[GRASS-dev] Managing the translations with Transifex

[Moritz Lennert]

On 27/06/17 12:06, Markus Neteler wrote:
> On Tue, Jun 27, 2017 at 9:08 AM, Moritz Lennert
> <mlennert at club.worldonline.be> wrote:
>> On 26/06/17 20:37, Markus Neteler wrote:
> ...
>> Sorry, Markus, but on my side, I'm just currently completely ignorant about
>> translation procedures and issues in GRASS. I didn't even know we had a
>> solution via transifex (which is great BTW !)...
>
> NP, it was loosely discussed here and there in the list.

Must have missed it, sorry. I think some of the discussion went on in 
the translation list...

>> - I don't know what the issue concerning empty translations is about.
>
> The current (new) flow is [Luca, correct me if I am wrong]:
>
> 1. messages are extracted from the source code in trunk with the
> common gettext mechanism (done by the transifex portal)
> 2. folks translate in the transifex portal
> 3. I or any other developer sometimes merges the translations back
> using this script:
>     grass-addons/tools/transifex_merge.sh
> 4. start from 1.
>
> Issue:
> - how to get messages intro relbranch72

Does transifex allow comparison and merge between projects ? I.e. would 
it be possible to create a relbranch72 project that imports from trunk, 
except for those strings that are different between the two ?

> - can step 3 be done with a cronjob? which frequency?

That obviously depends on its reliability, but as said, if you think 
that it +/- works maybe we should just take the risk and activate it. 
You should probably include a distinctive keyword in the svn message 
when comitting the merged files, so we can always reverse if necessary.

>
>> - I don't know what the issue concerning the choice between msgmerge and
>> transifex_merge.sh, and the backport, is about. I don't even know where to
>> find these files.

Sorry, my local grass-addons/tools directory was not up to date. That's 
why I didn't see the script.

> Well, it has been discussion, but with too much time in between.
> I just want to avoid to break stuff on the transifex portal.

How will the cron job risk breaking stuff in the portal ? I thought the 
it was just extracting from the portal and merging into svn ? Is there a 
risk of breakage when transifex then reads the po files from svn again ?


>> And from my ignorance a question: why don't we "just" move all translation
>> to transifex and so avoid the need for merges between translations from
>> different sources.
>
> This is how it is! But of course we must feed our .po files in the
> source code somehow.
> How this clarifies it for you.

It does, somewhat, thank you !

How does QGIS handle this issue ?

And is the integration between github and transifex easier than between 
svn and transifex ? That would make the case for migration even stronger.

One thing I take from this is that once this is ready, we should make a 
big call for translation, announcing the possibility to use transifex 
and very easily translate a few messages online.

Moritz