On Thu, Mar 27, 2014 at 12:05 PM, Jim Meyering <[email protected]> wrote:
> On Thu, Mar 27, 2014 at 5:54 AM, Norihiro Tanaka <[email protected]> wrote:
>> Jim,
>>
>> Thanks, I have added the comments to the patch and have slightly
>> modified the comment you wrote.
>
> Norihiro,
>
> I see you also added a test case.  Thanks for that.  I've amended that
> commit log message with a corresponding ChangeLog entry.
>
> You're right that "beg" must not be advanced, so I have adjusted the
> patch to factor out the call to dfaexec without changing "beg".  I've also
> augmented the first test to require an exact match of the sole input line,
> and changed the other test to require not just a nonzero exit, but $? = 1.

And here's the patch:
From cb1e339f576fa3a8431dd544edfc1d3b7792ddf2 Mon Sep 17 00:00:00 2001
From: Norihiro Tanaka <[email protected]>
Date: Wed, 26 Mar 2014 08:56:50 -0700
Subject: [PATCH] grep: perform the kwset-helping DFA match in narrower range

When kwsexec gives us the offset of a potential match, we compute
line begin/end and then run the DFA matcher to see if there really
is a match on that line.  When the beginning of the line, BEG, is
not on a multibyte character boundary, advance BEG until it on such
a boundary, before running the DFA search.
* src/dfasearch.c (EGexecute): As above.  Add a comment.
* tests/euc-mb: Add a test case that exercises this code.
This addresses http://debbugs.gnu.org/17095.
---
 src/dfasearch.c |  8 +++++++-
 tests/euc-mb    | 11 +++++++++--
 2 files changed, 16 insertions(+), 3 deletions(-)

diff --git a/src/dfasearch.c b/src/dfasearch.c
index 0b56960..d098a9b 100644
--- a/src/dfasearch.c
+++ b/src/dfasearch.c
@@ -236,6 +236,7 @@ EGexecute (char const *buf, size_t size, size_t *match_size,
               match = beg;
               while (beg > buf && beg[-1] != eol)
                 --beg;
+              char const *dfa_start = beg;
               if (kwsm.index < kwset_exact_matches)
                 {
                   if (!MBS_SUPPORT)
@@ -247,8 +248,13 @@ EGexecute (char const *buf, size_t size, size_t 
*match_size,
                       || !is_mb_middle (&mb_start, match, buflim,
                                         kwsm.size[0]))
                     goto success;
+                  /* The matched line starts in the middle of a multibyte
+                     character.  Perform the DFA search starting from the
+                     beginning of the next character.  */
+                  dfa_start = mb_start;
                 }
-              if (dfaexec (dfa, beg, (char *) end, 0, NULL, &backref) == NULL)
+              if (dfaexec (dfa, dfa_start, (char *) end, 0, NULL,
+                           &backref) == NULL)
                 continue;
             }
           else
diff --git a/tests/euc-mb b/tests/euc-mb
index c0af220..f44253f 100755
--- a/tests/euc-mb
+++ b/tests/euc-mb
@@ -30,7 +30,14 @@ fail=0
 # Does EUC-JP work at all?
 make_input BABA |euc_grep AB && fail=1

-# Whole line rejected after matching in the middle of a multibyte char?
-make_input BABAAB |euc_grep AB || fail=1
+# Here are two cases in which a KWSet search matches in the middle
+# of a multibyte character.  The first ensures that the DFA matcher
+# finds the real match at the end of line.  The second ensures that
+# while the KWSet match found a false positive, the DFA matcher
+# determines there is no match after all.
+make_input BABAAB |euc_grep AB > out || fail=1
+make_input BABAAB > exp || framework_failure_
+compare out exp || fail=1
+make_input BABABA |euc_grep AB; test $? = 1 || fail=1

 Exit $fail
-- 
1.9.0.258.g00eda23

Reply via email to