Re: [PATCH v4 02/73] gettext tests: add GETTEXT_POISON to simulate unfriendly translator

From: Junio C Hamano <gitster@pobox.com>
Date: Feb 25, 2011, 17:52 UTC
Message-ID: <7vvd08atnu.fsf@alter.siamese.dyndns.org>
In-Reply-To: <20110225070834.GB23037@elie>

Jonathan Nieder <jrnieder@gmail.com> writes:

[offtopic: why is Jakub on the CC list? -- I don't recall him saying much on this topic yet...]

> The idea of the GETTEXT_POISON facility is to simulate a worst-case
> translation, so that strings marked for translation that cause trouble
> can be noticed and dealt with.

I've never thought this was about a "worst-case translation".

Isn't this about emitting all the strings marked for translation in a form significantly different from the original, in the hope that such an output would help us find messages that should never be marked for translation by triggering test failures? I do not see any "worst-case", just different enough for the plumbing tests to break.

IOW, I don't think this is about helping translators in l10n, but about helping coders in i18n department.

> A string starting with "#" does not seem particularly well suited
> for that purpose, since it makes '#' characters that mark a comment
> and that translators must be careful not to remove harder to find.

Hence, I would suggest an alternative phrasing of the above part (and alternative way of thinking): '... since it easily hides a mistaken marking of a line that _has_ to always start with "#" followed by a translatable string.'