL;
9920 else if ((const GV *)sv == PL_statgv)
9925 =for apidoc sv_unref_flags
9927 Unsets the RV status of the SV, and decrements the reference count of
9928 whatever was being referenced by the RV. This can almost be thought of
9929 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9930 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9931 (otherwise the decrementing is conditional on the reference count being
9932 different from one or the reference being a readonly SV).
9939 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9941 SV* const target = SvRV(ref);
9943 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9945 if (SvWEAKREF(ref)) {
9946 sv_del_backref(target, ref);
9948 SvRV_set(ref, NULL);
9951 SvRV_set(ref, NULL);
9953 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9954 assigned to as BEGIN {$a = \"Foo"} will fail. */
9955 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9956 SvREFCNT_dec_NN(target);
9957 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9958 sv_2mortal(target); /* Schedule for freeing later */
9962 =for apidoc sv_untaint
9964 Untaint an SV. Use C<SvTAINTED_off> instead.
9970 Perl_sv_untaint(pTHX_ SV *const sv)
9972 PERL_ARGS_ASSERT_SV_UNTAINT;
9974 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9975 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9982 =for apidoc sv_tainted
9984 Test an SV for taintedness. Use C<SvTAINTED> instead.
9990 Perl_sv_tainted(pTHX_ SV *const sv)
9992 PERL_ARGS_ASSERT_SV_TAINTED;
9994 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9995 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9996 if (mg && (mg->mg_len & 1) )
10003 =for apidoc sv_setpviv
10005 Copies an integer into the given SV, also updating its string value.
10006 Does not handle 'set' magic. See C<sv_setpviv_mg>.
10012 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
10014 char buf[TYPE_CHARS(UV)];
10016 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
10018 PERL_ARGS_ASSERT_SV_SETPVIV;
10020 sv_setpvn(sv, ptr, ebuf - ptr);
10024 =for apidoc sv_setpviv_mg
10026 Like C<sv_setpviv>, but also handles 'set' magic.
10032 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
10034 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
10036 sv_setpviv(sv, iv);
10040 #if defined(PERL_IMPLICIT_CONTEXT)
10042 /* pTHX_ magic can't cope with varargs, so this is a no-context
10043 * version of the main function, (which may itself be aliased to us).
10044 * Don't access this version directly.
10048 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
10053 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
10055 va_start(args, pat);
10056 sv_vsetpvf(sv, pat, &args);
10060 /* pTHX_ magic can't cope with varargs, so this is a no-context
10061 * version of the main function, (which may itself be aliased to us).
10062 * Don't access this version directly.
10066 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
10071 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
10073 va_start(args, pat);
10074 sv_vsetpvf_mg(sv, pat, &args);
10080 =for apidoc sv_setpvf
10082 Works like C<sv_catpvf> but copies the text into the SV instead of
10083 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
10089 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
10093 PERL_ARGS_ASSERT_SV_SETPVF;
10095 va_start(args, pat);
10096 sv_vsetpvf(sv, pat, &args);
10101 =for apidoc sv_vsetpvf
10103 Works like C<sv_vcatpvf> but copies the text into the SV instead of
10104 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
10106 Usually used via its frontend C<sv_setpvf>.
10112 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
10114 PERL_ARGS_ASSERT_SV_VSETPVF;
10116 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
10120 =for apidoc sv_setpvf_mg
10122 Like C<sv_setpvf>, but also handles 'set' magic.
10128 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
10132 PERL_ARGS_ASSERT_SV_SETPVF_MG;
10134 va_start(args, pat);
10135 sv_vsetpvf_mg(sv, pat, &args);
10140 =for apidoc sv_vsetpvf_mg
10142 Like C<sv_vsetpvf>, but also handles 'set' magic.
10144 Usually used via its frontend C<sv_setpvf_mg>.
10150 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
10152 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
10154 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
10158 #if defined(PERL_IMPLICIT_CONTEXT)
10160 /* pTHX_ magic can't cope with varargs, so this is a no-context
10161 * version of the main function, (which may itself be aliased to us).
10162 * Don't access this version directly.
10166 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
10171 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
10173 va_start(args, pat);
10174 sv_vcatpvf(sv, pat, &args);
10178 /* pTHX_ magic can't cope with varargs, so this is a no-context
10179 * version of the main function, (which may itself be aliased to us).
10180 * Don't access this version directly.
10184 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
10189 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
10191 va_start(args, pat);
10192 sv_vcatpvf_mg(sv, pat, &args);
10198 =for apidoc sv_catpvf
10200 Processes its arguments like C<sprintf> and appends the formatted
10201 output to an SV. If the appended data contains "wide" characters
10202 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
10203 and characters >255 formatted with %c), the original SV might get
10204 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
10205 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
10206 valid UTF-8; if the original SV was bytes, the pattern should be too.
10211 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
10215 PERL_ARGS_ASSERT_SV_CATPVF;
10217 va_start(args, pat);
10218 sv_vcatpvf(sv, pat, &args);
10223 =for apidoc sv_vcatpvf
10225 Processes its arguments like C<vsprintf> and appends the formatted output
10226 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
10228 Usually used via its frontend C<sv_catpvf>.
10234 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
10236 PERL_ARGS_ASSERT_SV_VCATPVF;
10238 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
10242 =for apidoc sv_catpvf_mg
10244 Like C<sv_catpvf>, but also handles 'set' magic.
10250 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
10254 PERL_ARGS_ASSERT_SV_CATPVF_MG;
10256 va_start(args, pat);
10257 sv_vcatpvf_mg(sv, pat, &args);
10262 =for apidoc sv_vcatpvf_mg
10264 Like C<sv_vcatpvf>, but also handles 'set' magic.
10266 Usually used via its frontend C<sv_catpvf_mg>.
10272 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
10274 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
10276 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
10281 =for apidoc sv_vsetpvfn
10283 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
10286 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
10292 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
10293 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
10295 PERL_ARGS_ASSERT_SV_VSETPVFN;
10298 sv_vcatpvfn_flags(sv, pat, patlen, args, svargs, svmax, maybe_tainted, 0);
10303 * Warn of missing argument to sprintf, and then return a defined value
10304 * to avoid inappropriate "use of uninit" warnings [perl #71000].
10306 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
10308 S_vcatpvfn_missing_argument(pTHX) {
10309 if (ckWARN(WARN_MISSING)) {
10310 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
10311 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
10318 S_expect_number(pTHX_ char **const pattern)
10323 PERL_ARGS_ASSERT_EXPECT_NUMBER;
10325 switch (**pattern) {
10326 case '1': case '2': case '3':
10327 case '4': case '5': case '6':
10328 case '7': case '8': case '9':
10329 var = *(*pattern)++ - '0';
10330 while (isDIGIT(**pattern)) {
10331 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
10333 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
10341 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
10343 const int neg = nv < 0;
10346 PERL_ARGS_ASSERT_F0CONVERT;
10354 if (uv & 1 && uv == nv)
10355 uv--; /* Round to even */
10357 const unsigned dig = uv % 10;
10359 } while (uv /= 10);
10370 =for apidoc sv_vcatpvfn
10372 =for apidoc sv_vcatpvfn_flags
10374 Processes its arguments like C<vsprintf> and appends the formatted output
10375 to an SV. Uses an array of SVs if the C style variable argument list is
10376 missing (NULL). When running with taint checks enabled, indicates via
10377 C<maybe_tainted> if results are untrustworthy (often due to the use of
10380 If called as C<sv_vcatpvfn> or flags include C<SV_GMAGIC>, calls get magic.
10382 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
10387 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
10388 vecstr = (U8*)SvPV_const(vecsv,veclen);\
10389 vec_utf8 = DO_UTF8(vecsv);
10391 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
10394 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
10395 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
10397 PERL_ARGS_ASSERT_SV_VCATPVFN;
10399 sv_vcatpvfn_flags(sv, pat, patlen, args, svargs, svmax, maybe_tainted, SV_GMAGIC|SV_SMAGIC);
10403 Perl_sv_vcatpvfn_flags(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
10404 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted,
10410 const char *patend;
10413 static const char nullstr[] = "(null)";
10415 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
10416 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
10418 /* Times 4: a decimal digit takes more than 3 binary digits.
10419 * NV_DIG: mantissa takes than many decimal digits.
10420 * Plus 32: Playing safe. */
10421 char ebuf[IV_DIG * 4 + NV_DIG + 32];
10422 /* large enough for "%#.#f" --chip */
10423 /* what about long double NVs? --jhi */
10424 #ifdef USE_LOCALE_NUMERIC
10425 SV* oldlocale = NULL;
10428 PERL_ARGS_ASSERT_SV_VCATPVFN_FLAGS;
10429 PERL_UNUSED_ARG(maybe_tainted);
10431 if (flags & SV_GMAGIC)
10434 /* no matter what, this is a string now */
10435 (void)SvPV_force_nomg(sv, origlen);
10437 /* special-case "", "%s", and "%-p" (SVf - see below) */
10440 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
10442 const char * const s = va_arg(*args, char*);
10443 sv_catpv_nomg(sv, s ? s : nullstr);
10445 else if (svix < svmax) {
10446 /* we want get magic on the source but not the target. sv_catsv can't do that, though */
10447 SvGETMAGIC(*svargs);
10448 sv_catsv_nomg(sv, *svargs);
10451 S_vcatpvfn_missing_argument(aTHX);
10454 if (args && patlen == 3 && pat[0] == '%' &&
10455 pat[1] == '-' && pat[2] == 'p') {
10456 argsv = MUTABLE_SV(va_arg(*args, void*));
10457 sv_catsv_nomg(sv, argsv);
10461 #ifndef USE_LONG_DOUBLE
10462 /* special-case "%.<number>[gf]" */
10463 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
10464 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
10465 unsigned digits = 0;
10469 while (*pp >= '0' && *pp <= '9')
10470 digits = 10 * digits + (*pp++ - '0');
10471 if (pp - pat == (int)patlen - 1 && svix < svmax) {
10472 const NV nv = SvNV(*svargs);
10474 /* Add check for digits != 0 because it seems that some
10475 gconverts are buggy in this case, and we don't yet have
10476 a Configure test for this. */
10477 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
10478 /* 0, point, slack */
10479 V_Gconvert(nv, (int)digits, 0, ebuf);
10480 sv_catpv_nomg(sv, ebuf);
10481 if (*ebuf) /* May return an empty string for digits==0 */
10484 } else if (!digits) {
10487 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
10488 sv_catpvn_nomg(sv, p, l);
10494 #endif /* !USE_LONG_DOUBLE */
10496 if (!args && svix < svmax && DO_UTF8(*svargs))
10499 patend = (char*)pat + patlen;
10500 for (p = (char*)pat; p < patend; p = q) {
10503 bool vectorize = FALSE;
10504 bool vectorarg = FALSE;
10505 bool vec_utf8 = FALSE;
10511 bool has_precis = FALSE;
10513 const I32 osvix = svix;
10514 bool is_utf8 = FALSE; /* is this item utf8? */
10515 #ifdef HAS_LDBL_SPRINTF_BUG
10516 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10517 with sfio - Allen <allens@cpan.org> */
10518 bool fix_ldbl_sprintf_bug = FALSE;
10522 U8 utf8buf[UTF8_MAXBYTES+1];
10523 STRLEN esignlen = 0;
10525 const char *eptr = NULL;
10526 const char *fmtstart;
10529 const U8 *vecstr = NULL;
10536 /* we need a long double target in case HAS_LONG_DOUBLE but
10537 not USE_LONG_DOUBLE
10539 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10547 const char *dotstr = ".";
10548 STRLEN dotstrlen = 1;
10549 I32 efix = 0; /* explicit format parameter index */
10550 I32 ewix = 0; /* explicit width index */
10551 I32 epix = 0; /* explicit precision index */
10552 I32 evix = 0; /* explicit vector index */
10553 bool asterisk = FALSE;
10555 /* echo everything up to the next format specification */
10556 for (q = p; q < patend && *q != '%'; ++q) ;
10558 if (has_utf8 && !pat_utf8)
10559 sv_catpvn_nomg_utf8_upgrade(sv, p, q - p, nsv);
10561 sv_catpvn_nomg(sv, p, q - p);
10570 We allow format specification elements in this order:
10571 \d+\$ explicit format parameter index
10573 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10574 0 flag (as above): repeated to allow "v02"
10575 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10576 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10578 [%bcdefginopsuxDFOUX] format (mandatory)
10583 As of perl5.9.3, printf format checking is on by default.
10584 Internally, perl uses %p formats to provide an escape to
10585 some extended formatting. This block deals with those
10586 extensions: if it does not match, (char*)q is reset and
10587 the normal format processing code is used.
