Re: Translations not used in Evolution (was Re: Evolution and Evolution-Data-Server branched)

[Roozbeh Pournader <roozbeh farsiweb info>]

On Mon, 2005-09-19 at 13:54 +0530, Harish Krishnaswamy wrote:
> 2. All commits to 'po' were being made both to the stable branch and the
> HEAD. Any reason why this was not done in the case of Persian, Bengali
> translations?

Well, the double-checking-in is usually considered a good practice, but
is not necessarily easily done, specially in the last few days before a
major release. In short, let's say that there is usually a lot of
translations to review and/or check in before the deadline, and the
Internet connection from a random part of the world may not be good
enough.

So, someone who is checking in translations usually assumes that he has
six months to update different HEAD branches but only a few hours to
updated the stable branches. He will try to maximize the number of
stable branches, won't he?

BTW, it was not only these two languages, but also a few others.

>    However, it would be helpful if you could make incremental updates
> during the development cycle [2]  or have all your huge changes
> committed at least a day prior to the deadline. This would limit the
> damage caused by errors in packaging or missing commits.

People try that. We do incremental translation updates during the
instability period, but it is sometimes a waste of time and resources to
do that, simply because many messages may change or get removed
practically the next day.after the check-in.

That's the reason we have translation freezes, so we can work safely
being assured that not many message changes will happen. That one-month
is not usually enough either, and so we start earlier by asking
developers to notify the gnome-i18n list about message changes a while
before it starts.

> 1. I do not know what languages I should watch out for - before carving
> out a snapshot for pre-release testing.

You need not watch for any language. Translators take care of the
languages they care about if a maintainer announces his plan and sticks
to it (a pre-announcement of the time a maintainer will check the tree
out for final release would also be appreciated very much). They way it
happened this time, translators assumed that the Evolution 2.4.0 release
will be based on the gnome-2-12 tag, so they checked in their
translations there, instead of HEAD (or both branches).

> 2. The curtains are down not exactly when the tree is tagged but
> precedes it by a build-test-fix-verify cycle that ranges from half a day
> to one. So translators have no way of knowing for sure if their commits
> on tarballs_due_date -1 would make into a release unless they put in
> most of their work in advance or send me a personal alert.

That's understandable. So they try their best to get in their work as
early as possible. But they usually assume that a maintainer will not
release a tarball earlier than the day tarballs are due, unless he
announces so a while before. This time, I personally assumed that a
translation checked in on September 4 or before that would definitely
make it to the GNOME 2.12 releas.

> Sure. I will release  2.4.0.1 (from *gnome-2-12* branch) with the
> missing translations.

Thanks a lot. We would really appreciate that. (We also need to go ping
a few distributions, of course, which usually don't listen to us much if
it's a third world language, but well, who knows.)

Roozbeh