Re: [PATCH 03/65] i18n: archive: mark parseopt strings for translation

From: Nguyen Thai Ngoc Duy <pclouds@gmail.com>
Date: Jun 7, 2012, 12:53 UTC
Message-ID: <CACsJy8D=SMmhiyp-+pL5FxfcUGdLp9X4D42eV0uAGX3_LAa0xA@mail.gmail.com>
In-Reply-To: <20120607124155.GB12987@burratino>

On Thu, Jun 7, 2012 at 7:41 PM, Jonathan Nieder <jrnieder@gmail.com> wrote:

Show 9 quoted lines

> Nguyen Thai Ngoc Duy wrote:
>
>> No changes to the series then. Junio, pu material perhaps? The
>> breaking gettext poison thing is different and I plan to kill gettext
>> poison soon (whether i succeed is another question of course).
>
> If you mean to kill it by replacing it with a facility to run tests in
> another locale, wouldn't all these testsuite bugs still be testsuite
> bugs?

They still are and should be fixed when the new facility comes, so there's no point fixing them now. I played with the poison .po generation a bit and started to use libgettext-po, which would give us more control over reading/writing .po. We could even perform git-specific checks on po files..

The said plan is to build .mo for $LANG in po/build/poison-locale (po/build/locale remains good translation) and add a switch to make the test suite use poison-locale instead. Long term I'd like to switch to poison locale by default, unless LANG is C, maybe with an option to run with real .mo files.

-- 
Duy