10589 Currently defined extensions are:
10590 %p include pointer address (standard)
10591 %-p (SVf) include an SV (previously %_)
10592 %-<num>p include an SV with precision <num>
10594 %3p include a HEK with precision of 256
10595 %4p char* preceded by utf8 flag and length
10596 %<num>p (where num is 1 or > 4) reserved for future
10599 Robin Barker 2005-07-14 (but modified since)
10601 %1p (VDf) removed. RMB 2007-10-19
10608 else if (strnEQ(q, UTF8f, sizeof(UTF8f)-1)) { /* UTF8f */
10609 /* The argument has already gone through cBOOL, so the cast
10611 is_utf8 = (bool)va_arg(*args, int);
10612 elen = va_arg(*args, UV);
10613 eptr = va_arg(*args, char *);
10614 q += sizeof(UTF8f)-1;
10617 n = expect_number(&q);
10619 if (sv) { /* SVf */
10624 argsv = MUTABLE_SV(va_arg(*args, void*));
10625 eptr = SvPV_const(argsv, elen);
10626 if (DO_UTF8(argsv))
10630 else if (n==2 || n==3) { /* HEKf */
10631 HEK * const hek = va_arg(*args, HEK *);
10632 eptr = HEK_KEY(hek);
10633 elen = HEK_LEN(hek);
10634 if (HEK_UTF8(hek)) is_utf8 = TRUE;
10635 if (n==3) precis = 256, has_precis = TRUE;
10639 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10640 "internal %%<num>p might conflict with future printf extensions");
10646 if ( (width = expect_number(&q)) ) {
10661 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10690 if ( (ewix = expect_number(&q)) )
10699 if ((vectorarg = asterisk)) {
10712 width = expect_number(&q);
10715 if (vectorize && vectorarg) {
10716 /* vectorizing, but not with the default "." */
10718 vecsv = va_arg(*args, SV*);
10720 vecsv = (evix > 0 && evix <= svmax)
10721 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10723 vecsv = svix < svmax
10724 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10726 dotstr = SvPV_const(vecsv, dotstrlen);
10727 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10728 bad with tied or overloaded values that return UTF8. */
10729 if (DO_UTF8(vecsv))
10731 else if (has_utf8) {
10732 vecsv = sv_mortalcopy(vecsv);
10733 sv_utf8_upgrade(vecsv);
10734 dotstr = SvPV_const(vecsv, dotstrlen);
10741 i = va_arg(*args, int);
10743 i = (ewix ? ewix <= svmax : svix < svmax) ?
10744 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10746 width = (i < 0) ? -i : i;
10756 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10758 /* XXX: todo, support specified precision parameter */
10762 i = va_arg(*args, int);
10764 i = (ewix ? ewix <= svmax : svix < svmax)
10765 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10767 has_precis = !(i < 0);
10771 while (isDIGIT(*q))
10772 precis = precis * 10 + (*q++ - '0');
10781 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10782 vecsv = svargs[efix ? efix-1 : svix++];
10783 vecstr = (U8*)SvPV_const(vecsv,veclen);
10784 vec_utf8 = DO_UTF8(vecsv);
10786 /* if this is a version object, we need to convert
10787 * back into v-string notation and then let the
10788 * vectorize happen normally
10790 if (sv_isobject(vecsv) && sv_derived_from(vecsv, "version")) {
10791 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10792 Perl_ck_warner_d(aTHX_ packWARN(WARN_PRINTF),
10793 "vector argument not supported with alpha versions");
10796 vecsv = sv_newmortal();
10797 scan_vstring((char *)vecstr, (char *)vecstr + veclen,
10799 vecstr = (U8*)SvPV_const(vecsv, veclen);
10800 vec_utf8 = DO_UTF8(vecsv);
10814 case 'I': /* Ix, I32x, and I64x */
10815 # ifdef USE_64_BIT_INT
10816 if (q[1] == '6' && q[2] == '4') {
10822 if (q[1] == '3' && q[2] == '2') {
10826 # ifdef USE_64_BIT_INT
10832 #if IVSIZE >= 8 || defined(HAS_LONG_DOUBLE)
10844 #if IVSIZE >= 8 || defined(HAS_LONG_DOUBLE)
10845 if (*q == 'l') { /* lld, llf */
10854 if (*++q == 'h') { /* hhd, hhu */
10883 if (!vectorize && !args) {
10885 const I32 i = efix-1;
10886 argsv = (i >= 0 && i < svmax)
10887 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10889 argsv = (svix >= 0 && svix < svmax)
10890 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10894 switch (c = *q++) {
10901 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10903 (!UVCHR_IS_INVARIANT(uv) && SvUTF8(sv)))
10905 eptr = (char*)utf8buf;
10906 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10920 eptr = va_arg(*args, char*);
10922 elen = strlen(eptr);
10924 eptr = (char *)nullstr;
10925 elen = sizeof nullstr - 1;
10929 eptr = SvPV_const(argsv, elen);
10930 if (DO_UTF8(argsv)) {
10931 STRLEN old_precis = precis;
10932 if (has_precis && precis < elen) {
10933 STRLEN ulen = sv_or_pv_len_utf8(argsv, eptr, elen);
10934 STRLEN p = precis > ulen ? ulen : precis;
10935 precis = sv_or_pv_pos_u2b(argsv, eptr, p, 0);
10936 /* sticks at end */
10938 if (width) { /* fudge width (can't fudge elen) */
10939 if (has_precis && precis < elen)
10940 width += precis - old_precis;
10943 elen - sv_or_pv_len_utf8(argsv,eptr,elen);
10950 if (has_precis && precis < elen)
10957 if (alt || vectorize)
10959 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10980 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10989 esignbuf[esignlen++] = plus;
10993 case 'c': iv = (char)va_arg(*args, int); break;
10994 case 'h': iv = (short)va_arg(*args, int); break;
10995 case 'l': iv = va_arg(*args, long); break;
10996 case 'V': iv = va_arg(*args, IV); break;
10997 case 'z': iv = va_arg(*args, SSize_t); break;
10998 case 't': iv = va_arg(*args, ptrdiff_t); break;
10999 default: iv = va_arg(*args, int); break;
11001 case 'j': iv = va_arg(*args, intmax_t); break;
11005 iv = va_arg(*args, Quad_t); break;
11012 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
11014 case 'c': iv = (char)tiv; break;
11015 case 'h': iv = (short)tiv; break;
11016 case 'l': iv = (long)tiv; break;
11018 default: iv = tiv; break;
11021 iv = (Quad_t)tiv; break;
11027 if ( !vectorize ) /* we already set uv above */
11032 esignbuf[esignlen++] = plus;
11036 esignbuf[esignlen++] = '-';
11080 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
11091 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
11092 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
11093 case 'l': uv = va_arg(*args, unsigned long); break;
11094 case 'V': uv = va_arg(*args, UV); break;
11095 case 'z': uv = va_arg(*args, Size_t); break;
11096 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
11098 case 'j': uv = va_arg(*args, uintmax_t); break;
11100 default: uv = va_arg(*args, unsigned); break;
11103 uv = va_arg(*args, Uquad_t); break;
11110 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
11112 case 'c': uv = (unsigned char)tuv; break;
11113 case 'h': uv = (unsigned short)tuv; break;
11114 case 'l': uv = (unsigned long)tuv; break;
11116 default: uv = tuv; break;
11119 uv = (Uquad_t)tuv; break;
11128 char *ptr = ebuf + sizeof ebuf;
11129 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
11135 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
11139 } while (uv >>= 4);
11141 esignbuf[esignlen++] = '0';
11142 esignbuf[esignlen++] = c; /* 'x' or 'X' */
11148 *--ptr = '0' + dig;
11149 } while (uv >>= 3);
11150 if (alt && *ptr != '0')
11156 *--ptr = '0' + dig;
11157 } while (uv >>= 1);
11159 esignbuf[esignlen++] = '0';
11160 esignbuf[esignlen++] = c;
11163 default: /* it had better be ten or less */
11166 *--ptr = '0' + dig;
11167 } while (uv /= base);
11170 elen = (ebuf + sizeof ebuf) - ptr;
11174 zeros = precis - elen;
11175 else if (precis == 0 && elen == 1 && *eptr == '0'
11176 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
11179 /* a precision nullifies the 0 flag. */
11186 /* FLOATING POINT */
11189 c = 'f'; /* maybe %F isn't supported here */
11191 case 'e': case 'E':
11193 case 'g': case 'G':
11197 /* This is evil, but floating point is even more evil */
11199 /* for SV-style calling, we can only get NV
11200 for C-style calling, we assume %f is double;
11201 for simplicity we allow any of %Lf, %llf, %qf for long double
11205 #if defined(USE_LONG_DOUBLE)
11209 /* [perl #20339] - we should accept and ignore %lf rather than die */
11213 #if defined(USE_LONG_DOUBLE)
11214 intsize = args ? 0 : 'q';
11218 #if defined(HAS_LONG_DOUBLE)
11231 /* now we need (long double) if intsize == 'q', else (double) */
11233 #if LONG_DOUBLESIZE > DOUBLESIZE
11235 va_arg(*args, long double) :
11236 va_arg(*args, double)
11238 va_arg(*args, double)
11243 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
11244 else. frexp() has some unspecified behaviour for those three */
11245 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
11247 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
11248 will cast our (long double) to (double) */
11249 (void)Perl_frexp(nv, &i);
11250 if (i == PERL_INT_MIN)
11251 Perl_die(aTHX_ "panic: frexp");
11253 need = BIT_DIGITS(i);
11255 need += has_precis ? precis : 6; /* known default */
11260 #ifdef HAS_LDBL_SPRINTF_BUG
11261 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
11262 with sfio - Allen <allens@cpan.org> */
11265 # define MY_DBL_MAX DBL_MAX
11266 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
11267 # if DOUBLESIZE >= 8
11268 # define MY_DBL_MAX 1.7976931348623157E+308L
11270 # define MY_DBL_MAX 3.40282347E+38L
11274 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
11275 # define MY_DBL_MAX_BUG 1L
11277 # define MY_DBL_MAX_BUG MY_DBL_MAX
11281 # define MY_DBL_MIN DBL_MIN
11282 # else /* XXX guessing! -Allen */
11283 # if DOUBLESIZE >= 8
11284 # define MY_DBL_MIN 2.2250738585072014E-308L
11286 # define MY_DBL_MIN 1.17549435E-38L
11290 if ((intsize == 'q') && (c == 'f') &&
11291 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
11292 (need < DBL_DIG)) {
11293 /* it's going to be short enough that
11294 * long double precision is not needed */
11296 if ((nv <= 0L) && (nv >= -0L))
11297 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
11299 /* would use Perl_fp_class as a double-check but not
11300 * functional on IRIX - see perl.h comments */
11302 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
11303 /* It's within the range that a double can represent */
11304 #if defined(DBL_MAX) && !defined(DBL_MIN)
11305 if ((nv >= ((long double)1/DBL_MAX)) ||
11306 (nv <= (-(long double)1/DBL_MAX)))
11308 fix_ldbl_sprintf_bug = TRUE;
11311 if (fix_ldbl_sprintf_bug == TRUE) {
11321 # undef MY_DBL_MAX_BUG
11324 #endif /* HAS_LDBL_SPRINTF_BUG */
11326 need += 20; /* fudge factor */
11327 if (PL_efloatsize < need) {
11328 Safefree(PL_efloatbuf);
11329 PL_efloatsize = need + 20; /* more fudge */
11330 Newx(PL_efloatbuf, PL_efloatsize, char);
11331 PL_efloatbuf[0] = '\0';
11334 if ( !(width || left || plus || alt) && fill != '0'
11335 && has_precis && intsize != 'q' ) { /* Shortcuts */
11336 /* See earlier comment about buggy Gconvert when digits,
11338 if ( c == 'g' && precis) {
11339 V_Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
11340 /* May return an empty string for digits==0 */
11341 if (*PL_efloatbuf) {
11342 elen = strlen(PL_efloatbuf);
11343 goto float_converted;
11345 } else if ( c == 'f' && !precis) {
11346 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
11351 char *ptr = ebuf + sizeof ebuf;
11354 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
11355 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
11356 if (intsize == 'q') {
11357 /* Copy the one or more characters in a long double
11358 * format before the 'base' ([efgEFG]) character to
11359 * the format string. */
11360 static char const prifldbl[] = PERL_PRIfldbl;
11361 char const *p = prifldbl + sizeof(prifldbl) - 3;
11362 while (p >= prifldbl) { *--ptr = *p--; }
11367 do { *--ptr = '0' + (base % 10); } while (base /= 10);
11372 do { *--ptr = '0' + (base % 10); } while (base /= 10);
11384 /* No taint. Otherwise we are in the strange situation
11385 * where printf() taints but print($float) doesn't.
