Jumbo regexp patch applied (with minor fix-up tweaks):

	Subject: Version 7 of Jumbo RE patch available

p4raw-id: //depot/perl@267

Author: Ilya Zakharevich

dump.c

@@ -359,18 +359,17 @@ dump_pm(PMOP *pm)
 	dump("PMf_REPL = ");
 	dump_op(pm->op_pmreplroot);
     }
-    if (pm->op_pmshort) {
-	dump("PMf_SHORT = %s\n",SvPEEK(pm->op_pmshort));
-    }
-    if (pm->op_pmflags) {
+    if (pm->op_pmflags || (pm->op_pmregexp && pm->op_pmregexp->check_substr)) {
 	SV *tmpsv = newSVpv("", 0);
 	if (pm->op_pmflags & PMf_USED)
 	    sv_catpv(tmpsv, ",USED");
 	if (pm->op_pmflags & PMf_ONCE)
 	    sv_catpv(tmpsv, ",ONCE");
-	if (pm->op_pmflags & PMf_SCANFIRST)
+	if (pm->op_pmregexp && pm->op_pmregexp->check_substr
+	    && !(pm->op_pmregexp->reganch & ROPT_NOSCAN))
 	    sv_catpv(tmpsv, ",SCANFIRST");
-	if (pm->op_pmflags & PMf_ALL)
+	if (pm->op_pmregexp && pm->op_pmregexp->check_substr
+	    && pm->op_pmregexp->reganch & ROPT_CHECK_ALL)
 	    sv_catpv(tmpsv, ",ALL");
 	if (pm->op_pmflags & PMf_SKIPWHITE)
 	    sv_catpv(tmpsv, ",SKIPWHITE");

embed.h

@@ -185,7 +185,6 @@
 #define do_trans		Perl_do_trans
 #define do_vecset		Perl_do_vecset
 #define do_vop			Perl_do_vop
-#define doeval			Perl_doeval
 #define dofindlabel		Perl_dofindlabel
 #define dopoptoeval		Perl_dopoptoeval
 #define dounwind		Perl_dounwind
@@ -331,6 +330,7 @@
 #define magic_clearsig		Perl_magic_clearsig
 #define magic_existspack	Perl_magic_existspack
 #define magic_freedefelem	Perl_magic_freedefelem
+#define magic_freeregexp	Perl_magic_freeregexp
 #define magic_get		Perl_magic_get
 #define magic_getarylen		Perl_magic_getarylen
 #define magic_getdefelem	Perl_magic_getdefelem
@@ -890,32 +890,11 @@
 #define reall_srchlen		Perl_reall_srchlen
 #define ref			Perl_ref
 #define refkids			Perl_refkids
-#define regarglen		Perl_regarglen
-#define regbol			Perl_regbol
-#define regcode			Perl_regcode
-#define regdummy		Perl_regdummy
 #define regdump			Perl_regdump
-#define regendp			Perl_regendp
-#define regeol			Perl_regeol
-#define reginput		Perl_reginput
+#define regexec_flags		Perl_regexec_flags
 #define regkind			Perl_regkind
-#define reglastparen		Perl_reglastparen
-#define regmyendp		Perl_regmyendp
-#define regmyp_size		Perl_regmyp_size
-#define regmystartp		Perl_regmystartp
-#define regnarrate		Perl_regnarrate
-#define regnaughty		Perl_regnaughty
 #define regnext			Perl_regnext
-#define regnpar			Perl_regnpar
-#define regparse		Perl_regparse
-#define regprecomp		Perl_regprecomp
-#define regprev			Perl_regprev
 #define regprop			Perl_regprop
-#define regsawback		Perl_regsawback
-#define regsize			Perl_regsize
-#define regstartp		Perl_regstartp
-#define regtill			Perl_regtill
-#define regxend			Perl_regxend
 #define repeat_amg		Perl_repeat_amg
 #define repeat_ass_amg		Perl_repeat_ass_amg
 #define repeatcpy		Perl_repeatcpy
@@ -1046,6 +1025,7 @@
 #define sv_cmp			Perl_sv_cmp
 #define sv_cmp_locale		Perl_sv_cmp_locale
 #define sv_collxfrm		Perl_sv_collxfrm
+#define sv_compile_2op		Perl_sv_compile_2op
 #define sv_dec			Perl_sv_dec
 #define sv_derived_from		Perl_sv_derived_from
 #define sv_dump			Perl_sv_dump
@@ -1197,7 +1177,6 @@
 #define no_wrongref		Perl_no_wrongref
 #define pad_reset_pending	Perl_pad_reset_pending
 #define padix_floor		Perl_padix_floor
-#define regflags		Perl_regflags
 #define safecalloc		Perl_safecalloc
 #define safefree		Perl_safefree
 #define safemalloc		Perl_safemalloc

