X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=sv.c;h=dcc7a899a97c92446249858455d4425c0d7fcf93;hb=712d05cf9bbf674929d5e38c5bb989d87c51f19b;hp=62f6107c073d5381fd815b126479f8d896b512e0;hpb=9f985e4c436303340569d267e23865fa9c16fba7;p=p5sagit%2Fp5-mst-13.2.git diff --git a/sv.c b/sv.c index 62f6107..dcc7a89 100644 --- a/sv.c +++ b/sv.c @@ -35,13 +35,13 @@ * lib/utf8.t lib/Unicode/Collate/t/index.t * --jhi */ -#define ASSERT_UTF8_CACHE(cache) \ +# define ASSERT_UTF8_CACHE(cache) \ STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \ assert((cache)[2] <= (cache)[3]); \ assert((cache)[3] <= (cache)[1]);} \ } STMT_END #else -#define ASSERT_UTF8_CACHE(cache) NOOP +# define ASSERT_UTF8_CACHE(cache) NOOP #endif #ifdef PERL_OLD_COPY_ON_WRITE @@ -190,10 +190,10 @@ Perl_offer_nice_chunk(pTHX_ void *chunk, U32 chunk_size) # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv) /* Whilst I'd love to do this, it seems that things like to check on unreferenced scalars -# define POSION_SV_HEAD(sv) Poison(sv, 1, struct STRUCT_SV) +# define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV) */ -# define POSION_SV_HEAD(sv) Poison(&SvANY(sv), 1, void *), \ - Poison(&SvREFCNT(sv), 1, U32) +# define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \ + PoisonNew(&SvREFCNT(sv), 1, U32) #else # define SvARENA_CHAIN(sv) SvANY(sv) # define POSION_SV_HEAD(sv) @@ -678,6 +678,7 @@ Perl_sv_free_arenas(pTHX) void* Perl_get_arena(pTHX_ int arena_size) { + dVAR; struct arena_desc* adesc; struct arena_set *newroot, **aroot = (struct arena_set**) &PL_body_arenas; int curr; @@ -692,7 +693,7 @@ Perl_get_arena(pTHX_ int arena_size) newroot->set_size = ARENAS_PER_SET; newroot->next = *aroot; *aroot = newroot; - DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", *aroot)); + DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)*aroot)); } /* ok, now have arena-set with at least 1 empty/available arena-desc */ @@ -1089,7 +1090,7 @@ S_more_bodies (pTHX_ svtype sv_type) void ** const r3wt = &PL_body_roots[sv_type]; \ LOCK_SV_MUTEX; \ xpv = *((void **)(r3wt)) \ - ? *((void **)(r3wt)) : S_more_bodies(aTHX_ sv_type); \ + ? *((void **)(r3wt)) : more_bodies(sv_type); \ *(r3wt) = *(void**)(xpv); \ UNLOCK_SV_MUTEX; \ } STMT_END @@ -1319,7 +1320,7 @@ Perl_sv_upgrade(pTHX_ register SV *sv, U32 new_type) int length = old_type_details->copy; if (new_type_details->offset > old_type_details->offset) { - int difference + const int difference = new_type_details->offset - old_type_details->offset; offset += difference; length -= difference; @@ -1717,8 +1718,29 @@ Perl_looks_like_number(pTHX_ SV *sv) return grok_number(sbegin, len, NULL); } +STATIC bool +S_glob_2number(pTHX_ GV * const gv) +{ + const U32 wasfake = SvFLAGS(gv) & SVf_FAKE; + SV *const buffer = sv_newmortal(); + + /* FAKE globs can get coerced, so need to turn this off temporarily if it + is on. */ + SvFAKE_off(gv); + gv_efullname3(buffer, gv, "*"); + SvFLAGS(gv) |= wasfake; + + /* We know that all GVs stringify to something that is not-a-number, + so no need to test that. */ + if (ckWARN(WARN_NUMERIC)) + not_a_number(buffer); + /* We just want something true to return, so that S_sv_2iuv_common + can tail call us and return true. */ + return TRUE; +} + STATIC char * -S_glob_2inpuv(pTHX_ GV *gv, STRLEN *len, bool want_number) +S_glob_2pv(pTHX_ GV * const gv, STRLEN * const len) { const U32 wasfake = SvFLAGS(gv) & SVf_FAKE; SV *const buffer = sv_newmortal(); @@ -1729,17 +1751,9 @@ S_glob_2inpuv(pTHX_ GV *gv, STRLEN *len, bool want_number) gv_efullname3(buffer, gv, "*"); SvFLAGS(gv) |= wasfake; - if (want_number) { - /* We know that all GVs stringify to something that is not-a-number, - so no need to test that. */ - if (ckWARN(WARN_NUMERIC)) - not_a_number(buffer); - /* We just want something true to return, so that S_sv_2iuv_common - can tail call us and return true. */ - return (char *) 1; - } else { - return SvPV(buffer, *len); - } + assert(SvPOK(buffer)); + *len = SvCUR(buffer); + return SvPVX(buffer); } /* Actually, ISO C leaves conversion of UV to IV undefined, but @@ -2050,7 +2064,7 @@ S_sv_2iuv_common(pTHX_ SV *sv) { if ((NV)(SvIVX(sv)) == SvNVX(sv)) { SvIOK_on(sv); } else { - /*EMPTY*/; /* Integer is imprecise. NOK, IOKp */ + NOOP; /* Integer is imprecise. NOK, IOKp */ } /* UV will not work better than IV */ } else { @@ -2065,7 +2079,7 @@ S_sv_2iuv_common(pTHX_ SV *sv) { if ((NV)(SvUVX(sv)) == SvNVX(sv)) { SvIOK_on(sv); } else { - /*EMPTY*/; /* Integer is imprecise. NOK, IOKp, is UV */ + NOOP; /* Integer is imprecise. NOK, IOKp, is UV */ } } SvIsUV_on(sv); @@ -2109,9 +2123,8 @@ S_sv_2iuv_common(pTHX_ SV *sv) { } } else { - if (isGV_with_GP(sv)) { - return (bool)PTR2IV(glob_2inpuv((GV *)sv, NULL, TRUE)); - } + if (isGV_with_GP(sv)) + return glob_2number((GV *)sv); if (!