views:

53

answers:

2

I have a central git repo set up using gitolite.

I want to set up a hook such that whenever a user pushes to the repo, it performs a pull elsewhere followed by some automated testing.

So far, I only want to it perform the pull.

In the hooks directory I created the following script names post-update:

#!/bin/sh  
cd /home/git/www/epicac
git pull

When I invoke this script using ./post-update, it does exactly what I want.

However, whenever it's invoked automatically as I hook, I get: fatal: Not a git repository: '.'

Any idea why this might be happening?

A: 

You have various diagnostics to run as suggested in this SO answer.

In particular, check out the the value of GIT_DiR and GIT_WORK_TREE.

While the hook is running, GIT_DIR and (if the worktree can't be inferred from GIT_DIR) GIT_WORK_TREE are set.
That means your pull won't run with the repository in the directory you changed to.


See also blog post Using Git Inside a Git Hook:

Eventually we got our linux guru over and he noticed that the environment under which the git user runs is totally different when inside a hook.
Gitolite does a bunch of things to the env, but the one that was screwing us up was the setting of the GIT_DIR.
After we figured that out, the solution was as easy as:

ENV.delete 'GIT_DIR'

in our ruby script that is triggered by the 'post-receive' hook.


Same deal in Git Tip: Auto update working tree via post-receive hook, but with an elegant way out of this:

The solution?
It turns out the post-receive hook starts out with the GIT_DIR environment variable set to the repo/.git folder, so no matter what path you 'cd' into it will always try to run any following git commands there.
Fixing this is simply a matter of unsetting the GIT_DIR
(thanks to Ulrich Petri for the elegant env -i solution):

#!/bin/sh
cd ..
env -i git reset --hard
VonC