11388 #ifdef USE_LOCALE_NUMERIC
11389 if (! PL_numeric_standard && ! IN_SOME_LOCALE_FORM) {
11391 /* We use a mortal SV, so that any failures (such as if
11392 * warnings are made fatal) won't leak */
11393 char *oldlocale_string = setlocale(LC_NUMERIC, NULL);
11394 oldlocale = newSVpvn_flags(oldlocale_string,
11395 strlen(oldlocale_string),
11397 PL_numeric_standard = TRUE;
11398 setlocale(LC_NUMERIC, "C");
11402 /* hopefully the above makes ptr a very constrained format
11403 * that is safe to use, even though it's not literal */
11404 GCC_DIAG_IGNORE(-Wformat-nonliteral);
11405 #if defined(HAS_LONG_DOUBLE)
11406 elen = ((intsize == 'q')
11407 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
11408 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
11410 elen = my_sprintf(PL_efloatbuf, ptr, nv);
11415 eptr = PL_efloatbuf;
11417 #ifdef USE_LOCALE_NUMERIC
11418 if (PL_numeric_radix_sv && SvUTF8(PL_numeric_radix_sv)
11419 && instr(eptr, SvPVX_const(PL_numeric_radix_sv)))
11432 i = SvCUR(sv) - origlen;
11435 case 'c': *(va_arg(*args, char*)) = i; break;
11436 case 'h': *(va_arg(*args, short*)) = i; break;
11437 default: *(va_arg(*args, int*)) = i; break;
11438 case 'l': *(va_arg(*args, long*)) = i; break;
11439 case 'V': *(va_arg(*args, IV*)) = i; break;
11440 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
11441 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
11443 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
11447 *(va_arg(*args, Quad_t*)) = i; break;
11454 sv_setuv_mg(argsv, has_utf8 ? (UV)sv_len_utf8(sv) : (UV)i);
11455 continue; /* not "break" */
11462 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
11463 && ckWARN(WARN_PRINTF))
11465 SV * const msg = sv_newmortal();
11466 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
11467 (PL_op->op_type == OP_PRTF) ? "" : "s");
11468 if (fmtstart < patend) {
11469 const char * const fmtend = q < patend ? q : patend;
11471 sv_catpvs(msg, "\"%");
11472 for (f = fmtstart; f < fmtend; f++) {
11474 sv_catpvn_nomg(msg, f, 1);
11476 Perl_sv_catpvf(aTHX_ msg,
11477 "\\%03"UVof, (UV)*f & 0xFF);
11480 sv_catpvs(msg, "\"");
11482 sv_catpvs(msg, "end of string");
11484 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
11487 /* output mangled stuff ... */
11493 /* ... right here, because formatting flags should not apply */
11494 SvGROW(sv, SvCUR(sv) + elen + 1);
11496 Copy(eptr, p, elen, char);
11499 SvCUR_set(sv, p - SvPVX_const(sv));
11501 continue; /* not "break" */
11504 if (is_utf8 != has_utf8) {
11507 sv_utf8_upgrade(sv);
11510 const STRLEN old_elen = elen;
11511 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
11512 sv_utf8_upgrade(nsv);
11513 eptr = SvPVX_const(nsv);
11516 if (width) { /* fudge width (can't fudge elen) */
11517 width += elen - old_elen;
11523 have = esignlen + zeros + elen;
11525 croak_memory_wrap();
11527 need = (have > width ? have : width);
11530 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
11531 croak_memory_wrap();
11532 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
11534 if (esignlen && fill == '0') {
11536 for (i = 0; i < (int)esignlen; i++)
11537 *p++ = esignbuf[i];
11539 if (gap && !left) {
11540 memset(p, fill, gap);
11543 if (esignlen && fill != '0') {
11545 for (i = 0; i < (int)esignlen; i++)
11546 *p++ = esignbuf[i];
11550 for (i = zeros; i; i--)
11554 Copy(eptr, p, elen, char);
11558 memset(p, ' ', gap);
11563 Copy(dotstr, p, dotstrlen, char);
11567 vectorize = FALSE; /* done iterating over vecstr */
11574 SvCUR_set(sv, p - SvPVX_const(sv));
11582 #ifdef USE_LOCALE_NUMERIC /* Done outside loop, so don't have to save/restore
11585 setlocale(LC_NUMERIC, SvPVX(oldlocale));
11586 PL_numeric_standard = FALSE;
11591 /* =========================================================================
11593 =head1 Cloning an interpreter
11595 All the macros and functions in this section are for the private use of
11596 the main function, perl_clone().
11598 The foo_dup() functions make an exact copy of an existing foo thingy.
11599 During the course of a cloning, a hash table is used to map old addresses
11600 to new addresses. The table is created and manipulated with the
11601 ptr_table_* functions.
11605 * =========================================================================*/
11608 #if defined(USE_ITHREADS)
11610 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11611 #ifndef GpREFCNT_inc
11612 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11616 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11617 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11618 If this changes, please unmerge ss_dup.
11619 Likewise, sv_dup_inc_multiple() relies on this fact. */
11620 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11621 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11622 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11623 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11624 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11625 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11626 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11627 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11628 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11629 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11630 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11631 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11632 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11634 /* clone a parser */
11637 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11641 PERL_ARGS_ASSERT_PARSER_DUP;
11646 /* look for it in the table first */
11647 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11651 /* create anew and remember what it is */
11652 Newxz(parser, 1, yy_parser);
11653 ptr_table_store(PL_ptr_table, proto, parser);
11655 /* XXX these not yet duped */
11656 parser->old_parser = NULL;
11657 parser->stack = NULL;
11659 parser->stack_size = 0;
11660 /* XXX parser->stack->state = 0; */
11662 /* XXX eventually, just Copy() most of the parser struct ? */
11664 parser->lex_brackets = proto->lex_brackets;
11665 parser->lex_casemods = proto->lex_casemods;
11666 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11667 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11668 parser->lex_casestack = savepvn(proto->lex_casestack,
11669 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11670 parser->lex_defer = proto->lex_defer;
11671 parser->lex_dojoin = proto->lex_dojoin;
11672 parser->lex_expect = proto->lex_expect;
11673 parser->lex_formbrack = proto->lex_formbrack;
11674 parser->lex_inpat = proto->lex_inpat;
11675 parser->lex_inwhat = proto->lex_inwhat;
11676 parser->lex_op = proto->lex_op;
11677 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11678 parser->lex_starts = proto->lex_starts;
11679 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11680 parser->multi_close = proto->multi_close;
11681 parser->multi_open = proto->multi_open;
11682 parser->multi_start = proto->multi_start;
11683 parser->multi_end = proto->multi_end;
11684 parser->preambled = proto->preambled;
11685 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11686 parser->linestr = sv_dup_inc(proto->linestr, param);
11687 parser->expect = proto->expect;
11688 parser->copline = proto->copline;
11689 parser->last_lop_op = proto->last_lop_op;
11690 parser->lex_state = proto->lex_state;
11691 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11692 /* rsfp_filters entries have fake IoDIRP() */
11693 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11694 parser->in_my = proto->in_my;
11695 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11696 parser->error_count = proto->error_count;
11699 parser->linestr = sv_dup_inc(proto->linestr, param);
11702 char * const ols = SvPVX(proto->linestr);
11703 char * const ls = SvPVX(parser->linestr);
11705 parser->bufptr = ls + (proto->bufptr >= ols ?
11706 proto->bufptr - ols : 0);
11707 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11708 proto->oldbufptr - ols : 0);
11709 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11710 proto->oldoldbufptr - ols : 0);
11711 parser->linestart = ls + (proto->linestart >= ols ?
11712 proto->linestart - ols : 0);
11713 parser->last_uni = ls + (proto->last_uni >= ols ?
11714 proto->last_uni - ols : 0);
11715 parser->last_lop = ls + (proto->last_lop >= ols ?
11716 proto->last_lop - ols : 0);
11718 parser->bufend = ls + SvCUR(parser->linestr);
11721 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11725 parser->endwhite = proto->endwhite;
11726 parser->faketokens = proto->faketokens;
11727 parser->lasttoke = proto->lasttoke;
11728 parser->nextwhite = proto->nextwhite;
11729 parser->realtokenstart = proto->realtokenstart;
11730 parser->skipwhite = proto->skipwhite;
11731 parser->thisclose = proto->thisclose;
11732 parser->thismad = proto->thismad;
11733 parser->thisopen = proto->thisopen;
11734 parser->thisstuff = proto->thisstuff;
11735 parser->thistoken = proto->thistoken;
11736 parser->thiswhite = proto->thiswhite;
11738 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11739 parser->curforce = proto->curforce;
11741 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11742 Copy(proto->nexttype, parser->nexttype, 5, I32);
11743 parser->nexttoke = proto->nexttoke;
11746 /* XXX should clone saved_curcop here, but we aren't passed
11747 * proto_perl; so do it in perl_clone_using instead */
11753 /* duplicate a file handle */
11756 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11760 PERL_ARGS_ASSERT_FP_DUP;
11761 PERL_UNUSED_ARG(type);
11764 return (PerlIO*)NULL;
11766 /* look for it in the table first */
11767 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11771 /* create anew and remember what it is */
11772 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11773 ptr_table_store(PL_ptr_table, fp, ret);
11777 /* duplicate a directory handle */
11780 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11787 const Direntry_t *dirent;
11788 char smallbuf[256];
11794 PERL_UNUSED_CONTEXT;
11795 PERL_ARGS_ASSERT_DIRP_DUP;
11800 /* look for it in the table first */
11801 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11807 PERL_UNUSED_ARG(param);
11811 /* open the current directory (so we can switch back) */
11812 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11814 /* chdir to our dir handle and open the present working directory */
11815 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11816 PerlDir_close(pwd);
11817 return (DIR *)NULL;
11819 /* Now we should have two dir handles pointing to the same dir. */
11821 /* Be nice to the calling code and chdir back to where we were. */
11822 rc = fchdir(my_dirfd(pwd));
11823 /* XXX If this fails, then what? */
11824 PERL_UNUSED_VAR(rc);
11826 /* We have no need of the pwd handle any more. */
11827 PerlDir_close(pwd);
11830 # define d_namlen(d) (d)->d_namlen
11832 # define d_namlen(d) strlen((d)->d_name)
11834 /* Iterate once through dp, to get the file name at the current posi-
11835 tion. Then step back. */
11836 pos = PerlDir_tell(dp);
11837 if ((dirent = PerlDir_read(dp))) {
11838 len = d_namlen(dirent);
11839 if (len <= sizeof smallbuf) name = smallbuf;
11840 else Newx(name, len, char);
11841 Move(dirent->d_name, name, len, char);
11843 PerlDir_seek(dp, pos);
11845 /* Iterate through the new dir handle, till we find a file with the
11847 if (!dirent) /* just before the end */
11849 pos = PerlDir_tell(ret);
11850 if (PerlDir_read(ret)) continue; /* not there yet */
11851 PerlDir_seek(ret, pos); /* step back */
11855 const long pos0 = PerlDir_tell(ret);
11857 pos = PerlDir_tell(ret);
11858 if ((dirent = PerlDir_read(ret))) {
11859 if (len == d_namlen(dirent)
11860 && memEQ(name, dirent->d_name, len)) {
11862 PerlDir_seek(ret, pos); /* step back */
11865 /* else we are not there yet; keep iterating */
11867 else { /* This is not meant to happen. The best we can do is
11868 reset the iterator to the beginning. */
11869 PerlDir_seek(ret, pos0);
11876 if (name && name != smallbuf)
11881 ret = win32_dirp_dup(dp, param);
11884 /* pop it in the pointer table */
11886 ptr_table_store(PL_ptr_table, dp, ret);
11891 /* duplicate a typeglob */
11894 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11898 PERL_ARGS_ASSERT_GP_DUP;
11902 /* look for it in the table first */
11903 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11907 /* create anew and remember what it is */
11909 ptr_table_store(PL_ptr_table, gp, ret);
11912 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11913 on Newxz() to do this for us. */
11914 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11915 ret->gp_io = io_dup_inc(gp->gp_io, param);
11916 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11917 ret->gp_av = av_dup_inc(gp->gp_av, param);
11918 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11919 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11920 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11921 ret->gp_cvgen = gp->gp_cvgen;
11922 ret->gp_line = gp->gp_line;
11923 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11927 /* duplicate a chain of magic */
11930 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11932 MAGIC *mgret = NULL;
11933 MAGIC **mgprev_p = &mgret;
11935 PERL_ARGS_ASSERT_MG_DUP;
11937 for (; mg; mg = mg->mg_moremagic) {
11940 if ((param->flags & CLONEf_JOIN_IN)
11941 && mg->mg_type == PERL_MAGIC_backref)
11942 /* when joining, we let the individual SVs add themselves to
11943 * backref as needed. */
11946 Newx(nmg, 1, MAGIC);
11948 mgprev_p = &(nmg->mg_moremagic);
11950 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11951 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11952 from the original commit adding Perl_mg_dup() - revision 4538.