(SvFLAGS(sv) & SVs_PADTMP)) { if (!PL_localizing && ckWARN(WARN_UNINITIALIZED)) @@ -2461,7 +2474,7 @@ Perl_sv_2nv(pTHX_ register SV *sv) } else { if (isGV_with_GP(sv)) { - glob_2inpuv((GV *)sv, NULL, TRUE); + glob_2number((GV *)sv); return 0.0; } @@ -2644,8 +2657,14 @@ Perl_sv_2pv_flags(pTHX_ register SV *sv, STRLEN *lp, I32 flags) STRLEN len; if (SvIOKp(sv)) { - len = SvIsUV(sv) ? my_sprintf(tbuf,"%"UVuf, (UV)SvUVX(sv)) - : my_sprintf(tbuf,"%"IVdf, (IV)SvIVX(sv)); + len = SvIsUV(sv) +#ifdef USE_SNPRINTF + ? snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv)) + : snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv)); +#else + ? my_sprintf(tbuf, "%"UVuf, (UV)SvUVX(sv)) + : my_sprintf(tbuf, "%"IVdf, (IV)SvIVX(sv)); +#endif /* #ifdef USE_SNPRINTF */ } else { Gconvert(SvNVX(sv), NV_DIG, 0, tbuf); len = strlen(tbuf); @@ -2797,9 +2816,8 @@ Perl_sv_2pv_flags(pTHX_ register SV *sv, STRLEN *lp, I32 flags) #endif } else { - if (isGV_with_GP(sv)) { - return glob_2inpuv((GV *)sv, lp, FALSE); - } + if (isGV_with_GP(sv)) + return glob_2pv((GV *)sv, lp); if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED)) report_uninit(sv); @@ -3283,7 +3301,7 @@ S_glob_assign_ref(pTHX_ SV *dstr, SV *sstr) { it was a const and its value changed. */ if (CvCONST(cv) && CvCONST((CV*)sref) && cv_const_sv(cv) == cv_const_sv((CV*)sref)) { - /*EMPTY*/ + NOOP; /* They are 2 constant subroutines generated from the same constant. This probably means that they are really the "same" proxy subroutine @@ -3305,8 +3323,9 @@ S_glob_assign_ref(pTHX_ SV *dstr, SV *sstr) { } } if (!intro) - cv_ckproto(cv, (GV*)dstr, - SvPOK(sref) ? SvPVX_const(sref) : NULL); + cv_ckproto_len(cv, (GV*)dstr, + SvPOK(sref) ? SvPVX_const(sref) : NULL, + SvPOK(sref) ? SvCUR(sref) : 0); } GvCVGEN(dstr) = 0; /* Switch off cacheness. */ GvASSUMECV_on(dstr); @@ -3445,7 +3464,7 @@ Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV *sstr, I32 flags) case SVt_PVGV: if (dtype <= SVt_PVGV) { - S_glob_assign_glob(aTHX_ dstr, sstr, dtype); + glob_assign_glob(dstr, sstr, dtype); return; } /*FALLTHROUGH*/ @@ -3458,7 +3477,7 @@ Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV *sstr, I32 flags) if ((int)SvTYPE(sstr) != stype) { stype = SvTYPE(sstr); if (stype == SVt_PVGV && dtype <= SVt_PVGV) { - S_glob_assign_glob(aTHX_ dstr, sstr, dtype); + glob_assign_glob(dstr, sstr, dtype); return; } } @@ -3486,13 +3505,13 @@ Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV *sstr, I32 flags) GvMULTI_on(dstr); return; } - S_glob_assign_glob(aTHX_ dstr, sstr, dtype); + glob_assign_glob(dstr, sstr, dtype); return; } if (dtype >= SVt_PV) { if (dtype == SVt_PVGV) { - S_glob_assign_ref(aTHX_ dstr, sstr); + glob_assign_ref(dstr, sstr); return; } if (SvPVX_const(dstr)) { @@ -3881,21 +3900,27 @@ Perl_sv_setpv_mg(pTHX_ register SV *sv, register const char *ptr) } /* -=for apidoc sv_usepvn - -Tells an SV to use C to find its string value. Normally the string is -stored inside the SV but sv_usepvn allows the SV to use an outside string. -The C should point to memory that was allocated by C. The -string length, C, must be supplied. This function will realloc the -memory pointed to by C, so that pointer should not be freed or used by -the programmer after giving it to sv_usepvn. Does not handle 'set' magic. -See C. +=for apidoc sv_usepvn_flags + +Tells an SV to use C to find its string value. Normally the +string is stored inside the SV but sv_usepvn allows the SV to use an +outside string. The C should point to memory that was allocated +by C. The string length, C, must be supplied. By default +this function will realloc (i.e. move) the memory pointed to by C, +so that pointer should not be freed or used by the programmer after +giving it to sv_usepvn, and neither should any pointers from "behind" +that pointer (e.g. ptr + 1) be used. + +If C & SV_SMAGIC is true, will call SvSETMAGIC. If C & +SV_HAS_TRAILING_NUL is true, then C must be NUL, and the realloc +will be skipped. (i.e. the buffer is actually at least 1 byte longer than +C, and already meets the requirements for storing in C) =cut */ void -Perl_sv_usepvn(pTHX_ register SV *sv, register char *ptr, register STRLEN len) +Perl_sv_usepvn_flags(pTHX_ SV *sv, char *ptr, STRLEN len, U32 flags) { dVAR; STRLEN allocate; @@ -3903,34 +3928,43 @@ Perl_sv_usepvn(pTHX_ register SV *sv, register char *ptr, register STRLEN len) SvUPGRADE(sv, SVt_PV); if (!ptr) { (void)SvOK_off(sv); + if (flags & SV_SMAGIC) + SvSETMAGIC(sv); return; } if (SvPVX_const(sv)) SvPV_free(sv); - allocate = PERL_STRLEN_ROUNDUP(len + 1); - ptr = saferealloc (ptr, allocate); + if (flags & SV_HAS_TRAILING_NUL) + assert(ptr[len] == '\0'); + + allocate = (flags & SV_HAS_TRAILING_NUL) + ? len + 1: PERL_STRLEN_ROUNDUP(len + 1); + if (flags & SV_HAS_TRAILING_NUL) { + /* It's long enough - do nothing. + Specfically Perl_newCONSTSUB is relying on this. */ + } else { +#ifdef DEBUGGING + /* Force a move to shake out bugs in callers. */ + char *new_ptr = safemalloc(allocate); + Copy(ptr, new_ptr, len, char); + PoisonFree(ptr,len,char); + Safefree(ptr); + ptr = new_ptr; +#else + ptr = saferealloc (ptr, allocate); +#endif + } SvPV_set(sv, ptr); SvCUR_set(sv, len); SvLEN_set(sv, allocate); - *SvEND(sv) = '\0'; + if (!