diff options
| author | Karl Williamson <khw@cpan.org> | 2015-10-19 12:14:36 -0600 |
|---|---|---|
| committer | Steve Hay <steve.m.hay@googlemail.com> | 2015-10-30 21:15:52 +0000 |
| commit | 23dba96959ea693ba5c974f57c08db2d1d66bd4b (patch) | |
| tree | d80e05e7591ccc5ef900fd747f0a3a87f54e0603 | |
| parent | 9e053d5265641a359970d9a120a0247e29e78337 (diff) | |
| download | perl-23dba96959ea693ba5c974f57c08db2d1d66bd4b.tar.gz | |
"" =~ /\b{gcb}/ should fail; same \b{wb}, \b{sb}
The Unicode standard indicates that these breaks should succeed at the
beginning and end of text. It appears to me to be an oversight on their
part to not make an exception when there is no actual text. (Their test
suite does not cover this case.) I blindly implemented their algorithm
for 5.22, but it really is the wrong thing to do.
(cherry picked from commit a7a8bd1ed56dbdb7e63735924945bbb66b7e2e5c)
| -rw-r--r-- | regexec.c | 24 |
1 files changed, 17 insertions, 7 deletions
@@ -2073,13 +2073,17 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, FBC_BOUND(isWORDCHAR_L1, isWORDCHAR_uni, isWORDCHAR_utf8); break; case GCB_BOUND: - if (s == reginfo->strbeg) { /* GCB always matches at begin and - end */ + if (s == reginfo->strbeg) { if (reginfo->intuit || regtry(reginfo, &s)) { goto got_it; } + + /* Didn't match. Try at the next position (if there is one) */ s += (utf8_target) ? UTF8SKIP(s) : 1; + if (UNLIKELY(s >= reginfo->strend)) { + break; + } } if (utf8_target) { @@ -2118,13 +2122,14 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, break; case SB_BOUND: - if (s == reginfo->strbeg) { /* SB always matches at beginning */ + if (s == reginfo->strbeg) { if (reginfo->intuit || regtry(reginfo, &s)) { goto got_it; } - - /* Didn't match. Go try at the next position */ s += (utf8_target) ? UTF8SKIP(s) : 1; + if (UNLIKELY(s >= reginfo->strend)) { + break; + } } if (utf8_target) { @@ -2183,6 +2188,9 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, goto got_it; } s += (utf8_target) ? UTF8SKIP(s) : 1; + if (UNLIKELY(s >= reginfo->strend)) { + break; + } } if (utf8_target) { @@ -5617,8 +5625,10 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) case BOUNDU: /* /\b/u */ boundu: - if (utf8_target) { - + if (UNLIKELY(reginfo->strbeg >= reginfo->strend)) { + match = FALSE; + } + else if (utf8_target) { bound_utf8: switch((bound_type) FLAGS(scan)) { case TRADITIONAL_BOUND: |