Autore: sajolida Data: To: Tails localization discussion Oggetto: Re: [Tails-l10n] Weblate & manual reviews of Git
Sandro Knauß: >> I understand that Weblate is not yet ready for production so I guess
>> it's still a good time to discuss this.
>>
>> I would like to continue checking all changes on tails.git whenever I
>> fetch it:
>>
>> * Would it be possible to further lower the amount of commits done by
>> Weblate? What would be the other drawbacks?
>
> We have already "lazy commit" activated to lower the amounts of commits from
> beginning, that should wait as long as needed before committing. At least I
> know no easy solution to lower the amount of commits further.
Too bad :) Thanks for the clarification.
>> * Would it be possible to document tricks on how to do such a review
>> more easily (and maybe more consistently)?
>
> Properly ;D But as I don't know what you want to spot etc. so I can't help
> here.
Sure. Personally, I switched from reading the log to reading the diff
only. It's somehow easier to filter out all the PO files and I'll see
how it goes after some time.
>> * 2 of these commits were particularly scary (179e0bdaba and
>> fc87cf823e) and I'm glad I reviewed then :) I understand that emmapeel
>> did a mistake and then fixed it, which is fine.
>
> It was not not done by emmapeel, it were our scripts that had a race
> condition. (It was just triggered by changes made by emmapeel within Weblate).
> You can see that, because the committer is "tails-l10n@???" and not
> emmapeel themself.
>
>> How could prevent such mistakes from happening in the first place?
>
> Hopefully it is done as we solved #16844.
Good!
>> Has Weblate or Weblate admins the rights to override any file in
>> tails.git, even if it's not a PO file? Could we restrict this further?
>
> Weblate can only push PO file changes.
Nice!
> But the hook that restrict Weblate
> commits look at the overall change done by the commits that get pushed. That
> means, that Weblate couldn't push 179e0bdaba. The combination of
> 179e0bdaba+fc87cf823e has no changes other than PO files, that's why this got
> accepted.
I understand now. Thanks for putting all these safeguards in place :)