On Sun, Oct 19, 2014 at 6:24 PM, Jim Meyering <jim@meyering.net> wrote:
> On Sat, Oct 18, 2014 at 7:07 PM, Norihiro Tanaka <noritnk@kcn.ne.jp> wrote:
>> Jim Meyering <jim@meyering.net> wrote:
>>> dfa.c's match_mb_charset function *is* used, e.g., in a
>>> command like this one:
>>>
>>>   printf '\0' |src/grep -aE '^\s?$'
>>
>> Wow, just it isn't good.  I think that behavior of `fails' should be
>> same as of `trans' except `fails' checks accepted conditions, including
>> following part.  match_mb_charset() should be avoided as far as possible,
>> as it doesn't support collating symbols and equivalence classes.
>>
>>>               /* Falling back to the glibc matcher in this case gives
>>>                  better performance (up to 25% better on [a-z], for
>>>                  example) and enables support for collating symbols and
>>>                  equivalence classes.  */
>>>               if (d->states[s].has_mbcset && backref)
>>>                 {
>>>                   *backref = 1;
>>>                   goto done;
>>>                 }
>
> Nice change.  I've adjusted the commit log and added the test
> above, since no other code even excercised the
> now-inaccessible function. I will push it tomorrow.

By the way, I've also adjusted your preceding patch (see attached),
and will push it tomorrow, too.