global.sym

@@ -173,30 +173,11 @@ psig_name
 psig_ptr
 rcsid
 reall_srchlen
-regarglen
-regbol
-regcode
-regdummy
-regendp
-regeol
-regflags
-reginput
+regdump
+regexec_flags
 regkind
-reglastparen
-regmyendp
-regmyp_size
-regmystartp
-regnarrate
-regnaughty
-regnpar
-regparse
-regprecomp
-regprev
-regsawback
-regsize
-regstartp
-regtill
-regxend
+regnext
+regprop
 repeat_amg
 repeat_ass_amg
 retstack
@@ -429,7 +410,6 @@ do_tell
 do_trans
 do_vecset
 do_vop
-doeval
 dofindlabel
 dopoptoeval
 dounwind
@@ -526,6 +506,7 @@ magic_clearpack
 magic_clearsig
 magic_existspack
 magic_freedefelem
+magic_freeregexp
 magic_get
 magic_getarylen
 magic_getdefelem
@@ -1021,6 +1002,7 @@ q
 ref
 refkids
 regdump
+regexec_flags
 regnext
 regprop
 repeatcpy
@@ -1120,6 +1102,7 @@ sv_clean_objs
 sv_clear
 sv_cmp
 sv_cmp_locale
+sv_compile_2op
 sv_collxfrm
 sv_dec
 sv_derived_from

mg.c

@@ -418,7 +418,7 @@ magic_get(SV *sv, MAGIC *mg)
 		    }
 		    sv_setpvn(sv,s,i);
 		    if (tainting)
-			tainted = was_tainted || rx->exec_tainted;
+			tainted = was_tainted || RX_MATCH_TAINTED(rx);
 		    break;
 		}
 	    }
@@ -1305,6 +1305,14 @@ magic_setuvar(SV *sv, MAGIC *mg)
     return 0;
 }
 
+int
+magic_freeregexp(SV *sv, MAGIC *mg)
+{
+    regexp *re = (regexp *)mg->mg_obj;
+    ReREFCNT_dec(re);
+    return 0;
+}
+
 #ifdef USE_LOCALE_COLLATE
 int
 magic_setcollxfrm(SV *sv, MAGIC *mg)

op.c

@@ -522,6 +522,12 @@ find_thread_magical(char *name)
 	case ';':
 	    sv_setpv(sv, "\034");
 	    break;
+	case '&':
+	case '`':
+	case '\'':
+	    sawampersand = TRUE;
+	    SvREADONLY_on(sv);
+	    break;
 	}
 	sv_magic(sv, 0, 0, name, 1); 
 	DEBUG_L(PerlIO_printf(PerlIO_stderr(),
@@ -594,8 +600,7 @@ op_free(OP *o)
 	/* FALL THROUGH */
     case OP_PUSHRE:
     case OP_MATCH:
-	pregfree(cPMOPo->op_pmregexp);
-	SvREFCNT_dec(cPMOPo->op_pmshort);
+	ReREFCNT_dec(cPMOPo->op_pmregexp);
 	break;
     }
 
@@ -1914,7 +1919,12 @@ newUNOP(I32 type, I32 flags, OP *first)
     unop->op_first = first;
     unop->op_flags = flags | OPf_KIDS;
     unop->op_private = 1 | (flags >> 8);
-
+#if 1
+    if(type == OP_STUDY && first->op_type == OP_MATCH) {
+	first->op_type = OP_PUSHRE;
+	first->op_ppaddr = ppaddr[OP_PUSHRE];
+    }
+#endif
     unop = (UNOP*) CHECKOP(type, unop);
     if (unop->op_next)
 	return (OP*)unop;
@@ -2065,7 +2075,6 @@ pmruntime(OP *o, OP *expr, OP *repl)
 	pm->op_pmregexp = pregcomp(p, p + plen, pm);
 	if (strEQ("\\s+", pm->op_pmregexp->precomp))
 	    pm->op_pmflags |= PMf_WHITE;
-	hoistmust(pm);
 	op_free(expr);
     }
     else {
@@ -4446,7 +4455,6 @@ OP *
 ck_split(OP *o)
 {
     register OP *kid;
-    PMOP* pm;
 
     if (o->op_flags & OPf_STACKED)
 	return no_fh_allowed(o);
@@ -4471,11 +4479,6 @@ ck_split(OP *o)
 	cLISTOPo->op_first = kid;
 	kid->op_sibling = sibl;
     }
-    pm = (PMOP*)kid;
-    if (pm->op_pmshort && !(pm->op_pmflags & PMf_ALL)) {
-	SvREFCNT_dec(pm->op_pmshort);	/* can't use substring to optimize */
-	pm->op_pmshort = 0;
-    }
 
     kid->op_type = OP_PUSHRE;
     kid->op_ppaddr = ppaddr[OP_PUSHRE];

op.h

@@ -177,16 +177,14 @@ struct pmop {
     OP *	op_pmreplstart;
     PMOP *	op_pmnext;		/* list of all scanpats */
     REGEXP *	op_pmregexp;		/* compiled expression */
-    SV *	op_pmshort;		/* for a fast bypass of execute() */
     U16		op_pmflags;
     U16		op_pmpermflags;
-    char	op_pmslen;
 };
 
