Re: Commit gettext .po files without extra comments

Le dimanche 06 juin 2010 à 20:07 +0200, Ulrik Sverdrup a écrit :
> 2010/6/6 Petr Kovar <pmkovar gnome org>:
> > Hi Ulrik!
> >
> > Ulrik Sverdrup <ulrik sverdrup gmail com>, Fri, 4 Jun 2010 23:18:28 +0200:
> >
> > (...)
> >
> >> I have also sent this email to gnome-i18n because I want your feedback
> >> on this proposal. Summary: Filename+line comments are still available
> >> when updating the .po file, but not shown in diffs or stored in the
> >> repository history. All translators have to opt-in.
> >
> > So if I'm not missing something, this l10n work flow is not to be required
> > from translators and is completely optional, right?
> Exactly.
> I realize that gnome translators absolutely don't want different
> settings / practices for different modules, so this whole proposal is
> a bit a lost case for me, if it doesn't get many people interested.

I admit I was dubious at first. But after some thinking, I don't see any
major drawbacks to this practice. It even complies with the rule that
data that can be dynamically generated should not be committed to a VCS.

Just to be sure, I also asked gettext maintainer, Bruno Haible, about
his opinion on this matter. He answered me the following:

"As long as these PO files won't go into the hands of translators, this
is fine.

If they go into the hands of the translators, the "extracted comments"
(#. lines) and file references (#: lines) can be restored by running
'msgmerge' on an updated POT file. This step should be done before
you distribute the PO files to the translators; it should not be the
translator's job to do this.

It all depends how the workflow with the translators is organized."

The two major use-cases used by GNOME translators ( and
intltool-update) both regenerate automatically the #. and #: lines.

So I invite GTP committers to test this method with kupfer. Then, unless
we discover any major problem, it might even be suggested for other
GNOME modules.


> >
> > Anyway, I'd like to note that Vertimus running on provides
> > us already with quite a comfort diff utility that tries to ignore the
> > mentioned noise with comment line changes (because it updates both the
> > committed and uploaded translation catalog against the current
> > auto-generated POT file).
> Yes, this is because translations are reviewed before they enter the
> repository. I'm not mainly a translator, but a module maintainer, so
> I'd like to see this implemented so that diffs look much better when
> the changes have actually entered the repository.
> Ulrik
> _______________________________________________

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]