We have a staging version of our web application (it is basically a subversion working copy that no-one works on) that lives in '/apps/software'. Each developer has their own working copy in '~/apps/software'. I would like to utilise a simple post-commit hook script to update the staging copy every time a developer commits a change to the repository.
Sounds simple right? Well I've been banging my head against a brick wall on this for longer than I should. The hook script (called 'post-commit', located in /svn/software/hooks, permissions=777, user:group=apache:dev) is as follows (ignore the commented out bits for now):
#!/bin/sh
/usr/bin/svn update /apps/software >> /var/log/svn/software.log
# REPOS="$1"
# REV="$2"
# AUTHOR=`/usr/bin/svnlook author -r "$REV" "$REPOS"`
# LOG=`/usr/bin/svnlook log -r "$REV" "$REPOS"`
# EMAIL="[email protected]"
# echo "Commit log message as follows:-
#
# \"${LOG}\"
#
# The staging version has automatically been updated.
#
# See http://trac/projects/software/changeset/${REV} for more details." | /bin/mail -s "SVN : software : revision ${REV} committed by ${AUTHOR}" ${EMAIL}
That's it. The log file has the same permissions and user:group as the post-commit script and I have even given the staging copy the same user:group and permissions. Apache itself (we're using the apache subversion extension) is running under apache:dev as well. I know the hook is being executed, because the stuff that's commented out above sending an email works fine - it's just the update command that isn't.
I can also execute the post-commit hook script without environment variables using:
$ env - /svn/software/hooks/post-commit /svn/software <changeset>
and it runs fine, performing the 'svn update' no problems. I have even tried removing the '>>' to log file, but it doesn't make a difference.
Any help on this would be most appreciated...