11953 Similarly there is the annotation "XXX random ptr?" next to the
11954 assignment to nmg->mg_ptr. */
11957 /* FIXME for plugins
11958 if (nmg->mg_type == PERL_MAGIC_qr) {
11959 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11963 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11964 ? nmg->mg_type == PERL_MAGIC_backref
11965 /* The backref AV has its reference
11966 * count deliberately bumped by 1 */
11967 ? SvREFCNT_inc(av_dup_inc((const AV *)
11968 nmg->mg_obj, param))
11969 : sv_dup_inc(nmg->mg_obj, param)
11970 : sv_dup(nmg->mg_obj, param);
11972 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11973 if (nmg->mg_len > 0) {
11974 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11975 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11976 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11978 AMT * const namtp = (AMT*)nmg->mg_ptr;
11979 sv_dup_inc_multiple((SV**)(namtp->table),
11980 (SV**)(namtp->table), NofAMmeth, param);
11983 else if (nmg->mg_len == HEf_SVKEY)
11984 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11986 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11987 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11993 #endif /* USE_ITHREADS */
11995 struct ptr_tbl_arena {
11996 struct ptr_tbl_arena *next;
11997 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
12000 /* create a new pointer-mapping table */
12003 Perl_ptr_table_new(pTHX)
12006 PERL_UNUSED_CONTEXT;
12008 Newx(tbl, 1, PTR_TBL_t);
12009 tbl->tbl_max = 511;
12010 tbl->tbl_items = 0;
12011 tbl->tbl_arena = NULL;
12012 tbl->tbl_arena_next = NULL;
12013 tbl->tbl_arena_end = NULL;
12014 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
12018 #define PTR_TABLE_HASH(ptr) \
12019 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
12021 /* map an existing pointer using a table */
12023 STATIC PTR_TBL_ENT_t *
12024 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
12026 PTR_TBL_ENT_t *tblent;
12027 const UV hash = PTR_TABLE_HASH(sv);
12029 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
12031 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
12032 for (; tblent; tblent = tblent->next) {
12033 if (tblent->oldval == sv)
12040 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
12042 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
12044 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
12045 PERL_UNUSED_CONTEXT;
12047 return tblent ? tblent->newval : NULL;
12050 /* add a new entry to a pointer-mapping table */
12053 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
12055 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
12057 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
12058 PERL_UNUSED_CONTEXT;
12061 tblent->newval = newsv;
12063 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
12065 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
12066 struct ptr_tbl_arena *new_arena;
12068 Newx(new_arena, 1, struct ptr_tbl_arena);
12069 new_arena->next = tbl->tbl_arena;
12070 tbl->tbl_arena = new_arena;
12071 tbl->tbl_arena_next = new_arena->array;
12072 tbl->tbl_arena_end = new_arena->array
12073 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
12076 tblent = tbl->tbl_arena_next++;
12078 tblent->oldval = oldsv;
12079 tblent->newval = newsv;
12080 tblent->next = tbl->tbl_ary[entry];
12081 tbl->tbl_ary[entry] = tblent;
12083 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
12084 ptr_table_split(tbl);
12088 /* double the hash bucket size of an existing ptr table */
12091 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
12093 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
12094 const UV oldsize = tbl->tbl_max + 1;
12095 UV newsize = oldsize * 2;
12098 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
12099 PERL_UNUSED_CONTEXT;
12101 Renew(ary, newsize, PTR_TBL_ENT_t*);
12102 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
12103 tbl->tbl_max = --newsize;
12104 tbl->tbl_ary = ary;
12105 for (i=0; i < oldsize; i++, ary++) {
12106 PTR_TBL_ENT_t **entp = ary;
12107 PTR_TBL_ENT_t *ent = *ary;
12108 PTR_TBL_ENT_t **curentp;
12111 curentp = ary + oldsize;
12113 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
12115 ent->next = *curentp;
12125 /* remove all the entries from a ptr table */
12126 /* Deprecated - will be removed post 5.14 */
12129 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
12131 if (tbl && tbl->tbl_items) {
12132 struct ptr_tbl_arena *arena = tbl->tbl_arena;
12134 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
12137 struct ptr_tbl_arena *next = arena->next;
12143 tbl->tbl_items = 0;
12144 tbl->tbl_arena = NULL;
12145 tbl->tbl_arena_next = NULL;
12146 tbl->tbl_arena_end = NULL;
12150 /* clear and free a ptr table */
12153 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
12155 struct ptr_tbl_arena *arena;
12161 arena = tbl->tbl_arena;
12164 struct ptr_tbl_arena *next = arena->next;
12170 Safefree(tbl->tbl_ary);
12174 #if defined(USE_ITHREADS)
12177 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
12179 PERL_ARGS_ASSERT_RVPV_DUP;
12181 assert(!isREGEXP(sstr));
12183 if (SvWEAKREF(sstr)) {
12184 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
12185 if (param->flags & CLONEf_JOIN_IN) {
12186 /* if joining, we add any back references individually rather
12187 * than copying the whole backref array */
12188 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
12192 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
12194 else if (SvPVX_const(sstr)) {
12195 /* Has something there */
12197 /* Normal PV - clone whole allocated space */
12198 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
12199 /* sstr may not be that normal, but actually copy on write.
12200 But we are a true, independent SV, so: */
12204 /* Special case - not normally malloced for some reason */
12205 if (isGV_with_GP(sstr)) {
12206 /* Don't need to do anything here. */
12208 else if ((SvIsCOW(sstr))) {
12209 /* A "shared" PV - clone it as "shared" PV */
12211 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
12215 /* Some other special case - random pointer */
12216 SvPV_set(dstr, (char *) SvPVX_const(sstr));
12221 /* Copy the NULL */
12222 SvPV_set(dstr, NULL);
12226 /* duplicate a list of SVs. source and dest may point to the same memory. */
12228 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
12229 SSize_t items, CLONE_PARAMS *const param)
12231 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
12233 while (items-- > 0) {
12234 *dest++ = sv_dup_inc(*source++, param);
12240 /* duplicate an SV of any type (including AV, HV etc) */
12243 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12248 PERL_ARGS_ASSERT_SV_DUP_COMMON;
12250 if (SvTYPE(sstr) == (svtype)SVTYPEMASK) {
12251 #ifdef DEBUG_LEAKING_SCALARS_ABORT
12256 /* look for it in the table first */
12257 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
12261 if(param->flags & CLONEf_JOIN_IN) {
12262 /** We are joining here so we don't want do clone
12263 something that is bad **/
12264 if (SvTYPE(sstr) == SVt_PVHV) {
12265 const HEK * const hvname = HvNAME_HEK(sstr);
12267 /** don't clone stashes if they already exist **/
12268 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
12269 HEK_UTF8(hvname) ? SVf_UTF8 : 0));
12270 ptr_table_store(PL_ptr_table, sstr, dstr);
12274 else if (SvTYPE(sstr) == SVt_PVGV && !SvFAKE(sstr)) {
12275 HV *stash = GvSTASH(sstr);
12276 const HEK * hvname;
12277 if (stash && (hvname = HvNAME_HEK(stash))) {
12278 /** don't clone GVs if they already exist **/
12280 stash = gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
12281 HEK_UTF8(hvname) ? SVf_UTF8 : 0);
12283 stash, GvNAME(sstr),
12289 if (svp && *svp && SvTYPE(*svp) == SVt_PVGV) {
12290 ptr_table_store(PL_ptr_table, sstr, *svp);
12297 /* create anew and remember what it is */
12300 #ifdef DEBUG_LEAKING_SCALARS
12301 dstr->sv_debug_optype = sstr->sv_debug_optype;
12302 dstr->sv_debug_line = sstr->sv_debug_line;
12303 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
12304 dstr->sv_debug_parent = (SV*)sstr;
12305 FREE_SV_DEBUG_FILE(dstr);
12306 dstr->sv_debug_file = savesharedpv(sstr->sv_debug_file);
12309 ptr_table_store(PL_ptr_table, sstr, dstr);
12312 SvFLAGS(dstr) = SvFLAGS(sstr);
12313 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
12314 SvREFCNT(dstr) = 0; /* must be before any other dups! */
12317 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
12318 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
12319 (void*)PL_watch_pvx, SvPVX_const(sstr));
12322 /* don't clone objects whose class has asked us not to */
12323 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
12328 switch (SvTYPE(sstr)) {
12330 SvANY(dstr) = NULL;
12333 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
12335 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
12337 SvIV_set(dstr, SvIVX(sstr));
12341 SvANY(dstr) = new_XNV();
12342 SvNV_set(dstr, SvNVX(sstr));
12346 /* These are all the types that need complex bodies allocating. */
12348 const svtype sv_type = SvTYPE(sstr);
12349 const struct body_details *const sv_type_details
12350 = bodies_by_type + sv_type;
12354 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
12370 assert(sv_type_details->body_size);
12371 if (sv_type_details->arena) {
12372 new_body_inline(new_body, sv_type);
12374 = (void*)((char*)new_body - sv_type_details->offset);
12376 new_body = new_NOARENA(sv_type_details);
12380 SvANY(dstr) = new_body;
12383 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
12384 ((char*)SvANY(dstr)) + sv_type_details->offset,
12385 sv_type_details->copy, char);
12387 Copy(((char*)SvANY(sstr)),
12388 ((char*)SvANY(dstr)),
12389 sv_type_details->body_size + sv_type_details->offset, char);
12392 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
12393 && !isGV_with_GP(dstr)
12395 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
12396 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
12398 /* The Copy above means that all the source (unduplicated) pointers
12399 are now in the destination. We can check the flags and the
12400 pointers in either, but it's possible that there's less cache
12401 missing by always going for the destination.
12402 FIXME - instrument and check that assumption */
12403 if (sv_type >= SVt_PVMG) {
12404 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
12405 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
12406 } else if (sv_type == SVt_PVAV && AvPAD_NAMELIST(dstr)) {
12408 } else if (SvMAGIC(dstr))
12409 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
12410 if (SvOBJECT(dstr) && SvSTASH(dstr))
12411 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
12412 else SvSTASH_set(dstr, 0); /* don't copy DESTROY cache */
12415 /* The cast silences a GCC warning about unhandled types. */
12416 switch ((int)sv_type) {
12427 /* FIXME for plugins */
12428 dstr->sv_u.svu_rx = ((REGEXP *)dstr)->sv_any;
12429 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
12432 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
12433 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
12434 LvTARG(dstr) = dstr;
12435 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
12436 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
12438 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
12439 if (isREGEXP(sstr)) goto duprex;
12441 /* non-GP case already handled above */
12442 if(isGV_with_GP(sstr)) {
12443 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
12444 /* Don't call sv_add_backref here as it's going to be
12445 created as part of the magic cloning of the symbol
12446 table--unless this is during a join and the stash
12447 is not actually being cloned. */
12448 /* Danger Will Robinson - GvGP(dstr) isn't initialised
12449 at the point of this comment. */
12450 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
12451 if (param->flags & CLONEf_JOIN_IN)
12452 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
12453 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
12454 (void)GpREFCNT_inc(GvGP(dstr));
12458 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
12459 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
12460 /* I have no idea why fake dirp (rsfps)
12461 should be treated differently but otherwise
12462 we end up with leaks -- sky*/
12463 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
12464 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
12465 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
12467 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
12468 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
12469 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
12470 if (IoDIRP(dstr)) {
12471 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
12474 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
12476 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
12478 if (IoOFP(dstr) == IoIFP(sstr))
12479 IoOFP(dstr) = IoIFP(dstr);
12481 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
12482 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
12483 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
12484 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
12487 /* avoid cloning an empty array */
12488 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
12489 SV **dst_ary, **src_ary;
12490 SSize_t items = AvFILLp((const AV *)sstr) + 1;
12492 src_ary = AvARRAY((const AV *)sstr);
12493 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
12494 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
12495 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
12496 AvALLOC((const AV *)dstr) = dst_ary;
12497 if (AvREAL((const AV *)sstr)) {
12498 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
12502 while (items-- > 0)
12503 *dst_ary++ = sv_dup(*src_ary++, param);
12505 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
12506 while (items-- > 0) {
12507 *dst_ary++ = &PL_sv_undef;
12511 AvARRAY(MUTABLE_AV(dstr)) = NULL;
12512 AvALLOC((const AV *)dstr) = (SV**)NULL;
12513 AvMAX( (const AV *)dstr) = -1;
12514 AvFILLp((const AV *)dstr) = -1;
12518 if (HvARRAY((const HV *)sstr)) {
12520 const bool sharekeys = !!HvSHAREKEYS(sstr);
12521 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
12522 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
12524 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
12525 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
12527 HvARRAY(dstr) = (HE**)darray;
12528 while (i <= sxhv->xhv_max) {
12529 const HE * const source = HvARRAY(sstr)[i];
12530 HvARRAY(dstr)[i] = source
12531 ? he_dup(source, sharekeys, param) : 0;
12535 const struct xpvhv_aux * const saux = HvAUX(sstr);
12536 struct xpvhv_aux * const daux = HvAUX(dstr);
12537 /* This flag isn't copied. */
12540 if (saux->xhv_name_count) {
12541 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
12543 = saux->xhv_name_count < 0
12544 ? -saux->xhv_name_count
12545 : saux->xhv_name_count;
12546 HEK **shekp = sname + count;
12548 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
12549 dhekp = daux->xhv_name_u.xhvnameu_names + count;
12550 while (shekp-- > sname) {
12552 *dhekp = hek_dup(*shekp, param);
12556 daux->xhv_name_u.xhvnameu_name
12557 = hek_dup(saux->xhv_name_u.xhvnameu_name,
12560 daux->xhv_name_count = saux->xhv_name_count;
12562 daux->xhv_fill_lazy = saux->xhv_fill_lazy;
12563 daux->xhv_riter = saux->xhv_riter;
12564 daux->xhv_eiter = saux->xhv_eiter
12565 ? he_dup(saux->xhv_eiter,
12566 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
12567 /* backref array needs refcnt=2; see sv_add_backref */
12568 daux->xhv_backreferences =
12569 (param->flags & CLONEf_JOIN_IN)
12570 /* when joining, we let the individual GVs and
12571 * CVs add themselves to backref as
12572 * needed. This avoids pulling in stuff
12573 * that isn't required, and simplifies the
12574 * case where stashes aren't cloned back
12575 * if they already exist in the parent
12578 : saux->xhv_backreferences
12579 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
12580 ? MUTABLE_AV(SvREFCNT_inc(
12581 sv_dup_inc((const SV *)
12582 saux->xhv_backreferences, param)))
12583 : MUTABLE_AV(sv_dup((const SV *)
12584 saux->xhv_backreferences, param))
12587 daux->xhv_mro_meta = saux->xhv_mro_meta
12588 ? mro_meta_dup(saux->xhv_mro_meta, param)
12591 /* Record stashes for possible cloning in Perl_clone(). */
12593 av_push(param->stashes, dstr);
12597 HvARRAY(MUTABLE_HV(dstr)) = NULL;
12600 if (!(param->flags & CLONEf_COPY_STACKS)) {
12605 /* NOTE: not refcounted */
12606 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12607 hv_dup(CvSTASH(dstr), param);
12608 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12609 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12610 if (!CvISXSUB(dstr)) {
12612 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12614 CvSLABBED_off(dstr);
12615 } else if (CvCONST(dstr)) {
12616 CvXSUBANY(dstr).any_ptr =
12617 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12619 assert(!CvSLABBED(dstr));
12620 if (CvDYNFILE(dstr)) CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12622 SvANY((CV *)dstr)->xcv_gv_u.xcv_hek =
12623 share_hek_hek(CvNAME_HEK((CV *)sstr));
12624 /* don't dup if copying back - CvGV isn't refcounted, so the
12625 * duped GV may never be freed. A bit of a hack! DAPM */
12627 SvANY(MUTABLE_CV(dstr))->xcv_gv_u.xcv_gv =
12629 ? gv_dup_inc(CvGV(sstr), param)
12630 : (param->flags & CLONEf_JOIN_IN)
12632 : gv_dup(CvGV(sstr), param);
12634 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12636 CvWEAKOUTSIDE(sstr)
12637 ? cv_dup( CvOUTSIDE(dstr), param)
12638 : cv_dup_inc(CvOUTSIDE(dstr), param);
12648 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12650 PERL_ARGS_ASSERT_SV_DUP_INC;
12651 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12655 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12657 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12658 PERL_ARGS_ASSERT_SV_DUP;
12660 /* Track every SV that (at least initially) had a reference count of 0.
