Re: [PATCH] gitweb: Fix the author initials in blame for non-ASCII names

From: Jakub Narębski <jnareb@gmail.com>
Date: Aug 31, 2013, 16:42 UTC
Message-ID: <CANQwDwdrmyMLZ299t4T-o6uGmfbHL_A83R37_zy0-jyFXEq1+A@mail.gmail.com>
In-Reply-To: <74435A51-B404-4EC8-8CF8-6289F49E6DBD@gmail.com>

On Fri, Aug 30, 2013 at 11:39 PM, Kyle J. McKay <mackyle@gmail.com> wrote:

> On Aug 30, 2013, at 11:13, Junio C Hamano wrote:
>> Junio C Hamano <gitster@pobox.com> writes:
>>> Ævar Arnfjörð Bjarmason  <avarab@gmail.com> writes:

>>>> +       binmode $fh, ':utf8';

Show 8 quoted lines

> What happens if the author name is written in ISO-8859-1 instead of UTF-8 in
> the actual commit object itself?
>
> I'm pretty sure I've seen this where older commits have a ISO-8859-1 author
> name and then newer commits have a UTF-8 version of the same author's name.
>
> In fact, in the git repository itself, look at commit 0cb3f80d (UTF-8) and
> commit 7eb93c89 (ISO-8859-1) to see this in action.

Well, then you have a problem, though it is only with old history (before introduction of "encoding" header in commit object).

Better and more complete solution would be to use to_utf8() function instead of 'utf8' layer, which when finding invalid UTF-8 sequence uses $fallback_encoding (by default "latin1", i.e. ISO-8859-1) instead.

In my TODO list is creating PerlIO layer ':utf8-with-fallback' which would replace all those to_utf8() calls...

-- 
Jakub Narebski