(flags & SV_HAS_TRAILING_NUL)) { + *SvEND(sv) = '\0'; + } (void)SvPOK_only_UTF8(sv); /* validate pointer */ SvTAINT(sv); -} - -/* -=for apidoc sv_usepvn_mg - -Like C, but also handles 'set' magic. - -=cut -*/ - -void -Perl_sv_usepvn_mg(pTHX_ register SV *sv, register char *ptr, register STRLEN len) -{ - sv_usepvn(sv,ptr,len); - SvSETMAGIC(sv); + if (flags & SV_SMAGIC) + SvSETMAGIC(sv); } #ifdef PERL_OLD_COPY_ON_WRITE @@ -4489,6 +4523,8 @@ Perl_sv_magic(pTHX_ register SV *sv, SV *obj, int how, const char *name, I32 nam case PERL_MAGIC_qr: vtable = &PL_vtbl_regexp; break; + case PERL_MAGIC_hints: + /* As this vtable is all NULL, we can reuse it. */ case PERL_MAGIC_sig: vtable = &PL_vtbl_sig; break; @@ -4528,6 +4564,9 @@ Perl_sv_magic(pTHX_ register SV *sv, SV *obj, int how, const char *name, I32 nam case PERL_MAGIC_backref: vtable = &PL_vtbl_backref; break; + case PERL_MAGIC_hintselem: + vtable = &PL_vtbl_hintselem; + break; case PERL_MAGIC_ext: /* Reserved for use by extensions not perl internals. */ /* Useful for attaching extension internal data to perl vars. */ @@ -5038,10 +5077,8 @@ Perl_sv_clear(pTHX_ register SV *sv) } } if (type >= SVt_PVMG) { - HV *ourstash; - if ((type == SVt_PVMG || type == SVt_PVGV) && - (ourstash = OURSTASH(sv))) { - SvREFCNT_dec(ourstash); + if ((type == SVt_PVMG || type == SVt_PVGV) && SvPAD_OUR(sv)) { + SvREFCNT_dec(OURSTASH(sv)); } else if (SvMAGIC(sv)) mg_free(sv); if (type == SVt_PVMG && SvPAD_TYPED(sv)) @@ -5274,8 +5311,10 @@ UTF-8 bytes as a single character. Handles magic and type coercion. /* * The length is cached in PERL_UTF8_magic, in the mg_len field. Also the - * mg_ptr is used, by sv_pos_u2b(), see the comments of S_utf8_mg_pos_init(). - * (Note that the mg_len is not the length of the mg_ptr field.) + * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below. + * (Note that the mg_len is not the length of the mg_ptr field. + * This allows the cache to store the character length of the string without + * needing to malloc() extra storage to attach to the mg_ptr.) * */ @@ -5308,7 +5347,7 @@ Perl_sv_len_utf8(pTHX_ register SV *sv) PL_utf8cache = 0; Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVf " real %"UVf" for %"SVf, - (UV) ulen, (UV) real, sv); + (UV) ulen, (UV) real, (void*)sv); } } } @@ -5329,27 +5368,10 @@ Perl_sv_len_utf8(pTHX_ register SV *sv) } } -/* -=for apidoc sv_pos_u2b - -Converts the value pointed to by offsetp from a count of UTF-8 chars from -the start of the string, to a count of the equivalent number of bytes; if -lenp is non-zero, it does the same to lenp, but this time starting from -the offset, rather than from the start of the string. Handles magic and -type coercion. - -=cut -*/ - -/* - * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential - * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and - * byte offsets. See also the comments of S_utf8_mg_pos(). - * - */ - +/* Walk forwards to find the byte corresponding to the passed in UTF-8 + offset. */ static STRLEN -S_sv_pos_u2b_forwards(pTHX_ const U8 *const start, const U8 *const send, +S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send, STRLEN uoffset) { const U8 *s = start; @@ -5364,9 +5386,11 @@ S_sv_pos_u2b_forwards(pTHX_ const U8 *const start, const U8 *const send, return s - start; } - +/* Given the length of the string in both bytes and UTF-8 characters, decide + whether to walk forwards or backwards to find the byte corresponding to + the passed in UTF-8 offset. */ static STRLEN -S_sv_pos_u2b_midway(pTHX_ const U8 *const start, const U8 *send, +S_sv_pos_u2b_midway(const U8 *const start, const U8 *send, STRLEN uoffset, STRLEN uend) { STRLEN backw = uend - uoffset; @@ -5374,7 +5398,7 @@ S_sv_pos_u2b_midway(pTHX_ const U8 *const start, const U8 *send, /* The assumption is that going forwards is twice the speed of going forward (that's where the 2 * backw comes from). (The real figure of course depends on the UTF-8 data.) */ - return S_sv_pos_u2b_forwards(aTHX_ start, send, uoffset); + return sv_pos_u2b_forwards(start, send, uoffset); } while (backw--) { @@ -5385,11 +5409,19 @@ S_sv_pos_u2b_midway(pTHX_ const U8 *const start, const U8 *send, return send - start; } +/* For the string representation of the given scalar, find the byte + corresponding to the passed in UTF-8 offset. uoffset0 and boffset0 + give another position in the string, *before* the sought offset, which + (which is always true, as 0, 0 is a valid pair of positions), which should + help reduce the amount of linear searching. + If *mgp is non-NULL, it should point to the UTF-8 cache magic, which + will be used to reduce the amount of linear searching. The cache will be + created if necessary, and the found value offered to it for update. */ static STRLEN S_sv_pos_u2b_cached(pTHX_ SV *sv, MAGIC **mgp, const U8 *const start, const U8 *const send, STRLEN uoffset, STRLEN uoffset0, STRLEN boffset0) { - STRLEN boffset; + STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */ bool found = FALSE; assert (uoffset >= uoffset0); @@ -5417,20 +5449,34 @@ S_sv_pos_u2b_cached(pTHX_ SV *sv, MAGIC **mgp, const U8 *const start, if ((*mgp)->mg_len != -1) { /* And we know the end too. */ boffset = boffset0 - + S_sv_pos_u2b_midway(aTHX_ start + boffset0, send, + + sv_pos_u2b_midway(start + boffset0, send, uoffset - uoffset0, (*mgp)->mg_len - uoffset0); } else { boffset = boffset0 - + S_sv_pos_u2b_forwards(aTHX_ start + boffset0, + + sv_pos_u2b_forwards(start + boffset0, send, uoffset - uoffset0); } - } else { + } + else if (cache[2] < uoffset) { + /* We're between the two cache entries. */ + if (cache[2] > uoffset0) { + /* and the cache knows more than the passed in pair */ + uoffset0 = cache[2]; + boffset0 = cache[3]; + } + boffset = boffset0 - + S_sv_pos_u2b_midway(aTHX_ start + boffset0, + + sv_pos_u2b_midway(start + boffset0, start + cache[1], uoffset - uoffset0, cache[0] - uoffset0); + } else { + boffset = boffset0 + + sv_pos_u2b_midway(start + boffset0, + start + cache[3], + uoffset - uoffset0, + cache[2] - uoffset0); } found = TRUE; } @@ -5439,7 +5485,7 @@ S_sv_pos_u2b_cached(pTHX_ SV *sv, MAGIC **mgp, const U8 *const start, /* In fact, offset0 is either 0, or less than offset, so don't need to worry about the other possibility. */ boffset = boffset0 - + S_sv_pos_u2b_midway(aTHX_ start + boffset0, send, + + sv_pos_u2b_midway(start + boffset0, send, uoffset - uoffset0, (*mgp)->mg_len - uoffset0); found = TRUE; @@ -5448,7 +5494,7 @@ S_sv_pos_u2b_cached(pTHX_ SV *sv, MAGIC **mgp, const U8 *const start, if (!found || PL_utf8cache < 0) { const STRLEN real_boffset - = boffset0 + S_sv_pos_u2b_forwards(aTHX_ start + boffset0, + = boffset0 + sv_pos_u2b_forwards(start + boffset0, send, uoffset - uoffset0); if (found && PL_utf8cache < 0) { @@ -5459,7 +5505,7 @@ S_sv_pos_u2b_cached(pTHX_ SV *sv, MAGIC **mgp, const U8 *const start, PL_utf8cache = 0; Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVf " real %"UVf" for %"SVf, - (UV) boffset, (UV) real_boffset, sv); + (UV) boffset, (UV) real_boffset, (void*)sv); } } boffset = real_boffset; @@ -5469,6 +5515,26 @@ S_sv_pos_u2b_cached(pTHX_ SV *sv, MAGIC **mgp, const U8 *const start, return boffset; } + +/* +=for apidoc sv_pos_u2b + +Converts the value pointed to by offsetp from a count of UTF-8 chars from +the start of the string, to a count of the equivalent number of bytes; if +lenp is non-zero, it does the same to lenp, but this time starting from +the offset, rather than from the start of the string. Handles magic and +type coercion. + +=cut +*/ + +/* + * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential + * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and + * byte offsets. See also the comments of S_utf8_mg_pos_cache_update(). + * + */ + void Perl_sv_pos_u2b(pTHX_ register SV *sv, I32* offsetp, I32* lenp) { @@ -5483,16 +5549,16 @@ Perl_sv_pos_u2b(pTHX_ register SV *sv, I32* offsetp, I32* lenp) STRLEN uoffset = (STRLEN) *offsetp; const U8 * const send = start + len; MAGIC *mg = NULL; - STRLEN boffset = S_sv_pos_u2b_cached(aTHX_ sv, &mg, start, send, + const STRLEN boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0); *offsetp = (I32) boffset; if (lenp) { /* Convert the relative offset to absolute. */ - STRLEN uoffset2 = uoffset + (STRLEN) *lenp; - STRLEN boffset2 - = S_sv_pos_u2b_cached(aTHX_ sv, &mg, start, send, uoffset2, + const STRLEN uoffset2 = uoffset + (STRLEN) *lenp; + const STRLEN boffset2 + = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2, uoffset, boffset) - boffset; *lenp = boffset2; @@ -5507,23 +5573,29 @@ Perl_sv_pos_u2b(pTHX_ register SV *sv, I32* offsetp, I32* lenp) return; } -/* -=for apidoc sv_pos_b2u - -Converts the value pointed to by offsetp from a count of bytes from the -start of the string, to a count of the equivalent number of UTF-8 chars. -Handles magic and type coercion. - -=cut +/* Create and update the UTF8 magic offset cache, with the proffered utf8/ + byte length pairing. The (byte) length of the total SV is passed in too, + as blen, because for some (more esoteric) SVs, the call to SvPV_const() + may not have updated SvCUR, so we can't rely on reading it directly. + + The proffered utf8/byte length pairing isn't used if the cache already has + two pairs, and swapping either for the proffered pair would increase the + RMS of the intervals between known byte offsets. + + The cache itself consists of 4 STRLEN values + 0: larger UTF-8 offset + 1: corresponding byte offset + 2: smaller UTF-8 offset + 3: corresponding byte offset + + Unused cache pairs have the value 0, 0. + Keeping the cache "backwards" means that the invariant of + cache[0] >= cache[2] is maintained even with empty slots, which means that + the code that uses it doesn't need to worry if only 1 entry has actually + been set to non-zero. It also makes the "position beyond the end of the + cache" logic much simpler, as the first slot is always the one to start + from. */ - -/* - * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential - * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and - * byte offsets. - * - */ - static void S_utf8_mg_pos_cache_update(pTHX_ SV *sv, MAGIC **mgp, STRLEN byte, STRLEN utf8, STRLEN blen) @@ -5566,7 +5638,7 @@ S_utf8_mg_pos_cache_update(pTHX_ SV *sv, MAGIC **mgp, STRLEN byte, STRLEN utf8, SAVEI8(PL_utf8cache); PL_utf8cache = 0; Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVf - " real %"UVf" for %"SVf, (UV) utf8, (UV) realutf8, sv); + " real %"UVf" for %"SVf, (UV) utf8, (UV) realutf8, (void*)sv); } } @@ -5716,15 +5788,32 @@ S_sv_pos_b2u_midway(pTHX_ const U8 *s, const U8 *const target, const U8 *end, return endu; } +/* +=for apidoc sv_pos_b2u + +Converts the value pointed to by offsetp from a count of bytes from the +start of the string, to a count of the equivalent number of UTF-8 chars. +Handles magic and type coercion. + +=cut +*/ + +/* + * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential + * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and + * byte offsets. + * + */ void Perl_sv_pos_b2u(pTHX_ register SV* sv, I32* offsetp) { const U8* s; const STRLEN byte = *offsetp; - STRLEN len; + STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */ STRLEN blen; MAGIC* mg = NULL; const U8* send; + bool found = FALSE; if (!sv) return; @@ -5739,7 +5828,7 @@ Perl_sv_pos_b2u(pTHX_ register SV* sv, I32* offsetp) if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache && (mg = mg_find(sv, PERL_MAGIC_utf8))) { if (mg->mg_ptr) { - STRLEN *cache = (STRLEN *) mg->mg_ptr; + STRLEN * const cache = (STRLEN *) mg->mg_ptr; if (cache[1] == byte) { /* An exact match. */ *offsetp = cache[0]; @@ -5779,27 +5868,27 @@ Perl_sv_pos_b2u(pTHX_ register SV* sv, I32* offsetp) } ASSERT_UTF8_CACHE(cache); - if (PL_utf8cache < 0) { - const STRLEN reallen = S_sv_pos_b2u_forwards(aTHX_ s, send); - - if (len != reallen) { - /* Need to turn the assertions off otherwise we may recurse - infinitely while printing error messages. */ - SAVEI8(PL_utf8cache); - PL_utf8cache = 0; - Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVf - " real %"UVf" for %"SVf, - (UV) len, (UV) reallen, sv); - } - } + found = TRUE; } else if (mg->mg_len != -1) { len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len); - } else { - len = S_sv_pos_b2u_forwards(aTHX_ s, send); + found = TRUE; } } - else { - len = S_sv_pos_b2u_forwards(aTHX_ s, send); + if (!found || PL_utf8cache < 0) { + const STRLEN real_len = S_sv_pos_b2u_forwards(aTHX_ s, send); + + if (found && PL_utf8cache < 0) { + if (len != real_len) { + /* Need to turn the assertions off otherwise we may recurse + infinitely while printing error messages. */ + SAVEI8(PL_utf8cache); + PL_utf8cache = 0; + Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVf + " real %"UVf" for %"SVf, + (UV) len, (UV) real_len, (void*)sv); + } + } + len = real_len; } *offsetp = len; @@ -6129,7 +6218,6 @@ Perl_sv_gets(pTHX_ register SV *sv, register PerlIO *fp, I32 append) register I32 cnt; I32 i = 0; I32 rspara = 0; - I32 recsize; if (SvTHINKFIRST(sv)) sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV); @@ -6170,9 +6258,9 @@ Perl_sv_gets(pTHX_ register SV *sv, register PerlIO *fp, I32 append) } else if (RsSNARF(PL_rs)) { /* If it is a regular disk file use size from stat() as estimate - of amount we are going to read - may result in malloc-ing - more memory than we realy need if layers bellow reduce - size we read (e.g. CRLF or a gzip layer) + of amount we are going to read -- may result in mallocing + more memory than we really need if the layers below reduce + the size we read (e.g. CRLF or a gzip layer). */ Stat_t st; if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) { @@ -6187,9 +6275,10 @@ Perl_sv_gets(pTHX_ register SV *sv, register PerlIO *fp, I32 append) else if (RsRECORD(PL_rs)) { I32 bytesread; char *buffer; + U32 recsize; /* Grab the size of the record we're getting */ - recsize = SvIV(SvRV(PL_rs)); + recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */ buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append; /* Go yank in */ #ifdef VMS @@ -6442,7 +6531,7 @@ screamer2: * * - jik 9/25/96 */ - if (!(cnt < sizeof(buf) && PerlIO_eof(fp))) + if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp))) goto screamer2; } @@ -6892,12 +6981,15 @@ Perl_newSVhek(pTHX_ const HEK *hek) SvUTF8_on (sv); Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */ return sv; - } else if (flags & HVhek_REHASH) { + } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) { /* We don't have a pointer to the hv, so we have to replicate the flag into every HEK. This hv is using custom a hasing algorithm. Hence we can't return a shared string scalar, as that would contain the (wrong) hash value, and might get passed - into an hv routine with a regular hash */ + into an hv routine with a regular hash. + Similarly, a hash that isn't using shared hash keys has to have + the flag in every key so that we know not to try to call + share_hek_kek on it. */ SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek)); if (HEK_UTF8(hek)) @@ -6905,9 +6997,23 @@ Perl_newSVhek(pTHX_ const HEK *hek) return sv; } /* This will be overwhelminly the most common case. */ - return newSVpvn_share(HEK_KEY(hek), - (HEK_UTF8(hek) ? -HEK_LEN(hek) : HEK_LEN(hek)), - HEK_HASH(hek)); + { + /* Inline most of newSVpvn_share(), because share_hek_hek() is far + more efficient than sharepvn(). */ + SV *sv; + + new_SV(sv); + sv_upgrade(sv, SVt_PV); + SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek))); + SvCUR_set(sv, HEK_LEN(hek)); + SvLEN_set(sv, 0); + SvREADONLY_on(sv); + SvFAKE_on(sv); + SvPOK_on(sv); + if (HEK_UTF8(hek)) + SvUTF8_on(sv); + return sv; + } } } @@ -6931,6 +7037,8 @@ Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash) dVAR; register SV *sv; bool is_utf8 = FALSE; + const char *const orig_src = src; + if (len < 0) { STRLEN tmplen = -len; is_utf8 = TRUE; @@ -6950,6 +7058,8 @@ Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash) SvPOK_on(sv); if (is_utf8) SvUTF8_on(sv); + if (src != orig_src) + Safefree(src); return sv; } @@ -7261,7 +7371,7 @@ Perl_sv_2io(pTHX_ SV *sv) else io = 0; if (!io) - Perl_croak(aTHX_ "Bad filehandle: %"SVf, sv); + Perl_croak(aTHX_ "Bad filehandle: %"SVf, (void*)sv); break; } return io; @@ -7353,7 +7463,7 @@ Perl_sv_2cv(pTHX_ SV *sv, HV **st, GV **gvp, I32 lref) LEAVE; if (!GvCVu(gv)) Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"", - sv); + (void*)sv); } return GvCVu(gv); } @@ -7629,9 +7739,11 @@ Perl_newSVrv(pTHX_ SV *rv, const char *classname) sv_clear(rv); SvFLAGS(rv) = 0; SvREFCNT(rv) = refcnt; - } - if (SvTYPE(rv) < SVt_RV) + sv_upgrade(rv, SVt_RV); + } else if (SvROK(rv)) { + SvREFCNT_dec(SvRV(rv)); + } else if (SvTYPE(rv) < SVt_RV) sv_upgrade(rv, SVt_RV); else if (SvTYPE(rv) > SVt_RV) { SvPV_free(rv); @@ -9156,8 +9268,13 @@ Perl_sv_vcatpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV * --jhi */ #if defined(HAS_LONG_DOUBLE) elen = ((intsize == 'q') +# ifdef USE_SNPRINTF + ? snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv) + : snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv)); +# else ? my_sprintf(PL_efloatbuf, ptr, nv) : my_sprintf(PL_efloatbuf, ptr, (double)nv)); +# endif /* #ifdef USE_SNPRINTF */ #else elen = my_sprintf(PL_efloatbuf, ptr, nv); #endif @@ -9208,7 +9325,7 @@ Perl_sv_vcatpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV (UV)c & 0xFF); } else sv_catpvs(msg, "end of string"); - Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, msg); /* yes, this is reentrant */ + Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, (void*)msg); /* yes, this is reentrant */ } /* output mangled stuff ... */ @@ -9323,11 +9440,15 @@ ptr_table_* functions. #if defined(USE_ITHREADS) +/* XXX Remove this so it doesn't have to go thru the macro and return for nothing */ #ifndef GpREFCNT_inc # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL) #endif +/* Certain cases in Perl_ss_dup have been merged, by relying on the fact + that currently av_dup and hv_dup are the same as sv_dup. If this changes, + please unmerge ss_dup. */ #define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t)) #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t)) #define av_dup(s,t) (AV*)sv_dup((SV*)s,t) @@ -9525,7 +9646,7 @@ Perl_gp_dup(pTHX_ GP *gp, CLONE_PARAMS* param) ret->gp_cv = cv_dup_inc(gp->gp_cv, param); ret->gp_cvgen = gp->gp_cvgen; ret->gp_line = gp->gp_line; - ret->gp_file = gp->gp_file; /* points to COP.cop_file */ + ret->gp_file_hek = hek_dup(gp->gp_file_hek, param); return ret; } @@ -9635,7 +9756,7 @@ S_ptr_table_find(PTR_TBL_t *tbl, const void *sv) { if (tblent->oldval == sv) return tblent; } - return 0; + return NULL; } void * @@ -9643,7 +9764,7 @@ Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *tbl, const void *sv) { PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv); PERL_UNUSED_CONTEXT; - return tblent ? tblent->newval : (void *) 0; + return tblent ? tblent->newval : NULL; } /* add a new entry to a pointer-mapping table */ @@ -9876,7 +9997,7 @@ Perl_sv_dup(pTHX_ const SV *sstr, CLONE_PARAMS* param) case SVt_PVGV: if (GvUNIQUE((GV*)sstr)) { - /*EMPTY*/; /* Do sharing here, and fall through */ + NOOP; /* Do sharing here, and fall through */ } case SVt_PVIO: case SVt_PVFM: @@ -9921,9 +10042,8 @@ Perl_sv_dup(pTHX_ const SV *sstr, CLONE_PARAMS* param) missing by always going for the destination. FIXME - instrument and check that assumption */ if (sv_type >= SVt_PVMG) { - HV *ourstash; - if ((sv_type == SVt_PVMG) && (ourstash = OURSTASH(dstr))) { - OURSTASH_set(dstr, hv_dup_inc(ourstash, param)); + if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) { + OURSTASH_set(dstr, hv_dup_inc(OURSTASH(dstr), param)); } else if (SvMAGIC(dstr)) SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param)); if (SvSTASH(dstr)) @@ -9987,7 +10107,7 @@ Perl_sv_dup(pTHX_ const SV *sstr, CLONE_PARAMS* param) if (IoDIRP(dstr)) { IoDIRP(dstr) = dirp_dup(IoDIRP(dstr)); } else { - /*EMPTY*/; + NOOP; /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */ } } @@ -10309,23 +10429,12 @@ Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param) TOPINT(nss,ix) = i; switch (i) { case SAVEt_ITEM: /* normal string */ + case SAVEt_SV: /* scalar reference */ sv = (SV*)POPPTR(ss,ix); TOPPTR(nss,ix) = sv_dup_inc(sv, param); sv = (SV*)POPPTR(ss,ix); TOPPTR(nss,ix) = sv_dup_inc(sv, param); break; - case SAVEt_SV: /* scalar reference */ - sv = (SV*)POPPTR(ss,ix); - TOPPTR(nss,ix) = sv_dup_inc(sv, param); - gv = (GV*)POPPTR(ss,ix); - TOPPTR(nss,ix) = gv_dup_inc(gv, param); - break; - case SAVEt_GENERIC_PVREF: /* generic char* */ - c = (char*)POPPTR(ss,ix); - TOPPTR(nss,ix) = pv_dup(c); - ptr = POPPTR(ss,ix); - TOPPTR(nss,ix) = any_dup(ptr, proto_perl); - break; case SAVEt_SHARED_PVREF: /* char* in shared space */ c = (char*)POPPTR(ss,ix); TOPPTR(nss,ix) = savesharedpv(c); @@ -10339,15 +10448,10 @@ Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param) ptr = POPPTR(ss,ix); TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */ break; - case SAVEt_AV: /* array reference */ - av = (AV*)POPPTR(ss,ix); - TOPPTR(nss,ix) = av_dup_inc(av, param); - gv = (GV*)POPPTR(ss,ix); - TOPPTR(nss,ix) = gv_dup(gv, param); - break; case SAVEt_HV: /* hash reference */ - hv = (HV*)POPPTR(ss,ix); - TOPPTR(nss,ix) = hv_dup_inc(hv, param); + case SAVEt_AV: /* array reference */ + sv = POPPTR(ss,ix); + TOPPTR(nss,ix) = sv_dup_inc(sv, param); gv = (GV*)POPPTR(ss,ix); TOPPTR(nss,ix) = gv_dup(gv, param); break; @@ -10366,6 +10470,7 @@ Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param) case