New submission from Bryan O'Sullivan <bos@xxxxxxxxxxxxxx>:
I mirror some Mercurial repositories at my web site. When a big snowstorm blew
through Germany last week, Thomas's web servers died for a few days.
Apparently, one of them died in the middle of my cron job doing a pull. The
connection never timed out, and the cron job remained stuck.
Instead of making progress or failing, every subsequent cron job also got stuck
because it blocked waiting for the stuck repository to become unlocked. By this
morning, I had about 75 cron jobs stuck, none of which had notified me of
problems because they were all "still running".
It would be a big help if the locking code used a command line flag to control
its behaviour in the face of a locked repository. Right now, it loops forever
unconditionally. One alternative would be to simply fail if the repo was
locked; another would be to add a loop count, which if exceeded would result in
failure.
----------
messages: 238
nosy: bos
priority: feature
status: unread
title: add flag to make hg fail if repo or dirstate locked
____________________________________________________
Mercurial issue tracker <mercurial-bugs@xxxxxxxxxxx>
<http://www.selenic.com/mercurial/bts/issue79>
____________________________________________________
|
