Re: where do the canonical translations live, git or pootle

On 30/01/15 12:01, Lionel Elie Mamane wrote:
> On Fri, Jan 30, 2015 at 11:32:20AM +0000, Anthonys Lists wrote:
> > On 30/01/2015 08:30, Lionel Elie Mamane wrote:
>
> > > How about the git commit/push is made to*fail*  for the developer
> > > then? Some git commit hook or something. The commit fails unless the
> > > developer has set environment variable
> > > "YES_THIS_IS_AN_IMPORT_FROM_POOTLE=yes" or
> > > "AWARE_OF_TRANSLATION_WORKFLOW=yes".
>
> > Isn't this behaviour that git is supposed to detect? If you push a
> > change, then I push a change, git is supposed to detect that my copy
> > is "stale" and force a rebase before I can push?
>
> My guess is because the pootle-to-git push is not done as a git
> merge, but as 'overwrite what is in the working directory from a fresh
> git checkout'.
>
> You have in mind:
>
>  * pootle has a git checkout, and changes by translators are committed
>    there.
>  * Sometimes, that checkout is pushed.
>
> What I think happens is:
>
>  * Sometimes, someones makes "git pull", then exports the contents of
>    pootle to that workdir and then "git commit -a && git push".
In other words, classic tool abuse :-(

That was my reaction when I saw git mentioned - "shouldn't git prevent
that?". Not knowing how things are being done, I can't really make any
recommendations, but how hard is it to fix that workflow and use git
properly? Wouldn't that fix matters?

Cheers,
Wol