summaryrefslogtreecommitdiff
path: root/regexec.c
diff options
context:
space:
mode:
authorKarl Williamson <public@khwilliamson.com>2011-01-18 16:06:47 -0700
committerKarl Williamson <public@khwilliamson.com>2011-01-18 16:35:16 -0700
commit6895a8aa39da30e5d27a918f8082897ce5f21779 (patch)
tree47e6bfef368570db57c730e10e2926c1cf1463da /regexec.c
parent8e9da4d422fd1fb3711b88105d4e58f6b8f88877 (diff)
downloadperl-6895a8aa39da30e5d27a918f8082897ce5f21779.tar.gz
regexec.c: Don't rely on break stmts in macros
It is safer and clearer to have the break statement in each case statement at the source level
Diffstat (limited to 'regexec.c')
-rw-r--r--regexec.c24
1 files changed, 24 insertions, 0 deletions
diff --git a/regexec.c b/regexec.c
index 3f38828616..29058a1e47 100644
--- a/regexec.c
+++ b/regexec.c
@@ -1676,34 +1676,40 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s,
isALNUM_LC_utf8((U8*)s),
isALNUM_LC(*s)
);
+ break;
case ALNUMU:
REXEC_FBC_CSCAN_PRELOAD(
LOAD_UTF8_CHARCLASS_PERL_WORD(),
swash_fetch(RE_utf8_perl_word,(U8*)s, utf8_target),
isWORDCHAR_L1((U8) *s)
);
+ break;
case ALNUM:
REXEC_FBC_CSCAN_PRELOAD(
LOAD_UTF8_CHARCLASS_PERL_WORD(),
swash_fetch(RE_utf8_perl_word,(U8*)s, utf8_target),
isWORDCHAR((U8) *s)
);
+ break;
case ALNUMA:
/* Don't need to worry about utf8, as it can match only a single
* byte invariant character */
REXEC_FBC_CLASS_SCAN( isWORDCHAR_A(*s));
+ break;
case NALNUMU:
REXEC_FBC_CSCAN_PRELOAD(
LOAD_UTF8_CHARCLASS_PERL_WORD(),
swash_fetch(RE_utf8_perl_word,(U8*)s, utf8_target),
! isWORDCHAR_L1((U8) *s)
);
+ break;
case NALNUM:
REXEC_FBC_CSCAN_PRELOAD(
LOAD_UTF8_CHARCLASS_PERL_WORD(),
!swash_fetch(RE_utf8_perl_word, (U8*)s, utf8_target),
! isALNUM(*s)
);
+ break;
case NALNUMA:
REXEC_FBC_CSCAN(
!isWORDCHAR_A(*s),
@@ -1715,39 +1721,46 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s,
!isALNUM_LC_utf8((U8*)s),
!isALNUM_LC(*s)
);
+ break;
case SPACEU:
REXEC_FBC_CSCAN_PRELOAD(
LOAD_UTF8_CHARCLASS_PERL_SPACE(),
*s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target),
isSPACE_L1((U8) *s)
);
+ break;
case SPACE:
REXEC_FBC_CSCAN_PRELOAD(
LOAD_UTF8_CHARCLASS_PERL_SPACE(),
*s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target),
isSPACE((U8) *s)
);
+ break;
case SPACEA:
/* Don't need to worry about utf8, as it can match only a single
* byte invariant character */
REXEC_FBC_CLASS_SCAN( isSPACE_A(*s));
+ break;
case SPACEL:
REXEC_FBC_CSCAN_TAINT(
isSPACE_LC_utf8((U8*)s),
isSPACE_LC(*s)
);
+ break;
case NSPACEU:
REXEC_FBC_CSCAN_PRELOAD(
LOAD_UTF8_CHARCLASS_PERL_SPACE(),
!( *s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target)),
! isSPACE_L1((U8) *s)
);
+ break;
case NSPACE:
REXEC_FBC_CSCAN_PRELOAD(
LOAD_UTF8_CHARCLASS_PERL_SPACE(),
!(*s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target)),
! isSPACE((U8) *s)
);
+ break;
case NSPACEA:
REXEC_FBC_CSCAN(
!isSPACE_A(*s),
@@ -1759,27 +1772,32 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s,
!isSPACE_LC_utf8((U8*)s),
!isSPACE_LC(*s)
);
+ break;
case DIGIT:
REXEC_FBC_CSCAN_PRELOAD(
LOAD_UTF8_CHARCLASS_POSIX_DIGIT(),
swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target),
isDIGIT(*s)
);
+ break;
case DIGITA:
/* Don't need to worry about utf8, as it can match only a single
* byte invariant character */
REXEC_FBC_CLASS_SCAN( isDIGIT_A(*s));
+ break;
case DIGITL:
REXEC_FBC_CSCAN_TAINT(
isDIGIT_LC_utf8((U8*)s),
isDIGIT_LC(*s)
);
+ break;
case NDIGIT:
REXEC_FBC_CSCAN_PRELOAD(
LOAD_UTF8_CHARCLASS_POSIX_DIGIT(),
!swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target),
!isDIGIT(*s)
);
+ break;
case NDIGITA:
REXEC_FBC_CSCAN(
!isDIGIT_A(*s),
@@ -1791,31 +1809,37 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s,
!isDIGIT_LC_utf8((U8*)s),
!isDIGIT_LC(*s)
);
+ break;
case LNBREAK:
REXEC_FBC_CSCAN(
is_LNBREAK_utf8(s),
is_LNBREAK_latin1(s)
);
+ break;
case VERTWS:
REXEC_FBC_CSCAN(
is_VERTWS_utf8(s),
is_VERTWS_latin1(s)
);
+ break;
case NVERTWS:
REXEC_FBC_CSCAN(
!is_VERTWS_utf8(s),
!is_VERTWS_latin1(s)
);
+ break;
case HORIZWS:
REXEC_FBC_CSCAN(
is_HORIZWS_utf8(s),
is_HORIZWS_latin1(s)
);
+ break;
case NHORIZWS:
REXEC_FBC_CSCAN(
!is_HORIZWS_utf8(s),
!is_HORIZWS_latin1(s)
);
+ break;
case AHOCORASICKC:
case AHOCORASICK:
{