12661 We need to do this by holding an actual reference to it in this array.
12662 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12663 (akin to the stashes hash, and the perl stack), we come unstuck if
12664 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12665 thread) is manipulated in a CLONE method, because CLONE runs before the
12666 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12667 (and fix things up by giving each a reference via the temps stack).
12668 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12669 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12670 before the walk of unreferenced happens and a reference to that is SV
12671 added to the temps stack. At which point we have the same SV considered
12672 to be in use, and free to be re-used. Not good.
12674 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12675 assert(param->unreferenced);
12676 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12682 /* duplicate a context */
12685 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12687 PERL_CONTEXT *ncxs;
12689 PERL_ARGS_ASSERT_CX_DUP;
12692 return (PERL_CONTEXT*)NULL;
12694 /* look for it in the table first */
12695 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12699 /* create anew and remember what it is */
12700 Newx(ncxs, max + 1, PERL_CONTEXT);
12701 ptr_table_store(PL_ptr_table, cxs, ncxs);
12702 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12705 PERL_CONTEXT * const ncx = &ncxs[ix];
12706 if (CxTYPE(ncx) == CXt_SUBST) {
12707 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12710 ncx->blk_oldcop = (COP*)any_dup(ncx->blk_oldcop, param->proto_perl);
12711 switch (CxTYPE(ncx)) {
12713 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12714 ? cv_dup_inc(ncx->blk_sub.cv, param)
12715 : cv_dup(ncx->blk_sub.cv,param));
12716 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12717 ? av_dup_inc(ncx->blk_sub.argarray,
12720 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12722 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12723 ncx->blk_sub.oldcomppad);
12726 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12728 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12729 ncx->blk_eval.cv = cv_dup(ncx->blk_eval.cv, param);
12731 case CXt_LOOP_LAZYSV:
12732 ncx->blk_loop.state_u.lazysv.end
12733 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12734 /* We are taking advantage of av_dup_inc and sv_dup_inc
12735 actually being the same function, and order equivalence of
12737 We can assert the later [but only at run time :-(] */
12738 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12739 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12741 ncx->blk_loop.state_u.ary.ary
12742 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12743 case CXt_LOOP_LAZYIV:
12744 case CXt_LOOP_PLAIN:
12745 if (CxPADLOOP(ncx)) {
12746 ncx->blk_loop.itervar_u.oldcomppad
12747 = (PAD*)ptr_table_fetch(PL_ptr_table,
12748 ncx->blk_loop.itervar_u.oldcomppad);
12750 ncx->blk_loop.itervar_u.gv
12751 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12756 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12757 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12758 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12773 /* duplicate a stack info structure */
12776 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12780 PERL_ARGS_ASSERT_SI_DUP;
12783 return (PERL_SI*)NULL;
12785 /* look for it in the table first */
12786 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12790 /* create anew and remember what it is */
12791 Newxz(nsi, 1, PERL_SI);
12792 ptr_table_store(PL_ptr_table, si, nsi);
12794 nsi->si_stack = av_dup_inc(si->si_stack, param);
12795 nsi->si_cxix = si->si_cxix;
12796 nsi->si_cxmax = si->si_cxmax;
12797 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12798 nsi->si_type = si->si_type;
12799 nsi->si_prev = si_dup(si->si_prev, param);
12800 nsi->si_next = si_dup(si->si_next, param);
12801 nsi->si_markoff = si->si_markoff;
12806 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12807 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12808 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12809 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12810 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12811 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12812 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12813 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12814 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12815 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12816 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12817 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12818 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12819 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12820 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12821 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12824 #define pv_dup_inc(p) SAVEPV(p)
12825 #define pv_dup(p) SAVEPV(p)
12826 #define svp_dup_inc(p,pp) any_dup(p,pp)
12828 /* map any object to the new equivent - either something in the
12829 * ptr table, or something in the interpreter structure
12833 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12837 PERL_ARGS_ASSERT_ANY_DUP;
12840 return (void*)NULL;
12842 /* look for it in the table first */
12843 ret = ptr_table_fetch(PL_ptr_table, v);
12847 /* see if it is part of the interpreter structure */
12848 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12849 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12857 /* duplicate the save stack */
12860 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12863 ANY * const ss = proto_perl->Isavestack;
12864 const I32 max = proto_perl->Isavestack_max;
12865 I32 ix = proto_perl->Isavestack_ix;
12878 void (*dptr) (void*);
12879 void (*dxptr) (pTHX_ void*);
12881 PERL_ARGS_ASSERT_SS_DUP;
12883 Newxz(nss, max, ANY);
12886 const UV uv = POPUV(ss,ix);
12887 const U8 type = (U8)uv & SAVE_MASK;
12889 TOPUV(nss,ix) = uv;
12891 case SAVEt_CLEARSV:
12892 case SAVEt_CLEARPADRANGE:
12894 case SAVEt_HELEM: /* hash element */
12895 sv = (const SV *)POPPTR(ss,ix);
12896 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12898 case SAVEt_ITEM: /* normal string */
12899 case SAVEt_GVSV: /* scalar slot in GV */
12900 case SAVEt_SV: /* scalar reference */
12901 sv = (const SV *)POPPTR(ss,ix);
12902 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12905 case SAVEt_MORTALIZESV:
12906 case SAVEt_READONLY_OFF:
12907 sv = (const SV *)POPPTR(ss,ix);
12908 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12910 case SAVEt_SHARED_PVREF: /* char* in shared space */
12911 c = (char*)POPPTR(ss,ix);
12912 TOPPTR(nss,ix) = savesharedpv(c);
12913 ptr = POPPTR(ss,ix);
12914 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12916 case SAVEt_GENERIC_SVREF: /* generic sv */
12917 case SAVEt_SVREF: /* scalar reference */
12918 sv = (const SV *)POPPTR(ss,ix);
12919 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12920 ptr = POPPTR(ss,ix);
12921 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12923 case SAVEt_GVSLOT: /* any slot in GV */
12924 sv = (const SV *)POPPTR(ss,ix);
12925 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12926 ptr = POPPTR(ss,ix);
12927 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12928 sv = (const SV *)POPPTR(ss,ix);
12929 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12931 case SAVEt_HV: /* hash reference */
12932 case SAVEt_AV: /* array reference */
12933 sv = (const SV *) POPPTR(ss,ix);
12934 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12936 case SAVEt_COMPPAD:
12938 sv = (const SV *) POPPTR(ss,ix);
12939 TOPPTR(nss,ix) = sv_dup(sv, param);
12941 case SAVEt_INT: /* int reference */
12942 ptr = POPPTR(ss,ix);
12943 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12944 intval = (int)POPINT(ss,ix);
12945 TOPINT(nss,ix) = intval;
12947 case SAVEt_LONG: /* long reference */
12948 ptr = POPPTR(ss,ix);
12949 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12950 longval = (long)POPLONG(ss,ix);
12951 TOPLONG(nss,ix) = longval;
12953 case SAVEt_I32: /* I32 reference */
12954 ptr = POPPTR(ss,ix);
12955 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12957 TOPINT(nss,ix) = i;
12959 case SAVEt_IV: /* IV reference */
12960 case SAVEt_STRLEN: /* STRLEN/size_t ref */
12961 ptr = POPPTR(ss,ix);
12962 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12964 TOPIV(nss,ix) = iv;
12966 case SAVEt_HPTR: /* HV* reference */
12967 case SAVEt_APTR: /* AV* reference */
12968 case SAVEt_SPTR: /* SV* reference */
12969 ptr = POPPTR(ss,ix);
12970 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12971 sv = (const SV *)POPPTR(ss,ix);
12972 TOPPTR(nss,ix) = sv_dup(sv, param);
12974 case SAVEt_VPTR: /* random* reference */
12975 ptr = POPPTR(ss,ix);
12976 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12978 case SAVEt_INT_SMALL:
12979 case SAVEt_I32_SMALL:
12980 case SAVEt_I16: /* I16 reference */
12981 case SAVEt_I8: /* I8 reference */
12983 ptr = POPPTR(ss,ix);
12984 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12986 case SAVEt_GENERIC_PVREF: /* generic char* */
12987 case SAVEt_PPTR: /* char* reference */
12988 ptr = POPPTR(ss,ix);
12989 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12990 c = (char*)POPPTR(ss,ix);
12991 TOPPTR(nss,ix) = pv_dup(c);
12993 case SAVEt_GP: /* scalar reference */
12994 gp = (GP*)POPPTR(ss,ix);
12995 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12996 (void)GpREFCNT_inc(gp);
12997 gv = (const GV *)POPPTR(ss,ix);
12998 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
13001 ptr = POPPTR(ss,ix);
13002 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
13003 /* these are assumed to be refcounted properly */
13005 switch (((OP*)ptr)->op_type) {
13007 case OP_LEAVESUBLV:
13011 case OP_LEAVEWRITE:
13012 TOPPTR(nss,ix) = ptr;
13015 (void) OpREFCNT_inc(o);
13019 TOPPTR(nss,ix) = NULL;
13024 TOPPTR(nss,ix) = NULL;
13026 case SAVEt_FREECOPHH:
13027 ptr = POPPTR(ss,ix);
13028 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
13030 case SAVEt_ADELETE:
13031 av = (const AV *)POPPTR(ss,ix);
13032 TOPPTR(nss,ix) = av_dup_inc(av, param);
13034 TOPINT(nss,ix) = i;
13037 hv = (const HV *)POPPTR(ss,ix);
13038 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
13040 TOPINT(nss,ix) = i;
13043 c = (char*)POPPTR(ss,ix);
13044 TOPPTR(nss,ix) = pv_dup_inc(c);
13046 case SAVEt_STACK_POS: /* Position on Perl stack */
13048 TOPINT(nss,ix) = i;
13050 case SAVEt_DESTRUCTOR:
13051 ptr = POPPTR(ss,ix);
13052 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
13053 dptr = POPDPTR(ss,ix);
13054 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
13055 any_dup(FPTR2DPTR(void *, dptr),
13058 case SAVEt_DESTRUCTOR_X:
13059 ptr = POPPTR(ss,ix);
13060 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
13061 dxptr = POPDXPTR(ss,ix);
13062 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
13063 any_dup(FPTR2DPTR(void *, dxptr),
13066 case SAVEt_REGCONTEXT:
13068 ix -= uv >> SAVE_TIGHT_SHIFT;
13070 case SAVEt_AELEM: /* array element */
13071 sv = (const SV *)POPPTR(ss,ix);
13072 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
13074 TOPINT(nss,ix) = i;
13075 av = (const AV *)POPPTR(ss,ix);
13076 TOPPTR(nss,ix) = av_dup_inc(av, param);
13079 ptr = POPPTR(ss,ix);
13080 TOPPTR(nss,ix) = ptr;
13083 ptr = POPPTR(ss,ix);
13084 ptr = cophh_copy((COPHH*)ptr);
13085 TOPPTR(nss,ix) = ptr;
13087 TOPINT(nss,ix) = i;
13088 if (i & HINT_LOCALIZE_HH) {
13089 hv = (const HV *)POPPTR(ss,ix);
13090 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
13093 case SAVEt_PADSV_AND_MORTALIZE:
13094 longval = (long)POPLONG(ss,ix);
13095 TOPLONG(nss,ix) = longval;
13096 ptr = POPPTR(ss,ix);
13097 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
13098 sv = (const SV *)POPPTR(ss,ix);
13099 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
13101 case SAVEt_SET_SVFLAGS:
13103 TOPINT(nss,ix) = i;
13105 TOPINT(nss,ix) = i;
13106 sv = (const SV *)POPPTR(ss,ix);
13107 TOPPTR(nss,ix) = sv_dup(sv, param);
13109 case SAVEt_COMPILE_WARNINGS:
13110 ptr = POPPTR(ss,ix);
13111 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
13114 ptr = POPPTR(ss,ix);
13115 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
13119 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
13127 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
13128 * flag to the result. This is done for each stash before cloning starts,
13129 * so we know which stashes want their objects cloned */
13132 do_mark_cloneable_stash(pTHX_ SV *const sv)
13134 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
13136 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
13137 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
13138 if (cloner && GvCV(cloner)) {
13145 mXPUSHs(newSVhek(hvname));
13147 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
13154 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
13162 =for apidoc perl_clone
13164 Create and return a new interpreter by cloning the current one.
13166 perl_clone takes these flags as parameters:
13168 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
13169 without it we only clone the data and zero the stacks,
13170 with it we copy the stacks and the new perl interpreter is
13171 ready to run at the exact same point as the previous one.
13172 The pseudo-fork code uses COPY_STACKS while the
13173 threads->create doesn't.
13175 CLONEf_KEEP_PTR_TABLE -
13176 perl_clone keeps a ptr_table with the pointer of the old
13177 variable as a key and the new variable as a value,
13178 this allows it to check if something has been cloned and not
13179 clone it again but rather just use the value and increase the
13180 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
13181 the ptr_table using the function
13182 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
13183 reason to keep it around is if you want to dup some of your own
13184 variable who are outside the graph perl scans, example of this
13185 code is in threads.xs create.
