diff -up perl-5.10.1/ext/re/t/regop.t.git perl-5.10.1/ext/re/t/regop.t --- perl-5.10.1/ext/re/t/regop.t.git 2009-12-21 19:31:07.564141841 +0100 +++ perl-5.10.1/ext/re/t/regop.t 2009-12-21 19:31:55.158142088 +0100 @@ -233,12 +233,12 @@ anchored "ABC" at 0 #Freeing REx: "(\\.COM|\\.EXE|\\.BAT|\\.CMD|\\.VBS|\\.VBE|\\.JS|\\.JSE|\\."...... %MATCHED% floating ""$ at 3..4 (checking floating) -1:1[1] 3:2[1] 5:2[64] 45:83[1] 47:84[1] 48:85[0] -stclass EXACTF <.> minlen 3 -Found floating substr ""$ at offset 30... -Does not contradict STCLASS... -Guessed: match at offset 26 -Matching stclass EXACTF <.> against ".exe" +#1:1[1] 3:2[1] 5:2[64] 45:83[1] 47:84[1] 48:85[0] +#stclass EXACTF <.> minlen 3 +#Found floating substr ""$ at offset 30... +#Does not contradict STCLASS... +#Guessed: match at offset 26 +#Matching stclass EXACTF <.> against ".exe" --- #Compiling REx "[q]" #size 12 nodes Got 100 bytes for offset annotations. @@ -258,4 +258,4 @@ Got 100 bytes for offset annotations. Offsets: [12] 1:1[3] 3:4[0] %MATCHED% -Freeing REx: "[q]" \ No newline at end of file +Freeing REx: "[q]" diff -up perl-5.10.1/regcomp.c.git perl-5.10.1/regcomp.c --- perl-5.10.1/regcomp.c.git 2009-12-21 19:32:05.893141719 +0100 +++ perl-5.10.1/regcomp.c 2009-12-21 19:33:35.106141384 +0100 @@ -2820,13 +2820,16 @@ S_study_chunk(pTHX_ RExC_state_t *pRExC_ } } else { /* - Currently we assume that the trie can handle unicode and ascii - matches fold cased matches. If this proves true then the following - define will prevent tries in this situation. - - #define TRIE_TYPE_IS_SAFE (UTF || optype==EXACT) -*/ + Currently we do not believe that the trie logic can + handle case insensitive matching properly when the + pattern is not unicode (thus forcing unicode semantics). + If/when this is fixed the following define can be swapped + in below to fully enable trie logic. #define TRIE_TYPE_IS_SAFE 1 + +*/ +#define TRIE_TYPE_IS_SAFE (UTF || optype==EXACT) + if ( last && TRIE_TYPE_IS_SAFE ) { make_trie( pRExC_state, startbranch, first, cur, tail, count, diff -up perl-5.10.1/regexec.c.git perl-5.10.1/regexec.c --- perl-5.10.1/regexec.c.git 2009-12-21 19:33:50.570141632 +0100 +++ perl-5.10.1/regexec.c 2009-12-21 19:36:41.300142175 +0100 @@ -1006,16 +1006,15 @@ Perl_re_intuit_start(pTHX_ REGEXP * cons #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \ uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \ - UV uvc_unfolded = 0; \ switch (trie_type) { \ case trie_utf8_fold: \ if ( foldlen>0 ) { \ - uvc_unfolded = uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ + uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ foldlen -= len; \ uscan += len; \ len=0; \ } else { \ - uvc_unfolded = uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ + uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \ foldlen -= UNISKIP( uvc ); \ uscan = foldbuf + UNISKIP( uvc ); \ @@ -1054,9 +1053,6 @@ uvc, charid, foldlen, foldbuf, uniflags) charid = (U16)SvIV(*svpp); \ } \ } \ - if (!charid && trie_type == trie_utf8_fold && !UTF) { \ - charid = trie->charmap[uvc_unfolded]; \ - } \ } STMT_END #define REXEC_FBC_EXACTISH_CHECK(CoNd) \