ssh: send/accept GIT_* env vars

[Kacper Kornet]

On Wed, May 04, 2011 at 03:25:07PM +0200, Artur Frysiak wrote:
> On Wed, May 4, 2011 at 15:18, Kacper Kornet <kornet at camk.edu.pl> wrote:
> > On Wed, May 04, 2011 at 11:51:44AM +0200, Arkadiusz Miskiewicz wrote:
> >> On Wednesday 04 of May 2011, Kacper Kornet wrote:
> >> > On Wed, May 04, 2011 at 11:37:16AM +0200, Arkadiusz Miskiewicz wrote:
> >> > > On Tuesday 03 of May 2011, Kacper Kornet wrote:
> >> > > > What is the reason that GIT_ variables are copied by ssh to remote
> >> > > > session? I mean the following line in our default /etc/ssh/ssh_config.

> >> > > > AcceptEnv LANG LC_* LANGUAGE TZ GIT_*

> >> > > To avoid a need to set all that on each remote system.

> >> > > > I have just discovered that it break at least one feature in gitolite.

> >> > > And the bug is where?

> >> > The creation of wildcard repositories
> >> > http://sitaramc.github.com/gitolite/doc/wildcard-repositories.html

> >> > does not work. I can patch gitolite for it. But there is a small chance
> >> > the patch would be accepted upstream.

> >> How so? gitolite doesn't accept patches that fix bugs?

> > So the gitolite author's answet to the patch is:

> >> You seem to have a very unusual environment.  I'd need to understand
> >> in more detail why and what this is before I would take this patch.

> > So if someone (glen?) can give good a good reason why the GIT_*
> > variables are propagated in PLD it would be great.

> Kacper, can you provide more information? Which GIT_* variable is a
> problem for gitolite?

It is GIT_DIR. The problem is that git clone sets GIT_DIR variable.
Therefore remote git init fails during creation of wildcard repository
in this case.
-- 
  Kacper