 #define PMf_USED	0x0001		/* pm has been used once already */
 #define PMf_ONCE	0x0002		/* use pattern only once per reset */
-#define PMf_SCANFIRST	0x0004		/* initial constant not anchored */
-#define PMf_ALL		0x0008		/* initial constant is whole pat */
+#define PMf_REVERSED	0x0004		/* Should be matched right->left */
+/*#define PMf_ALL		0x0008*/		/* initial constant is whole pat */
 #define PMf_SKIPWHITE	0x0010		/* skip leading whitespace for split */
 #define PMf_FOLD	0x0020		/* case insensitivity */
 #define PMf_CONST	0x0040		/* subst replacement is constant */

perl.c

@@ -418,36 +418,6 @@ perl_destruct(register PerlInterpreter *sv_interp)
 
     /* defgv, aka *_ should be taken care of elsewhere */
 
-#if 0  /* just about all regexp stuff, seems to be ok */
-
-    /* shortcuts to regexp stuff */
-    leftgv = Nullgv;
-    ampergv = Nullgv;
-
-    SAVEFREEOP(curpm);
-    SAVEFREEOP(oldlastpm); /* for saving regexp context during debugger */
-
-    regprecomp = NULL;	/* uncompiled string. */
-    regparse = NULL;	/* Input-scan pointer. */
-    regxend = NULL;	/* End of input for compile */
-    regnpar = 0;	/* () count. */
-    regcode = NULL;	/* Code-emit pointer; &regdummy = don't. */
-    regsize = 0;	/* Code size. */
-    regnaughty = 0;	/* How bad is this pattern? */
-    regsawback = 0;	/* Did we see \1, ...? */
-
-    reginput = NULL;		/* String-input pointer. */
-    regbol = NULL;		/* Beginning of input, for ^ check. */
-    regeol = NULL;		/* End of input, for $ check. */
-    regstartp = (char **)NULL;	/* Pointer to startp array. */
-    regendp = (char **)NULL;	/* Ditto for endp. */
-    reglastparen = 0;		/* Similarly for lastparen. */
-    regtill = NULL;		/* How far we are required to go. */
-    regflags = 0;		/* are we folding, multilining? */
-    regprev = (char)NULL;	/* char before regbol, \n if none */
-
-#endif /* if 0 */
-
     /* clean up after study() */
     SvREFCNT_dec(lastscream);
     lastscream = Nullsv;

perl.h

@@ -1748,29 +1748,6 @@ EXT U32		hints;		/* various compilation flags */
 #define HINT_STRICT_VARS	0x00000400
 #define HINT_LOCALE		0x00000800
 
-/**************************************************************************/
-/* This regexp stuff is global since it always happens within 1 expr eval */
-/**************************************************************************/
-
-EXT char *	regprecomp;	/* uncompiled string. */
-EXT char *	regparse;	/* Input-scan pointer. */
-EXT char *	regxend;	/* End of input for compile */
-EXT I32		regnpar;	/* () count. */
-EXT char *	regcode;	/* Code-emit pointer; &regdummy = don't. */
-EXT I32		regsize;	/* Code size. */
-EXT I32		regnaughty;	/* How bad is this pattern? */
-EXT I32		regsawback;	/* Did we see \1, ...? */
-
-EXT char *	reginput;	/* String-input pointer. */
-EXT char *	regbol;		/* Beginning of input, for ^ check. */
-EXT char *	regeol;		/* End of input, for $ check. */
-EXT char **	regstartp;	/* Pointer to startp array. */
-EXT char **	regendp;	/* Ditto for endp. */
-EXT U32 *	reglastparen;	/* Similarly for lastparen. */
-EXT char *	regtill;	/* How far we are required to go. */
-EXT U16		regflags;	/* are we folding, multilining? */
-EXT char	regprev;	/* char before regbol, \n if none */
-
 EXT bool	do_undump;	/* -u or dump seen? */
 EXT VOL U32	debug;
 
@@ -2072,6 +2049,8 @@ EXT MGVTBL vtbl_mutex =	{0,	0,	0,	0,	magic_mutexfree};
 EXT MGVTBL vtbl_defelem = {magic_getdefelem,magic_setdefelem,
 					0,	0,	magic_freedefelem};
 
+EXT MGVTBL vtbl_regexp = {0,0,0,0, magic_freeregexp};
+
 #ifdef USE_LOCALE_COLLATE
 EXT MGVTBL vtbl_collxfrm = {0,
 				magic_setcollxfrm,
@@ -2114,6 +2093,7 @@ EXT MGVTBL vtbl_mutex;
 #endif /* USE_THREADS */
 
 EXT MGVTBL vtbl_defelem;
+EXT MGVTBL vtbl_regexp;
 
 #ifdef USE_LOCALE_COLLATE
 EXT MGVTBL vtbl_collxfrm;