Re: proposal for splitting translation tests into a separate project

On 11/19/2015 04:12 PM, Matthew Miller wrote:
> On Thu, Nov 19, 2015 at 03:06:34PM -0500, Chris Lumens wrote:
>> We're never going to hear if something passes a threshold or not.  At
>> the most, we might find out if a whole new language needs to be added,
>> but that's pretty rare.
> Probably worth noting this:
> https://fedoraproject.org/wiki/Fedora_Khmer_Translation_Sprint
>
> I think it's completely fair to tell the G11N team to ask when they
> want new languages pulled in Anaconda.
>

Fair enough, I just brought up automatic additions in relation to 
automatic ejections. Just glancing at the top of the msgfmt output, 
Afrikaans, Amharic, Belarusian and Bosnian are all about 98% 
untranslated. All of those languages are "in" anaconda and have projects 
in Zanata and everything, so if removed them and then someone came along 
and updated the translations for Afrikaans, would F4A G11N notify us? 
For a less extreme case, Marathi is 58% untranslated. It probably hasn't 
been updated in a couple of releases (it would be nice to base this 
decision on that kind of a metric, but that sounds like a real pain to 
figure out). I would consider it inactive and go ahead and remove it 
LINGUAS. A hypothetical Marathi translator may not agree, so it might be 
good to just automatically check back on it since we're pulling it down 
anyway.

I'm also not real sure how to handle thresholds against the translation 
schedules. Presumably the percentage of untranslated strings will get 
higher and higher as we approach a string freeze, though we can probably 
find a threshold that kicks out the inactive ones without kicking out 
everything a week or so before beta.