13187 CLONEf_CLONE_HOST -
13188 This is a win32 thing, it is ignored on unix, it tells perls
13189 win32host code (which is c++) to clone itself, this is needed on
13190 win32 if you want to run two threads at the same time,
13191 if you just want to do some stuff in a separate perl interpreter
13192 and then throw it away and return to the original one,
13193 you don't need to do anything.
13198 /* XXX the above needs expanding by someone who actually understands it ! */
13199 EXTERN_C PerlInterpreter *
13200 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
13203 perl_clone(PerlInterpreter *proto_perl, UV flags)
13206 #ifdef PERL_IMPLICIT_SYS
13208 PERL_ARGS_ASSERT_PERL_CLONE;
13210 /* perlhost.h so we need to call into it
13211 to clone the host, CPerlHost should have a c interface, sky */
13213 if (flags & CLONEf_CLONE_HOST) {
13214 return perl_clone_host(proto_perl,flags);
13216 return perl_clone_using(proto_perl, flags,
13218 proto_perl->IMemShared,
13219 proto_perl->IMemParse,
13221 proto_perl->IStdIO,
13225 proto_perl->IProc);
13229 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
13230 struct IPerlMem* ipM, struct IPerlMem* ipMS,
13231 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
13232 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
13233 struct IPerlDir* ipD, struct IPerlSock* ipS,
13234 struct IPerlProc* ipP)
13236 /* XXX many of the string copies here can be optimized if they're
13237 * constants; they need to be allocated as common memory and just
13238 * their pointers copied. */
13241 CLONE_PARAMS clone_params;
13242 CLONE_PARAMS* const param = &clone_params;
13244 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
13246 PERL_ARGS_ASSERT_PERL_CLONE_USING;
13247 #else /* !PERL_IMPLICIT_SYS */
13249 CLONE_PARAMS clone_params;
13250 CLONE_PARAMS* param = &clone_params;
13251 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
13253 PERL_ARGS_ASSERT_PERL_CLONE;
13254 #endif /* PERL_IMPLICIT_SYS */
13256 /* for each stash, determine whether its objects should be cloned */
13257 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
13258 PERL_SET_THX(my_perl);
13261 PoisonNew(my_perl, 1, PerlInterpreter);
13264 PL_defstash = NULL; /* may be used by perl malloc() */
13267 PL_scopestack_name = 0;
13269 PL_savestack_ix = 0;
13270 PL_savestack_max = -1;
13271 PL_sig_pending = 0;
13273 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
13274 # ifdef DEBUG_LEAKING_SCALARS
13275 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
13277 #else /* !DEBUGGING */
13278 Zero(my_perl, 1, PerlInterpreter);
13279 #endif /* DEBUGGING */
13281 #ifdef PERL_IMPLICIT_SYS
13282 /* host pointers */
13284 PL_MemShared = ipMS;
13285 PL_MemParse = ipMP;
13292 #endif /* PERL_IMPLICIT_SYS */
13295 param->flags = flags;
13296 /* Nothing in the core code uses this, but we make it available to
13297 extensions (using mg_dup). */
13298 param->proto_perl = proto_perl;
13299 /* Likely nothing will use this, but it is initialised to be consistent
13300 with Perl_clone_params_new(). */
13301 param->new_perl = my_perl;
13302 param->unreferenced = NULL;
13305 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
13307 PL_body_arenas = NULL;
13308 Zero(&PL_body_roots, 1, PL_body_roots);
13312 PL_sv_arenaroot = NULL;
13314 PL_debug = proto_perl->Idebug;
13316 /* dbargs array probably holds garbage */
13319 PL_compiling = proto_perl->Icompiling;
13321 /* pseudo environmental stuff */
13322 PL_origargc = proto_perl->Iorigargc;
13323 PL_origargv = proto_perl->Iorigargv;
13325 #if !NO_TAINT_SUPPORT
13326 /* Set tainting stuff before PerlIO_debug can possibly get called */
13327 PL_tainting = proto_perl->Itainting;
13328 PL_taint_warn = proto_perl->Itaint_warn;
13330 PL_tainting = FALSE;
13331 PL_taint_warn = FALSE;
13334 PL_minus_c = proto_perl->Iminus_c;
13336 PL_localpatches = proto_perl->Ilocalpatches;
13337 PL_splitstr = proto_perl->Isplitstr;
13338 PL_minus_n = proto_perl->Iminus_n;
13339 PL_minus_p = proto_perl->Iminus_p;
13340 PL_minus_l = proto_perl->Iminus_l;
13341 PL_minus_a = proto_perl->Iminus_a;
13342 PL_minus_E = proto_perl->Iminus_E;
13343 PL_minus_F = proto_perl->Iminus_F;
13344 PL_doswitches = proto_perl->Idoswitches;
13345 PL_dowarn = proto_perl->Idowarn;
13346 #ifdef PERL_SAWAMPERSAND
13347 PL_sawampersand = proto_perl->Isawampersand;
13349 PL_unsafe = proto_perl->Iunsafe;
13350 PL_perldb = proto_perl->Iperldb;
13351 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
13352 PL_exit_flags = proto_perl->Iexit_flags;
13354 /* XXX time(&PL_basetime) when asked for? */
13355 PL_basetime = proto_perl->Ibasetime;
13357 PL_maxsysfd = proto_perl->Imaxsysfd;
13358 PL_statusvalue = proto_perl->Istatusvalue;
13360 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
13362 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
13365 /* RE engine related */
13366 PL_regmatch_slab = NULL;
13367 PL_reg_curpm = NULL;
13369 PL_sub_generation = proto_perl->Isub_generation;
13371 /* funky return mechanisms */
13372 PL_forkprocess = proto_perl->Iforkprocess;
13374 /* internal state */
13375 PL_maxo = proto_perl->Imaxo;
13377 PL_main_start = proto_perl->Imain_start;
13378 PL_eval_root = proto_perl->Ieval_root;
13379 PL_eval_start = proto_perl->Ieval_start;
13381 PL_filemode = proto_perl->Ifilemode;
13382 PL_lastfd = proto_perl->Ilastfd;
13383 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
13386 PL_gensym = proto_perl->Igensym;
13388 PL_laststatval = proto_perl->Ilaststatval;
13389 PL_laststype = proto_perl->Ilaststype;
13392 PL_profiledata = NULL;
13394 PL_generation = proto_perl->Igeneration;
13396 PL_in_clean_objs = proto_perl->Iin_clean_objs;
13397 PL_in_clean_all = proto_perl->Iin_clean_all;
13399 PL_delaymagic_uid = proto_perl->Idelaymagic_uid;
13400 PL_delaymagic_euid = proto_perl->Idelaymagic_euid;
13401 PL_delaymagic_gid = proto_perl->Idelaymagic_gid;
13402 PL_delaymagic_egid = proto_perl->Idelaymagic_egid;
13403 PL_nomemok = proto_perl->Inomemok;
13404 PL_an = proto_perl->Ian;
13405 PL_evalseq = proto_perl->Ievalseq;
13406 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
13407 PL_origalen = proto_perl->Iorigalen;
13409 PL_sighandlerp = proto_perl->Isighandlerp;
13411 PL_runops = proto_perl->Irunops;
13413 PL_subline = proto_perl->Isubline;
13416 PL_cryptseen = proto_perl->Icryptseen;
13419 #ifdef USE_LOCALE_COLLATE
13420 PL_collation_ix = proto_perl->Icollation_ix;
13421 PL_collation_standard = proto_perl->Icollation_standard;
13422 PL_collxfrm_base = proto_perl->Icollxfrm_base;
13423 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
13424 #endif /* USE_LOCALE_COLLATE */
13426 #ifdef USE_LOCALE_NUMERIC
13427 PL_numeric_standard = proto_perl->Inumeric_standard;
13428 PL_numeric_local = proto_perl->Inumeric_local;
13429 #endif /* !USE_LOCALE_NUMERIC */
13431 /* Did the locale setup indicate UTF-8? */
13432 PL_utf8locale = proto_perl->Iutf8locale;
13433 /* Unicode features (see perlrun/-C) */
13434 PL_unicode = proto_perl->Iunicode;
13436 /* Pre-5.8 signals control */
13437 PL_signals = proto_perl->Isignals;
13439 /* times() ticks per second */
13440 PL_clocktick = proto_perl->Iclocktick;
13442 /* Recursion stopper for PerlIO_find_layer */
13443 PL_in_load_module = proto_perl->Iin_load_module;
13445 /* sort() routine */
13446 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
13448 /* Not really needed/useful since the reenrant_retint is "volatile",
13449 * but do it for consistency's sake. */
13450 PL_reentrant_retint = proto_perl->Ireentrant_retint;
13452 /* Hooks to shared SVs and locks. */
13453 PL_sharehook = proto_perl->Isharehook;
13454 PL_lockhook = proto_perl->Ilockhook;
13455 PL_unlockhook = proto_perl->Iunlockhook;
13456 PL_threadhook = proto_perl->Ithreadhook;
13457 PL_destroyhook = proto_perl->Idestroyhook;
13458 PL_signalhook = proto_perl->Isignalhook;
13460 PL_globhook = proto_perl->Iglobhook;
13463 PL_last_swash_hv = NULL; /* reinits on demand */
13464 PL_last_swash_klen = 0;
13465 PL_last_swash_key[0]= '\0';
13466 PL_last_swash_tmps = (U8*)NULL;
13467 PL_last_swash_slen = 0;
13469 PL_srand_called = proto_perl->Isrand_called;
13470 Copy(&(proto_perl->Irandom_state), &PL_random_state, 1, PL_RANDOM_STATE_TYPE);
13472 if (flags & CLONEf_COPY_STACKS) {
13473 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
13474 PL_tmps_ix = proto_perl->Itmps_ix;
13475 PL_tmps_max = proto_perl->Itmps_max;
13476 PL_tmps_floor = proto_perl->Itmps_floor;
13478 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13479 * NOTE: unlike the others! */
13480 PL_scopestack_ix = proto_perl->Iscopestack_ix;
13481 PL_scopestack_max = proto_perl->Iscopestack_max;
13483 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
13484 * NOTE: unlike the others! */
13485 PL_savestack_ix = proto_perl->Isavestack_ix;
13486 PL_savestack_max = proto_perl->Isavestack_max;
13489 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
13490 PL_top_env = &PL_start_env;
13492 PL_op = proto_perl->Iop;
13495 PL_Xpv = (XPV*)NULL;
13496 my_perl->Ina = proto_perl->Ina;
13498 PL_statbuf = proto_perl->Istatbuf;
13499 PL_statcache = proto_perl->Istatcache;
13502 PL_timesbuf = proto_perl->Itimesbuf;
13505 #if !NO_TAINT_SUPPORT
13506 PL_tainted = proto_perl->Itainted;
13508 PL_tainted = FALSE;
13510 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
13512 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
13514 PL_restartjmpenv = proto_perl->Irestartjmpenv;
13515 PL_restartop = proto_perl->Irestartop;
13516 PL_in_eval = proto_perl->Iin_eval;
13517 PL_delaymagic = proto_perl->Idelaymagic;
13518 PL_phase = proto_perl->Iphase;
13519 PL_localizing = proto_perl->Ilocalizing;
13521 PL_hv_fetch_ent_mh = NULL;
13522 PL_modcount = proto_perl->Imodcount;
13523 PL_lastgotoprobe = NULL;
13524 PL_dumpindent = proto_perl->Idumpindent;
13526 PL_efloatbuf = NULL; /* reinits on demand */
13527 PL_efloatsize = 0; /* reinits on demand */
13531 PL_colorset = 0; /* reinits PL_colors[] */
13532 /*PL_colors[6] = {0,0,0,0,0,0};*/
13534 /* Pluggable optimizer */
13535 PL_peepp = proto_perl->Ipeepp;
13536 PL_rpeepp = proto_perl->Irpeepp;
13537 /* op_free() hook */
13538 PL_opfreehook = proto_perl->Iopfreehook;
13540 #ifdef USE_REENTRANT_API
13541 /* XXX: things like -Dm will segfault here in perlio, but doing
13542 * PERL_SET_CONTEXT(proto_perl);
13543 * breaks too many other things
13545 Perl_reentrant_init(aTHX);
13548 /* create SV map for pointer relocation */
13549 PL_ptr_table = ptr_table_new();
13551 /* initialize these special pointers as early as possible */
13553 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13554 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13555 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13557 /* create (a non-shared!) shared string table */
13558 PL_strtab = newHV();
13559 HvSHAREKEYS_off(PL_strtab);
13560 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13561 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13563 Zero(PL_sv_consts, SV_CONSTS_COUNT, SV*);
13565 /* This PV will be free'd special way so must set it same way op.c does */
13566 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13567 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13569 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13570 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13571 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13572 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13574 param->stashes = newAV(); /* Setup array of objects to call clone on */
13575 /* This makes no difference to the implementation, as it always pushes
13576 and shifts pointers to other SVs without changing their reference
13577 count, with the array becoming empty before it is freed. However, it
13578 makes it conceptually clear what is going on, and will avoid some
13579 work inside av.c, filling slots between AvFILL() and AvMAX() with
13580 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13581 AvREAL_off(param->stashes);
13583 if (!(flags & CLONEf_COPY_STACKS)) {
13584 param->unreferenced = newAV();
13587 #ifdef PERLIO_LAYERS
13588 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13589 PerlIO_clone(aTHX_ proto_perl, param);
13592 PL_envgv = gv_dup_inc(proto_perl->Ienvgv, param);
13593 PL_incgv = gv_dup_inc(proto_perl->Iincgv, param);
13594 PL_hintgv = gv_dup_inc(proto_perl->Ihintgv, param);
13595 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13596 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13597 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13600 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13601 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13602 PL_inplace = SAVEPV(proto_perl->Iinplace);
13603 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13605 /* magical thingies */
13607 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13609 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13610 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13611 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13614 /* Clone the regex array */
13615 /* ORANGE FIXME for plugins, probably in the SV dup code.
