#metabrainz

(possibly because it’s not needed on these nodes either?)

borg-backup is deprecated, superseded by ansible borgmatic

👍

yvanzo: I guess we'll need to decide how to encode those comments in the source files. easiest way would be adding another argument to l, ln, lp

I need to go through servers to remove those dirs

btw, borg backups were triggered by a timer that should not be active anymore

bitmap: we didn't have any optional arguments right now, right? So having a final one that can be ommitted would be safe?

I think so, I haven't played with xgettext in a while

for JS it won't be an issue since we have a custom extraction script

akshaaatt: can you check that tests are run? Because i keep noticing that broken tests that result in failure sometimes get away with it

take for example the bp tests

those were broken for a long time

the lint and build tests run fine

yvanzo: if we extract it like bitmap proposed, hopefully sending it to weblate would be trivial? :)

Yes, we just have to build the POT file correctly.

yeah iiuc they'd just be added as # comments in the pot file

Comments should not make the current translations obsolete either, only changing context does.

(context as in msgctxt)

Another thing that came across the discussion recently as well about l ln lp: We should probably avoid localizing the admin UI. Maybe adding an l_admin to avoid that from happening?

jasje can you please link me to the tests that were/are broken?

akshaaatt: https://github.com/metabrainz/listenbrainz-andr...

this was broken but is now fixed

Fixed jn your PR or on main?

In*

i made changes to testUsername (Jasjeet) and yim_data.json (jasjeet) was outdated and resulted in failure but it didn't get fixed

yvanzo: oh, to signal that it shouldn't be translated? I'm fine with it

akshaaatt: the DaoTest one is old but i remember fixing it

the yim test one is the latest example

I'm not sure why we want to actively not translate admin stuff - I think it'd be lovely if admins could also eventually use the site in their own language :)

But I don't mind if you really prefer that

zas: when I run remote commands on herb & kiki, this warning shows up: "mesg: ttyname failed: Inappropriate ioctl for device"

We messed up the gradle file for tests jasje

*****i made changes to testUsername (Jasjeet) and yim_data.json (jasjeet) was outdated and resulted in failure but the test didn't fail on pr run

akshaaatt: ?

where

are most admin pages translated already?

bitmap: To not even propagate it to any POT file, a like we have l_statistics for statistics.pot.

Probably mostly just to Italian :)

Translations status is not brilliant atm. It doesn’t seem reasonable to ask translating messages that will be used by nobody because admins are English speakers anyway.

For example, there is no Italian-speaker among admins.

At the very least, I'd be happy to have l_admin

It would also probably prevent further issues such as MBS-10553.

MBS-10553: User report reason is sent to admins translated https://tickets.metabrainz.org/browse/MBS-10553

We can have no POT file, or one with low priority

The user report reason should be translated anyway, it's not for admins, it's user-facing :)

(just the admins can get it in English, I guess)

having a separate domain for them might be an option if we can signal that the strings are of low priority on weblate

Transifex had that, at least

reosarevok: It "is sent to admins" but that is just a related issue, not the same.

Sure :)

reosarevok: Are you translating/using Spanish or Estonian UI?

I've translated Spanish a long time ago

Not really using it though (I am not used to use online stuff in Spanish, it's always felt weird to me :D )

Not asking for useless translations is also respecting translators work.

I get it

It'll be kinda weird when we have a page that both admins and non-admins can use with some different strings, but I can live with that

jasje: I just checked again. At first I thought it had something to do with the path for the tests

But we do run the unit tests through github actions during CI

Anyway, the admin UI messages are here for now, it is not something that is required for migrating to Weblate, but just something that was related to what we discussed this week.

There’s no way to run instrumentation tests during CI

And I don’t think we should even aim for that

Sure :)

the only niche scenario I can think of is a user wanting to use the test server in their own language and needing the admin UIs to reproduce something

reosarevok: Good point, probably just on admin-only pages then?

yvanzo: works for me

Ok, will open a ticket.

akshaaatt: whaattt

i mean UI tests are actually the real tests

yvanzo: is there anything else we need to decide for weblate? What are the next MBS steps?

yvanzo: how do you run those remote commands? btw, kiki & herb are meant to be retired very soon

akshaaatt: https://github.com/ReactiveCircus/android-emula...

reosarevok, bitmap: The repository is set for weblate to push/pull po/pot files from the branch 'translations' for test purpose.

Would it be alright to keep it pushing to 'translations' but pulling from 'beta'?

We would just have to review changes in 'translations' branch before getting them into 'beta'.

sounds good to me, as long as the release process is documented somewhere

Hey reosarevok any luck on the PR?

yvanzo: yeah, please document that a bit so I can follow it better :)

zas: also have a related syswiki PR about the new server procedure

Pratha-Fish: sorry, didn't get to that today - maybe bitmap can take a quick look? :) If not, I will in the morning

Can you post the link as a reminder?

Pratha-Fish: can you link the PR?

oops

sure

bitmap, reosarevok: Ok, by the way please do not update Transifex anymore.

Not actually jasje. I have seen that repo you shared

It’s expensive to run these instrumentation tests

but they are very important afterall

bitmap reosarevok here's the PR

Even though we do have free open source plan for github and can run it, I don’t think it’s a good idea to run them over CI

What we can do is make sure to run them during releases

so whos gonna run it

But not for CI

human error is way too much in this

I’ll add them to the deployment pipeline

alright

akshaaatt: how much time does it take btw?

Time for?

you said it was expensive

was it in terms of time?

No

akshaaatt: we run selenium tests with (Jenkins) CI, but not sure how that compares to your stuff :)

then?

The actual computation

reosarevok: in our case, we need to setup an emulator all together

And then boot it up, load the app there, and do it

yvanzo: Sure :) But we should amend RELEASING.md then :)

re: not using transifex

Tbh github actions are damn expensive for commerical companies. They’re not even useful for CI stuff because of the expense

but its cloud right?

So?

It runs on actual servers

so why do we worry about computation, its githubs problem right?

reosarevok: I mean during the current migration, thanks.

If you see the configurations we do for GH, you’ll see we use ubuntu for it

Energy as a resource is everyone’s problem jasje. Why waste it unnecessarily?

yvanzo: so on Monday we should pull from transifex but not push to it?

GH actions*

reosarevok: no transifex at all, I hope to have completed the switchover by then.

i mean if you go by that

sure

run it on deployment

Oh, ok :)

Sure. Cool

but its just that you will need to run and tell the contributor yourself about failing test

and what test its failing

why is it failing

Sure, we can handle that.

alright

lucifer: I forgot to mention I fixed the tests on this PR https://github.com/metabrainz/listenbrainz-serv... yesterday

> Energy as a resource is everyone’s problem jasje. Why waste it unnecessarily?

akshaaatt: speaking wise words. <3