Re: git archive stuck at mar 21?

[Ken Raeburn <raeburn@raeburn.org>]

On Mar 25, 2010, at 10:34, Randal L. Schwartz wrote:
> Óscar> The "bzr-to-git" is manually operated by Andreas Schwab, so the git
> Óscar> mirror is updated when he has time enough to do it.
> 
> Manually as in "he types a command"?  It would seem that could be completely
> cron-driven, no?

I could guess maybe it's got to be watched for errors or updated for new branch names or something.  Or maybe it just accesses the bzr repo with his ssh key that he doesn't want accessible via random cron jobs.  I've got a couple of mostly-automated tasks where scripts sit around waiting to be run with my SSH and Kerberos credentials.

One of the ideas being kicked around a while back, if I recall correctly, was an improved git-bzr that wouldn't require manual bzr commands to recover from conflicts, maybe could find and mirror all branches instead of having them explicitly named, could have a hook to guard against wrong-way merges to trunk (per Emacs dev policy), etc.  Perhaps even something that could be run on savannah as an additional access method to the bzr repo, automatically syncing with bzr as needed?  I haven't got the detailed understanding of either system at the moment to make this work, but has anyone else been thinking about working on it?

Ken