13616 newSViv(PTR2IV(CALLREGDUPE(
13617 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13619 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13620 PL_regex_pad = AvARRAY(PL_regex_padav);
13622 PL_stashpadmax = proto_perl->Istashpadmax;
13623 PL_stashpadix = proto_perl->Istashpadix ;
13624 Newx(PL_stashpad, PL_stashpadmax, HV *);
13627 for (; o < PL_stashpadmax; ++o)
13628 PL_stashpad[o] = hv_dup(proto_perl->Istashpad[o], param);
13631 /* shortcuts to various I/O objects */
13632 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13633 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13634 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13635 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13636 PL_argvgv = gv_dup_inc(proto_perl->Iargvgv, param);
13637 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13638 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13640 /* shortcuts to regexp stuff */
13641 PL_replgv = gv_dup_inc(proto_perl->Ireplgv, param);
13643 /* shortcuts to misc objects */
13644 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13646 /* shortcuts to debugging objects */
13647 PL_DBgv = gv_dup_inc(proto_perl->IDBgv, param);
13648 PL_DBline = gv_dup_inc(proto_perl->IDBline, param);
13649 PL_DBsub = gv_dup_inc(proto_perl->IDBsub, param);
13650 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13651 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13652 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13654 /* symbol tables */
13655 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13656 PL_curstash = hv_dup_inc(proto_perl->Icurstash, param);
13657 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13658 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13659 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13661 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13662 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13663 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13664 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13665 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13666 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13667 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13668 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13670 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13672 /* subprocess state */
13673 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13675 if (proto_perl->Iop_mask)
13676 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13679 /* PL_asserting = proto_perl->Iasserting; */
13681 /* current interpreter roots */
13682 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13684 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13687 /* runtime control stuff */
13688 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13690 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13692 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13694 /* interpreter atexit processing */
13695 PL_exitlistlen = proto_perl->Iexitlistlen;
13696 if (PL_exitlistlen) {
13697 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13698 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13701 PL_exitlist = (PerlExitListEntry*)NULL;
13703 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13704 if (PL_my_cxt_size) {
13705 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13706 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13707 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13708 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13709 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13713 PL_my_cxt_list = (void**)NULL;
13714 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13715 PL_my_cxt_keys = (const char**)NULL;
13718 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13719 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13720 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13721 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13723 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13725 PAD_CLONE_VARS(proto_perl, param);
13727 #ifdef HAVE_INTERP_INTERN
13728 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13731 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13733 #ifdef PERL_USES_PL_PIDSTATUS
13734 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13736 PL_osname = SAVEPV(proto_perl->Iosname);
13737 PL_parser = parser_dup(proto_perl->Iparser, param);
13739 /* XXX this only works if the saved cop has already been cloned */
13740 if (proto_perl->Iparser) {
13741 PL_parser->saved_curcop = (COP*)any_dup(
13742 proto_perl->Iparser->saved_curcop,
13746 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13748 #ifdef USE_LOCALE_COLLATE
13749 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13750 #endif /* USE_LOCALE_COLLATE */
13752 #ifdef USE_LOCALE_NUMERIC
13753 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13754 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13755 #endif /* !USE_LOCALE_NUMERIC */
13757 /* Unicode inversion lists */
13758 PL_Latin1 = sv_dup_inc(proto_perl->ILatin1, param);
13759 PL_UpperLatin1 = sv_dup_inc(proto_perl->IUpperLatin1, param);
13760 PL_AboveLatin1 = sv_dup_inc(proto_perl->IAboveLatin1, param);
13762 PL_NonL1NonFinalFold = sv_dup_inc(proto_perl->INonL1NonFinalFold, param);
13763 PL_HasMultiCharFold= sv_dup_inc(proto_perl->IHasMultiCharFold, param);
13765 /* utf8 character class swashes */
13766 for (i = 0; i < POSIX_SWASH_COUNT; i++) {
13767 PL_utf8_swash_ptrs[i] = sv_dup_inc(proto_perl->Iutf8_swash_ptrs[i], param);
13769 for (i = 0; i < POSIX_CC_COUNT; i++) {
13770 PL_Posix_ptrs[i] = sv_dup_inc(proto_perl->IPosix_ptrs[i], param);
13771 PL_L1Posix_ptrs[i] = sv_dup_inc(proto_perl->IL1Posix_ptrs[i], param);
13772 PL_XPosix_ptrs[i] = sv_dup_inc(proto_perl->IXPosix_ptrs[i], param);
13774 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13775 PL_utf8_X_regular_begin = sv_dup_inc(proto_perl->Iutf8_X_regular_begin, param);
13776 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13777 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13778 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13779 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13780 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13781 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13782 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13783 PL_utf8_perl_idstart = sv_dup_inc(proto_perl->Iutf8_perl_idstart, param);
13784 PL_utf8_perl_idcont = sv_dup_inc(proto_perl->Iutf8_perl_idcont, param);
13785 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13786 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13787 PL_utf8_foldable = sv_dup_inc(proto_perl->Iutf8_foldable, param);
13788 PL_utf8_charname_begin = sv_dup_inc(proto_perl->Iutf8_charname_begin, param);
13789 PL_utf8_charname_continue = sv_dup_inc(proto_perl->Iutf8_charname_continue, param);
13791 if (proto_perl->Ipsig_pend) {
13792 Newxz(PL_psig_pend, SIG_SIZE, int);
13795 PL_psig_pend = (int*)NULL;
13798 if (proto_perl->Ipsig_name) {
13799 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13800 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13802 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13805 PL_psig_ptr = (SV**)NULL;
13806 PL_psig_name = (SV**)NULL;
13809 if (flags & CLONEf_COPY_STACKS) {
13810 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13811 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13812 PL_tmps_ix+1, param);
13814 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13815 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13816 Newxz(PL_markstack, i, I32);
13817 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13818 - proto_perl->Imarkstack);
13819 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13820 - proto_perl->Imarkstack);
13821 Copy(proto_perl->Imarkstack, PL_markstack,
13822 PL_markstack_ptr - PL_markstack + 1, I32);
13824 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13825 * NOTE: unlike the others! */
13826 Newxz(PL_scopestack, PL_scopestack_max, I32);
13827 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13830 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13831 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13833 /* reset stack AV to correct length before its duped via
13834 * PL_curstackinfo */
13835 AvFILLp(proto_perl->Icurstack) =
13836 proto_perl->Istack_sp - proto_perl->Istack_base;
13838 /* NOTE: si_dup() looks at PL_markstack */
13839 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13841 /* PL_curstack = PL_curstackinfo->si_stack; */
13842 PL_curstack = av_dup(proto_perl->Icurstack, param);
13843 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13845 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13846 PL_stack_base = AvARRAY(PL_curstack);
13847 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13848 - proto_perl->Istack_base);
13849 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13851 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13852 PL_savestack = ss_dup(proto_perl, param);
13856 ENTER; /* perl_destruct() wants to LEAVE; */
13859 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13860 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13862 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13863 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13864 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13865 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13866 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13867 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13869 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13871 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13872 PL_firstgv = gv_dup_inc(proto_perl->Ifirstgv, param);
13873 PL_secondgv = gv_dup_inc(proto_perl->Isecondgv, param);
13875 PL_stashcache = newHV();
13877 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13878 proto_perl->Iwatchaddr);
13879 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13880 if (PL_debug && PL_watchaddr) {
13881 PerlIO_printf(Perl_debug_log,
13882 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13883 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13884 PTR2UV(PL_watchok));
13887 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13888 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13889 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13891 /* Call the ->CLONE method, if it exists, for each of the stashes
13892 identified by sv_dup() above.
13894 while(av_len(param->stashes) != -1) {
13895 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13896 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13897 if (cloner && GvCV(cloner)) {
13902 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13904 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13910 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13911 ptr_table_free(PL_ptr_table);
13912 PL_ptr_table = NULL;
13915 if (!(flags & CLONEf_COPY_STACKS)) {
13916 unreferenced_to_tmp_stack(param->unreferenced);
13919 SvREFCNT_dec(param->stashes);
13921 /* orphaned? eg threads->new inside BEGIN or use */
13922 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13923 SvREFCNT_inc_simple_void(PL_compcv);
13924 SAVEFREESV(PL_compcv);
13931 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13933 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13935 if (AvFILLp(unreferenced) > -1) {
13936 SV **svp = AvARRAY(unreferenced);
13937 SV **const last = svp + AvFILLp(unreferenced);
13941 if (SvREFCNT(*svp) == 1)
13943 } while (++svp <= last);
13945 EXTEND_MORTAL(count);
13946 svp = AvARRAY(unreferenced);
13949 if (SvREFCNT(*svp) == 1) {
13950 /* Our reference is the only one to this SV. This means that
13951 in this thread, the scalar effectively has a 0 reference.
13952 That doesn't work (cleanup never happens), so donate our
13953 reference to it onto the save stack. */
13954 PL_tmps_stack[++PL_tmps_ix] = *svp;
13956 /* As an optimisation, because we are already walking the
13957 entire array, instead of above doing either
13958 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13959 release our reference to the scalar, so that at the end of
13960 the array owns zero references to the scalars it happens to
13961 point to. We are effectively converting the array from
13962 AvREAL() on to AvREAL() off. This saves the av_clear()
13963 (triggered by the SvREFCNT_dec(unreferenced) below) from
13964 walking the array a second time. */
13965 SvREFCNT_dec(*svp);
13968 } while (++svp <= last);
13969 AvREAL_off(unreferenced);
13971 SvREFCNT_dec_NN(unreferenced);
13975 Perl_clone_params_del(CLONE_PARAMS *param)
13977 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13979 PerlInterpreter *const to = param->new_perl;
13981 PerlInterpreter *const was = PERL_GET_THX;
13983 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13989 SvREFCNT_dec(param->stashes);
13990 if (param->unreferenced)
13991 unreferenced_to_tmp_stack(param->unreferenced);
14001 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
14004 /* Need to play this game, as newAV() can call safesysmalloc(), and that
14005 does a dTHX; to get the context from thread local storage.
14006 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
14007 a version that passes in my_perl. */
14008 PerlInterpreter *const was = PERL_GET_THX;
14009 CLONE_PARAMS *param;
14011 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
14017 /* Given that we've set the context, we can do this unshared. */
14018 Newx(param, 1, CLONE_PARAMS);
14021 param->proto_perl = from;
14022 param->new_perl = to;
14023 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
14024 AvREAL_off(param->stashes);
14025 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
14033 #endif /* USE_ITHREADS */
14036 Perl_init_constants(pTHX)
14038 SvREFCNT(&PL_sv_undef) = SvREFCNT_IMMORTAL;
14039 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
14040 SvANY(&PL_sv_undef) = NULL;
14042 SvANY(&PL_sv_no) = new_XPVNV();
14043 SvREFCNT(&PL_sv_no) = SvREFCNT_IMMORTAL;
14044 SvFLAGS(&PL_sv_no) = SVt_PVNV|SVf_READONLY
14045 |SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
14048 SvANY(&PL_sv_yes) = new_XPVNV();
14049 SvREFCNT(&PL_sv_yes) = SvREFCNT_IMMORTAL;
14050 SvFLAGS(&PL_sv_yes) = SVt_PVNV|SVf_READONLY
14051 |SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
14054 SvPV_set(&PL_sv_no, (char*)PL_No);
14055 SvCUR_set(&PL_sv_no, 0);
14056 SvLEN_set(&PL_sv_no, 0);
14057 SvIV_set(&PL_sv_no, 0);
14058 SvNV_set(&PL_sv_no, 0);
14060 SvPV_set(&PL_sv_yes, (char*)PL_Yes);
14061 SvCUR_set(&PL_sv_yes, 1);
14062 SvLEN_set(&PL_sv_yes, 0);
14063 SvIV_set(&PL_sv_yes, 1);
14064 SvNV_set(&PL_sv_yes, 1);
14068 =head1 Unicode Support
14070 =for apidoc sv_recode_to_utf8
14072 The encoding is assumed to be an Encode object, on entry the PV
14073 of the sv is assumed to be octets in that encoding, and the sv
14074 will be converted into Unicode (and UTF-8).
14076 If the sv already is UTF-8 (or if it is not POK), or if the encoding
14077 is not a reference, nothing is done to the sv. If the encoding is not
14078 an C<Encode::XS> Encoding object, bad things will happen.
14079 (See F<lib/encoding.pm> and L<Encode>.)
14081 The PV of the sv is returned.
14086 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
14090 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
14092 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
14101 if (SvPADTMP(nsv)) {
14102 nsv = sv_newmortal();
14103 SvSetSV_nosteal(nsv, sv);
14112 Passing sv_yes is wrong - it needs to be or'ed set of constants
14113 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
14114 remove converted chars from source.
14116 Both will default the value - let them.
14118 XPUSHs(&PL_sv_yes);
14121 call_method("decode", G_SCALAR);
14125 s = SvPV_const(uni, len);
14126 if (s != SvPVX_const(sv)) {
14127 SvGROW(sv, len + 1);
14128 Move(s, SvPVX(sv), len + 1, char);
14129 SvCUR_set(sv, len);
14134 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
14135 /* clear pos and any utf8 cache */
14136 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
14139 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
14140 magic_setutf8(sv,mg); /* clear UTF8 cache */
14145 return SvPOKp(sv) ? SvPVX(sv) : NULL;
14149 =for apidoc sv_cat_decode
14151 The encoding is assumed to be an Encode object, the PV of the ssv is
14152 assumed to be octets in that encoding and decoding the input starts
14153 from the position which (PV + *offset) pointed to. The dsv will be
14154 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
14155 when the string tstr appears in decoding output or the input ends on
14156 the PV of the ssv. The value which the offset points will be modified
14157 to the last input position on the ssv.
