From: Fredrik Kuivinen <frekui@gmail.com>
Date: Sun, 28 Aug 2011 11:31:49 GMT
Subject: Re: [PATCH v2 5/5] Use kwset in grep
Message-ID: <CALx8hKTuZ5hG6fGUnTbRPLifF=N-SJcfmgYAKUVDO-jgzrKRaA@mail.gmail.com>
In-Reply-To: <4E51F998.50801@gnu.org>

On Mon, Aug 22, 2011 at 08:39, Paolo Bonzini <bonzini@gnu.org> wrote:
> On 08/21/2011 12:42 AM, Fredrik Kuivinen wrote:
>>
>> +               if (opt->regflags&  REG_ICASE || p->ignore_case) {
>> +                       static char trans[256];
>> +                       int i;
>> +                       for (i = 0; i<  256; i++)
>> +                               trans[i] = tolower(i);
>> +                       p->kws = kwsalloc(trans);
>> +               } else {
>> +                       p->kws = kwsalloc(NULL);
>> +               }
>
> Of course, this makes absolutely no sense for MB_CUR_MAX > 1.  It's worth
> mentioning that grep instead uses a loop with mbrtowc/towlower/wcrtomb.
>  This in turn will remove the need for the complex kwset code. :)

Good catch. At least it is not a regression from the current behavior,
neither our own strcasestr in compat/ nor strcasestr in glibc can
handle MB_CUR_MAX > 1. My original idea was to make use of kwset also
for the case when more than one fixed string is given to git-grep, but
I didn't find a nice way to refactor the code to make that possible.

> The "mbtolower" code" dates to after the license change, but I wrote it and
> I give permission to use it under GPLv2.  See commits 70e23616 and 30af8050
> in the GNU grep repository.
>
> Should still be good enough for most uses, so I'll give my
>
> Acked-by: Paolo Bonzini <bonzini@gnu.org>

Thanks.

- Fredrik