[Kde-scm-interest] the permissions confusion

[Oswald Buddenhagen]

On Sat, Dec 12, 2009 at 06:21:31PM +0100, Thiago Macieira wrote:
> Em Sábado 12. Dezembro 2009, às 17.35.52, Oswald Buddenhagen escreveu:
> > installing such a system also doesn't preclude allowing overrides for
> > urgent cases (say, obvious build breakage and no maintainer is to be
> > found on IRC). at qtdf, we have GIT_FORCE=yes-please to enable forced
> > pushes in justified cases.
> 
> And this relies on trusting people not to abuse. The hook checks and the 
> override were created to avoid mistakes with wide-reaching constraints (the 
> privacy issue).
> 
yes, but we don't want to put there an inpenetrable lock anyway. we just
want to make the policy obvious, and help "living it". imagine such a
message from the pre-receive hook:

************************************************************************

Whoopsie! Looks like this project has chosen to restrict commit access.

You have the following options to proceed from here:

1) run 'git magic-push'. This will submit your patch(es) for review and
integration by the maintainers.

2) run 'GIT_OVERRIDE=acl git push'. This will permit you to do the push
nonetheless. Before doing that, consider the following:

   * Is your patch really that urgent?
   * Is there really no maintainer reachable on short notice (IRC)?
   * Is your hotfix really correct? Did you have it reviewed by anyone?
   * Abuse of this facility WILL have consequences.

************************************************************************

should serve the purpose, no?

> But we also have a rule of reviewing commits, yet there are still a
> lot of unreviewed commits going in.
> 
well, yes. but it's usually people committing to the code they are
responsible for, so it's not directly related to acls. it's way harder
to get that issue fixed.