14159 Returns TRUE if the terminator was found, else returns FALSE.
14164 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
14165 SV *ssv, int *offset, char *tstr, int tlen)
14170 PERL_ARGS_ASSERT_SV_CAT_DECODE;
14172 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
14183 offsv = newSViv(*offset);
14185 mPUSHp(tstr, tlen);
14187 call_method("cat_decode", G_SCALAR);
14189 ret = SvTRUE(TOPs);
14190 *offset = SvIV(offsv);
14196 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
14201 /* ---------------------------------------------------------------------
14203 * support functions for report_uninit()
14206 /* the maxiumum size of array or hash where we will scan looking
14207 * for the undefined element that triggered the warning */
14209 #define FUV_MAX_SEARCH_SIZE 1000
14211 /* Look for an entry in the hash whose value has the same SV as val;
14212 * If so, return a mortal copy of the key. */
14215 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
14221 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
14223 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
14224 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
14227 array = HvARRAY(hv);
14229 for (i=HvMAX(hv); i>=0; i--) {
14231 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
14232 if (HeVAL(entry) != val)
14234 if ( HeVAL(entry) == &PL_sv_undef ||
14235 HeVAL(entry) == &PL_sv_placeholder)
14239 if (HeKLEN(entry) == HEf_SVKEY)
14240 return sv_mortalcopy(HeKEY_sv(entry));
14241 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
14247 /* Look for an entry in the array whose value has the same SV as val;
14248 * If so, return the index, otherwise return -1. */
14251 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
14255 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
14257 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
14258 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
14261 if (val != &PL_sv_undef) {
14262 SV ** const svp = AvARRAY(av);
14265 for (i=AvFILLp(av); i>=0; i--)
14272 /* varname(): return the name of a variable, optionally with a subscript.
14273 * If gv is non-zero, use the name of that global, along with gvtype (one
14274 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
14275 * targ. Depending on the value of the subscript_type flag, return:
14278 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
14279 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
14280 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
14281 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
14284 Perl_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
14285 const SV *const keyname, I32 aindex, int subscript_type)
14288 SV * const name = sv_newmortal();
14289 if (gv && isGV(gv)) {
14291 buffer[0] = gvtype;
14294 /* as gv_fullname4(), but add literal '^' for $^FOO names */
14296 gv_fullname4(name, gv, buffer, 0);
14298 if ((unsigned int)SvPVX(name)[1] <= 26) {
14300 buffer[1] = SvPVX(name)[1] + 'A' - 1;
14302 /* Swap the 1 unprintable control character for the 2 byte pretty
14303 version - ie substr($name, 1, 1) = $buffer; */
14304 sv_insert(name, 1, 1, buffer, 2);
14308 CV * const cv = gv ? ((CV *)gv) : find_runcv(NULL);
14312 assert(!cv || SvTYPE(cv) == SVt_PVCV || SvTYPE(cv) == SVt_PVFM);
14314 if (!cv || !CvPADLIST(cv))
14316 av = *PadlistARRAY(CvPADLIST(cv));
14317 sv = *av_fetch(av, targ, FALSE);
14318 sv_setsv_flags(name, sv, 0);
14321 if (subscript_type == FUV_SUBSCRIPT_HASH) {
14322 SV * const sv = newSV(0);
14323 *SvPVX(name) = '$';
14324 Perl_sv_catpvf(aTHX_ name, "{%s}",
14325 pv_pretty(sv, SvPVX_const(keyname), SvCUR(keyname), 32, NULL, NULL,
14326 PERL_PV_PRETTY_DUMP | PERL_PV_ESCAPE_UNI_DETECT ));
14327 SvREFCNT_dec_NN(sv);
14329 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
14330 *SvPVX(name) = '$';
14331 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
14333 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
14334 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
14335 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
14343 =for apidoc find_uninit_var
14345 Find the name of the undefined variable (if any) that caused the operator
14346 to issue a "Use of uninitialized value" warning.
14347 If match is true, only return a name if its value matches uninit_sv.
14348 So roughly speaking, if a unary operator (such as OP_COS) generates a
14349 warning, then following the direct child of the op may yield an
14350 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
14351 other hand, with OP_ADD there are two branches to follow, so we only print
14352 the variable name if we get an exact match.
14354 The name is returned as a mortal SV.
14356 Assumes that PL_op is the op that originally triggered the error, and that
14357 PL_comppad/PL_curpad points to the currently executing pad.
14363 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
14369 const OP *o, *o2, *kid;
14371 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
14372 uninit_sv == &PL_sv_placeholder)))
14375 switch (obase->op_type) {
14382 const bool pad = ( obase->op_type == OP_PADAV
14383 || obase->op_type == OP_PADHV
14384 || obase->op_type == OP_PADRANGE
14387 const bool hash = ( obase->op_type == OP_PADHV
14388 || obase->op_type == OP_RV2HV
14389 || (obase->op_type == OP_PADRANGE
14390 && SvTYPE(PAD_SVl(obase->op_targ)) == SVt_PVHV)
14394 int subscript_type = FUV_SUBSCRIPT_WITHIN;
14396 if (pad) { /* @lex, %lex */
14397 sv = PAD_SVl(obase->op_targ);
14401 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
14402 /* @global, %global */
14403 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
14406 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
14408 else if (obase == PL_op) /* @{expr}, %{expr} */
14409 return find_uninit_var(cUNOPx(obase)->op_first,
14411 else /* @{expr}, %{expr} as a sub-expression */
14415 /* attempt to find a match within the aggregate */
14417 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14419 subscript_type = FUV_SUBSCRIPT_HASH;
14422 index = find_array_subscript((const AV *)sv, uninit_sv);
14424 subscript_type = FUV_SUBSCRIPT_ARRAY;
14427 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
14430 return varname(gv, hash ? '%' : '@', obase->op_targ,
14431 keysv, index, subscript_type);
14435 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
14437 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
14438 if (!gv || !GvSTASH(gv))
14440 if (match && (GvSV(gv) != uninit_sv))
14442 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
14445 return find_uninit_var(cUNOPx(obase)->op_first, uninit_sv, 1);
14448 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
14450 return varname(NULL, '$', obase->op_targ,
14451 NULL, 0, FUV_SUBSCRIPT_NONE);
14454 gv = cGVOPx_gv(obase);
14455 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
14457 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
14459 case OP_AELEMFAST_LEX:
14462 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
14463 if (!av || SvRMAGICAL(av))
14465 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14466 if (!svp || *svp != uninit_sv)
14469 return varname(NULL, '$', obase->op_targ,
14470 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14473 gv = cGVOPx_gv(obase);
14478 AV *const av = GvAV(gv);
14479 if (!av || SvRMAGICAL(av))
14481 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14482 if (!svp || *svp != uninit_sv)
14485 return varname(gv, '$', 0,
14486 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14491 o = cUNOPx(obase)->op_first;
14492 if (!o || o->op_type != OP_NULL ||
14493 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
14495 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
14500 bool negate = FALSE;
14502 if (PL_op == obase)
14503 /* $a[uninit_expr] or $h{uninit_expr} */
14504 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
14507 o = cBINOPx(obase)->op_first;
14508 kid = cBINOPx(obase)->op_last;
14510 /* get the av or hv, and optionally the gv */
14512 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
14513 sv = PAD_SV(o->op_targ);
14515 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
14516 && cUNOPo->op_first->op_type == OP_GV)
14518 gv = cGVOPx_gv(cUNOPo->op_first);
14522 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
14527 if (kid && kid->op_type == OP_NEGATE) {
14529 kid = cUNOPx(kid)->op_first;
14532 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
14533 /* index is constant */
14536 kidsv = sv_2mortal(newSVpvs("-"));
14537 sv_catsv(kidsv, cSVOPx_sv(kid));
14540 kidsv = cSVOPx_sv(kid);
14544 if (obase->op_type == OP_HELEM) {
14545 HE* he = hv_fetch_ent(MUTABLE_HV(sv), kidsv, 0, 0);
14546 if (!he || HeVAL(he) != uninit_sv)
14550 SV * const opsv = cSVOPx_sv(kid);
14551 const IV opsviv = SvIV(opsv);
14552 SV * const * const svp = av_fetch(MUTABLE_AV(sv),
14553 negate ? - opsviv : opsviv,
14555 if (!svp || *svp != uninit_sv)
14559 if (obase->op_type == OP_HELEM)
14560 return varname(gv, '%', o->op_targ,
14561 kidsv, 0, FUV_SUBSCRIPT_HASH);
14563 return varname(gv, '@', o->op_targ, NULL,
14564 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14565 FUV_SUBSCRIPT_ARRAY);
14568 /* index is an expression;
14569 * attempt to find a match within the aggregate */
14570 if (obase->op_type == OP_HELEM) {
14571 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14573 return varname(gv, '%', o->op_targ,
14574 keysv, 0, FUV_SUBSCRIPT_HASH);
14578 = find_array_subscript((const AV *)sv, uninit_sv);
14580 return varname(gv, '@', o->op_targ,
14581 NULL, index, FUV_SUBSCRIPT_ARRAY);
14586 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
14588 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
14594 /* only examine RHS */
14595 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
14598 o = cUNOPx(obase)->op_first;
14599 if ( o->op_type == OP_PUSHMARK
14600 || (o->op_type == OP_NULL && o->op_targ == OP_PUSHMARK)
14604 if (!o->op_sibling) {
14605 /* one-arg version of open is highly magical */
14607 if (o->op_type == OP_GV) { /* open FOO; */
14609 if (match && GvSV(gv) != uninit_sv)
14611 return varname(gv, '$', 0,
14612 NULL, 0, FUV_SUBSCRIPT_NONE);
14614 /* other possibilities not handled are:
14615 * open $x; or open my $x; should return '${*$x}'
14616 * open expr; should return '$'.expr ideally
14622 /* ops where $_ may be an implicit arg */
14627 if ( !(obase->op_flags & OPf_STACKED)) {
14628 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14629 ? PAD_SVl(obase->op_targ)
14632 sv = sv_newmortal();
14633 sv_setpvs(sv, "$_");
14642 match = 1; /* print etc can return undef on defined args */
14643 /* skip filehandle as it can't produce 'undef' warning */
14644 o = cUNOPx(obase)->op_first;
14645 if ((obase->op_flags & OPf_STACKED)
14647 ( o->op_type == OP_PUSHMARK
14648 || (o->op_type == OP_NULL && o->op_targ == OP_PUSHMARK)))
14649 o = o->op_sibling->op_sibling;
14653 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14654 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14656 /* the following ops are capable of returning PL_sv_undef even for
14657 * defined arg(s) */
14676 case OP_GETPEERNAME:
14724 case OP_SMARTMATCH:
14733 /* XXX tmp hack: these two may call an XS sub, and currently
14734 XS subs don't have a SUB entry on the context stack, so CV and
14735 pad determination goes wrong, and BAD things happen. So, just
14736 don't try to determine the value under those circumstances.
14737 Need a better fix at dome point. DAPM 11/2007 */
14743 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14744 if (gv && GvSV(gv) == uninit_sv)
14745 return newSVpvs_flags("$.", SVs_TEMP);
14750 /* def-ness of rval pos() is independent of the def-ness of its arg */
14751 if ( !(obase->op_flags & OPf_MOD))
14756 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14757 return newSVpvs_flags("${$/}", SVs_TEMP);
14762 if (!(obase->op_flags & OPf_KIDS))
14764 o = cUNOPx(obase)->op_first;
14770 /* This loop checks all the kid ops, skipping any that cannot pos-
14771 * sibly be responsible for the uninitialized value; i.e., defined
14772 * constants and ops that return nothing. If there is only one op
14773 * left that is not skipped, then we *know* it is responsible for
14774 * the uninitialized value. If there is more than one op left, we
14775 * have to look for an exact match in the while() loop below.
14776 * Note that we skip padrange, because the individual pad ops that
14777 * it replaced are still in the tree, so we work on them instead.
14780 for (kid=o; kid; kid = kid->op_sibling) {
14782 const OPCODE type = kid->op_type;
14783 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14784 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14785 || (type == OP_PUSHMARK)
14786 || (type == OP_PADRANGE)
14790 if (o2) { /* more than one found */
14797 return find_uninit_var(o2, uninit_sv, match);
14799 /* scan all args */
14801 sv = find_uninit_var(o, uninit_sv, 1);
14813 =for apidoc report_uninit
14815 Print appropriate "Use of uninitialized variable" warning.
14821 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14825 SV* varname = NULL;
14826 if (uninit_sv && PL_curpad) {
14827 varname = find_uninit_var(PL_op, uninit_sv,0);
14829 sv_insert(varname, 0, 0, " ", 1);
14831 /* PL_warn_uninit_sv is constant */
14832 GCC_DIAG_IGNORE(-Wformat-nonliteral);
14833 /* diag_listed_as: Use of uninitialized value%s */
14834 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit_sv,
14835 SVfARG(varname ? varname : &PL_sv_no),
14836 " in ", OP_DESC(PL_op));
14840 /* PL_warn_uninit is constant */
14841 GCC_DIAG_IGNORE(-Wformat-nonliteral);
14842 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14850 * c-indentation-style: bsd
14851 * c-basic-offset: 4
14852 * indent-tabs-mode: nil
14855 * ex: set ts=8 sts=4 sw=4 et: