Bringing this topic from meeting discussion to mailing list for
Often after completion of translation package get updated with more
string or sometime translated .po file does not get pulled while
building package. This causes L10N string breakage.
Anaconda is major candidate where we fail often with string breakage.
We started this discussion in Fedora 23. Now we are thinking how we
could monitor this better. Following are some initial discussions
happened in G11N meeting. Just though to share with you all :)
05:17:28 <pravins> #idea String Breakage monitoring (Jens, Akira)
05:18:01 <pravins> juhp: tagoh_ any further idea on this.
05:18:14 <pravins> I think we are tracking in i18n side.
05:19:20 <pravins> We did not had any discussion last time, since juhp
and tagoh_ both were not available :)
05:19:25 <juhp> I have been thinking about it a little...
05:19:37 <tagoh_> not started to work on that but we know where
upstream repo is from zanata. it could be used to check by comparing
the POT in zanata and upstream?
05:20:08 <pravins> nice idea.
05:20:24 <pravins> How about discussing this as well in mailing list?
Even devel list is fine i think.
05:20:54 <juhp> my current idea is to try track koji builds for some
specific packages looking for string changes
05:21:36 <fujiwarat> upstream is unstable.
05:21:55 <juhp> part of it might be also tracking out of date POT/po files
05:21:59 <pravins> umm, to me comparing .pot periodically or with
script looks more promising.
05:22:33 <pravins> #idea String breakage monitoring - 1. Comparing POT
in Zanata and upstream.
05:22:43 <aeng> how often do we check?
05:22:55 <pravins> We do not
05:23:01 <aeng> or plan to do the comparison?
05:23:16 <tagoh_> assuming that they are properly branching and link
to zanata. if it doesn't work, they might use zanata wrongly I
05:23:57 <pravins> aeng: that is idea proposed by tagoh_ and juhp ,
to have some script or automation.
05:24:23 <juhp> what is in zanata might not be in fedora though
05:24:29 <aeng> yup
05:24:45 <pravins> #idea String breakage monitoring - 1. Track Koji
builds for some specific packages looking for string changes.
05:24:48 <tagoh_> juhp: right. so need to pick up projects we want to
keep on track
05:25:06 <juhp> aeng, I would suggest daily or whenever there is a
build - depending on the approach we take
05:25:16 <pravins> as we discussed Anacodda is major candidate here :)
05:25:31 <juhp> tagoh_, ah no - I mean strings in zanata might not be
in fedora :)
05:25:44 <juhp> additionally we might want to track projects not in Zanata
05:26:16 <aeng> so the script might need to be dynamic enough
05:26:25 <juhp> but leveraging zanata for this certainly also makes
some sense :)
05:26:52 <juhp> or maybe we should look at both zanata, upstream and
05:27:06 <juhp> aeng, yeah
05:27:13 <tagoh_> juhp: what is the original POT to compare then?
checking on the buildsys would need a cache to check right?
05:27:41 <juhp> tagoh_, right so the script might have to
generate/cache POT files even
05:28:23 <juhp> but if we really what to track fedora string breakage
then seems to me we need to look at what is actually in Fedora
05:29:51 <juhp> also s/breakage/changes/
05:30:16 <pravins> yeah, comparing with Fedora make sense
05:30:23 <tagoh_> hm, that said those translations are coming from zanata :)
05:31:30 <tagoh_> and detecting on the build time doesn't address
anything. we can just know "they broke it again".
g11n mailing list