Author: Cyril Brulebois Date: To: intrigeri CC: tails-l10n, tails-rm Subject: Re: [Tails-l10n] Translation status in 3.15
Hey,
intrigeri <intrigeri@???> (2019-07-09): > Cyril Brulebois:
> > Being in the middle of preparing 3.15, I'm going to double check the
> > changes in the jenkins-tools submodule, but it seems check_po is now
> > waaay slower than it used to be, possibly because it's now checking
> > every single PO file instead of just po/*?
>
> AFAICT check_po already checked all PO files before the recent
> changes.
>
> But the website PO files for all languages enabled on our translation
> platform are now in our Git tree so for every translatable page, we're
> now checking 15 PO files instead of 7 previously.
>
> It might also be that the new Python implementation of check_po
> introduced some performance bottlenecks. It might be that cheap
> low-hanging fruits can be spotted via basic profiling (if not done
> yet). OTOH just looking at htop(1) suggests that the bottleneck is
> still i18nspector CPU consumption, which is the part that
> hasn't changed.
I'm unlikely to go do some archeology any time soon, but I thought I'd
mention the (unfortunately number-less) “feeling” of slowness compared
to past experiences before I forget.
> In passing, while testing this I've noticed a regression: check_po
> used to (rightfully) skip files in ./tmp/ (leftovers of running
> ./import-translations) but the new version does not. Will file
> a ticket.
I wasn't sure whether something had been only partially imported when
the files under tmp/ popped up, thanks for confirming there's an issue
to fix here. And maybe that could explain part of the extended runtime
I'm perceiving.