SAVEt_I32: /* I32 reference */ case SAVEt_I16: /* I16 reference */ case SAVEt_I8: /* I8 reference */ + case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */ ptr = POPPTR(ss,ix); TOPPTR(nss,ix) = any_dup(ptr, proto_perl); i = POPINT(ss,ix); @@ -10377,6 +10482,8 @@ Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param) iv = POPIV(ss,ix); TOPIV(nss,ix) = iv; break; + case SAVEt_HPTR: /* HV* reference */ + case SAVEt_APTR: /* AV* reference */ case SAVEt_SPTR: /* SV* reference */ ptr = POPPTR(ss,ix); TOPPTR(nss,ix) = any_dup(ptr, proto_perl); @@ -10389,24 +10496,13 @@ Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param) ptr = POPPTR(ss,ix); TOPPTR(nss,ix) = any_dup(ptr, proto_perl); break; + case SAVEt_GENERIC_PVREF: /* generic char* */ case SAVEt_PPTR: /* char* reference */ ptr = POPPTR(ss,ix); TOPPTR(nss,ix) = any_dup(ptr, proto_perl); c = (char*)POPPTR(ss,ix); TOPPTR(nss,ix) = pv_dup(c); break; - case SAVEt_HPTR: /* HV* reference */ - ptr = POPPTR(ss,ix); - TOPPTR(nss,ix) = any_dup(ptr, proto_perl); - hv = (HV*)POPPTR(ss,ix); - TOPPTR(nss,ix) = hv_dup(hv, param); - break; - case SAVEt_APTR: /* AV* reference */ - ptr = POPPTR(ss,ix); - TOPPTR(nss,ix) = any_dup(ptr, proto_perl); - av = (AV*)POPPTR(ss,ix); - TOPPTR(nss,ix) = av_dup(av, param); - break; case SAVEt_NSTAB: gv = (GV*)POPPTR(ss,ix); TOPPTR(nss,ix) = gv_dup(gv, param); @@ -10518,6 +10614,17 @@ Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param) case SAVEt_HINTS: i = POPINT(ss,ix); TOPINT(nss,ix) = i; + ptr = POPPTR(ss,ix); + if (ptr) { + HINTS_REFCNT_LOCK; + ((struct refcounted_he *)ptr)->refcounted_he_refcnt++; + HINTS_REFCNT_UNLOCK; + } + TOPPTR(nss,ix) = ptr; + if (i & HINT_LOCALIZE_HH) { + hv = (HV*)POPPTR(ss,ix); + TOPPTR(nss,ix) = hv_dup_inc(hv, param); + } break; case SAVEt_COMPPAD: av = (AV*)POPPTR(ss,ix); @@ -10545,8 +10652,70 @@ Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param) sv = (SV*)POPPTR(ss,ix); TOPPTR(nss,ix) = sv_dup(sv, param); break; + case SAVEt_RE_STATE: + { + const struct re_save_state *const old_state + = (struct re_save_state *) + (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE); + struct re_save_state *const new_state + = (struct re_save_state *) + (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE); + + Copy(old_state, new_state, 1, struct re_save_state); + ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE; + + new_state->re_state_bostr + = pv_dup(old_state->re_state_bostr); + new_state->re_state_reginput + = pv_dup(old_state->re_state_reginput); + new_state->re_state_regeol + = pv_dup(old_state->re_state_regeol); + new_state->re_state_regstartp + = any_dup(old_state->re_state_regstartp, proto_perl); + new_state->re_state_regendp + = any_dup(old_state->re_state_regendp, proto_perl); + new_state->re_state_reglastparen + = any_dup(old_state->re_state_reglastparen, proto_perl); + new_state->re_state_reglastcloseparen + = any_dup(old_state->re_state_reglastcloseparen, + proto_perl); + /* XXX This just has to be broken. The old save_re_context + code did SAVEGENERICPV(PL_reg_start_tmp); + PL_reg_start_tmp is char **. + Look above to what the dup code does for + SAVEt_GENERIC_PVREF + It can never have worked. + So this is merely a faithful copy of the exiting bug: */ + new_state->re_state_reg_start_tmp + = (char **) pv_dup((char *) + old_state->re_state_reg_start_tmp); + /* I assume that it only ever "worked" because no-one called + (pseudo)fork while the regexp engine had re-entered itself. + */ +#ifdef PERL_OLD_COPY_ON_WRITE + new_state->re_state_nrs + = sv_dup(old_state->re_state_nrs, param); +#endif + new_state->re_state_reg_magic + = any_dup(old_state->re_state_reg_magic, proto_perl); + new_state->re_state_reg_oldcurpm + = any_dup(old_state->re_state_reg_oldcurpm, proto_perl); + new_state->re_state_reg_curpm + = any_dup(old_state->re_state_reg_curpm, proto_perl); + new_state->re_state_reg_oldsaved + = pv_dup(old_state->re_state_reg_oldsaved); + new_state->re_state_reg_poscache + = pv_dup(old_state->re_state_reg_poscache); + new_state->re_state_reg_starttry + = pv_dup(old_state->re_state_reg_starttry); + break; + } + case SAVEt_COMPILE_WARNINGS: + ptr = POPPTR(ss,ix); + TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr); + break; default: - Perl_croak(aTHX_ "panic: ss_dup inconsistency"); + Perl_croak(aTHX_ "panic: ss_dup inconsistency (%"IVdf")", (IV) i); } } @@ -10675,7 +10844,7 @@ perl_clone_using(PerlInterpreter *proto_perl, UV flags, PERL_SET_THX(my_perl); # ifdef DEBUGGING - Poison(my_perl, 1, PerlInterpreter); + PoisonNew(my_perl, 1, PerlInterpreter); PL_op = NULL; PL_curcop = NULL; PL_markstack = 0; @@ -10709,7 +10878,7 @@ perl_clone_using(PerlInterpreter *proto_perl, UV flags, PERL_SET_THX(my_perl); # ifdef DEBUGGING - Poison(my_perl, 1, PerlInterpreter); + PoisonNew(my_perl, 1, PerlInterpreter); PL_op = NULL; PL_curcop = NULL; PL_markstack = 0; @@ -10764,7 +10933,7 @@ perl_clone_using(PerlInterpreter *proto_perl, UV flags, SvREFCNT(&PL_sv_no) = (~(U32)0)/2; SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV; - SvPV_set(&PL_sv_no, SAVEPVN(PL_No, 0)); + SvPV_set(&PL_sv_no, savepvn(PL_No, 0)); SvCUR_set(&PL_sv_no, 0); SvLEN_set(&PL_sv_no, 1); SvIV_set(&PL_sv_no, 0); @@ -10775,7 +10944,7 @@ perl_clone_using(PerlInterpreter *proto_perl, UV flags, SvREFCNT(&PL_sv_yes) = (~(U32)0)/2; SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV; - SvPV_set(&PL_sv_yes, SAVEPVN(PL_Yes, 1)); + SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1)); SvCUR_set(&PL_sv_yes, 1); SvLEN_set(&PL_sv_yes, 2); SvIV_set(&PL_sv_yes, 1); @@ -10798,10 +10967,14 @@ perl_clone_using(PerlInterpreter *proto_perl, UV flags, ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file); ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling); - if (!specialWARN(PL_compiling.cop_warnings)) - PL_compiling.cop_warnings = sv_dup_inc(PL_compiling.cop_warnings, param); + PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings); if (!specialCopIO(PL_compiling.cop_io)) PL_compiling.cop_io = sv_dup_inc(PL_compiling.cop_io, param); + if (PL_compiling.cop_hints) { + HINTS_REFCNT_LOCK; + PL_compiling.cop_hints->refcounted_he_refcnt++; + HINTS_REFCNT_UNLOCK; + } PL_curcop = (COP*)any_dup(proto_perl->Tcurcop, proto_perl); /* pseudo environmental stuff */ @@ -10855,7 +11028,6 @@ perl_clone_using(PerlInterpreter *proto_perl, UV flags, PL_formfeed = sv_dup(proto_perl->Iformfeed, param); PL_maxsysfd = proto_perl->Imaxsysfd; - PL_multiline = proto_perl->Imultiline; PL_statusvalue = proto_perl->Istatusvalue; #ifdef VMS PL_statusvalue_vms = proto_perl->Istatusvalue_vms; @@ -11363,47 +11535,8 @@ perl_clone_using(PerlInterpreter *proto_perl, UV flags, PL_watchok = NULL; PL_regdummy = proto_perl->Tregdummy; - PL_regprecomp = NULL; - PL_regnpar = 0; - PL_regsize = 0; PL_colorset = 0; /* reinits PL_colors[] */ /*PL_colors[6] = {0,0,0,0,0,0};*/ - PL_reginput = NULL; - PL_regbol = NULL; - PL_regeol = NULL; - PL_regstartp = (I32*)NULL; - PL_regendp = (I32*)NULL; - PL_reglastparen = (U32*)NULL; - PL_reglastcloseparen = (U32*)NULL; - PL_regtill = NULL; - PL_reg_start_tmp = (char**)NULL; - PL_reg_start_tmpl = 0; - PL_regdata = (struct reg_data*)NULL; - PL_bostr = NULL; - PL_reg_flags = 0; - PL_reg_eval_set = 0; - PL_regnarrate = 0; - PL_regprogram = (regnode*)NULL; - PL_regindent = 0; - PL_regcc = (CURCUR*)NULL; - PL_reg_call_cc = (struct re_cc_state*)NULL; - PL_reg_re = (regexp*)NULL; - PL_reg_ganch = NULL; - PL_reg_sv = NULL; - PL_reg_match_utf8 = FALSE; - PL_reg_magic = (MAGIC*)NULL; - PL_reg_oldpos = 0; - PL_reg_oldcurpm = (PMOP*)NULL; - PL_reg_curpm = (PMOP*)NULL; - PL_reg_oldsaved = NULL; - PL_reg_oldsavedlen = 0; -#ifdef PERL_OLD_COPY_ON_WRITE - PL_nrs = NULL; -#endif - PL_reg_maxiter = 0; - PL_reg_leftiter = 0; - PL_reg_poscache = NULL; - PL_reg_poscache_size= 0; /* RE engine - function pointers */ PL_regcompp = proto_perl->Tregcompp; @@ -11411,9 +11544,9 @@ perl_clone_using(PerlInterpreter *proto_perl, UV flags, PL_regint_start = proto_perl->Tregint_start; PL_regint_string = proto_perl->Tregint_string; PL_regfree = proto_perl->Tregfree; - + Zero(&PL_reg_state, 1, struct re_save_state); PL_reginterp_cnt = 0; - PL_reg_starttry = 0; + PL_regmatch_slab = NULL; /* Pluggable optimizer */ PL_peepp = proto_perl->Tpeepp; @@ -11620,16 +11753,17 @@ STATIC I32 S_find_array_subscript(pTHX_ AV *av, SV* val) { dVAR; - SV** svp; - I32 i; if (!av || SvMAGICAL(av) || !AvARRAY(av) || (AvFILLp(av) > FUV_MAX_SEARCH_SIZE)) return -1; - svp = AvARRAY(av); - for (i=AvFILLp(av); i>=0; i--) { - if (svp[i] == val && svp[i] != &PL_sv_undef) - return i; + if (val != &PL_sv_undef) { + SV ** const svp = AvARRAY(av); + I32 i; + + for (i=AvFILLp(av); i>=0; i--) + if (svp[i] == val) + return i; } return -1; } @@ -11766,12 +11900,12 @@ S_find_uninit_var(pTHX_ OP* obase, SV* uninit_sv, bool match) /* attempt to find a match within the aggregate */ if (hash) { - keysv = S_find_hash_subscript(aTHX_ (HV*)sv, uninit_sv); + keysv = find_hash_subscript((HV*)sv, uninit_sv); if (keysv) subscript_type = FUV_SUBSCRIPT_HASH; } else { - index = S_find_array_subscript(aTHX_ (AV*)sv, uninit_sv); + index = find_array_subscript((AV*)sv, uninit_sv); if (index >= 0) subscript_type = FUV_SUBSCRIPT_ARRAY; } @@ -11887,13 +12021,13 @@ S_find_uninit_var(pTHX_ OP* obase, SV* uninit_sv, bool match) /* index is an expression; * attempt to find a match within the aggregate */ if (obase->op_type == OP_HELEM) { - SV * const keysv = S_find_hash_subscript(aTHX_ (HV*)sv, uninit_sv); + SV * const keysv = find_hash_subscript((HV*)sv, uninit_sv); if (keysv) return varname(gv, '%', o->op_targ, keysv, 0, FUV_SUBSCRIPT_HASH); } else { - const I32 index = S_find_array_subscript(aTHX_ (AV*)sv, uninit_sv); + const I32 index = find_array_subscript((AV*)sv, uninit_sv); if (index >= 0) return varname(gv, '@', o->op_targ, NULL, index, FUV_SUBSCRIPT_ARRAY); @@ -11985,13 +12119,14 @@ S_find_uninit_var(pTHX_ OP* obase, SV* uninit_sv, bool match) * or are optimized away, then it's unambiguous */ o2 = NULL; for (kid=o; kid; kid = kid->op_sibling) { - if (kid && - ( (kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) - || (kid->op_type == OP_NULL && ! (kid->op_flags & OPf_KIDS)) - || (kid->op_type == OP_PUSHMARK) + if (kid) { + const OPCODE type = kid->op_type; + if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid))) + || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS)) + || (type == OP_PUSHMARK) ) - ) continue; + } if (o2) { /* more than one found */ o2 = NULL; break;