On Mon, Nov 05, 2001 at 03:05:23PM +0100, H.Merijn Brand wrote: > On Mon 05 Nov 2001 15:00, Jarkko Hietaniemi <[EMAIL PROTECTED]> wrote: > > > Automated smoke report for patch 12847 > > > > Fixed in 12848. Your smokes seem always to pick the most inconvenient > > spot :-) > > Blame it to time differences. It rsyncs at 22:45 local time. Be sure that > rsync snap is stable at that moment int time :-P > > I could start later though, but then my reports will be less complete, because > I *will* have to leave for home before they end. Your choice. (I'd prefer to > keep this time in order not to disturb the `normal' users too much - for > whatever degree of normal, I'm one of those too)
Hmmm. (1) How about looking for my snapshot announcements in p5p (or "patchlevel.h" updates in perl5-changes) and launching smokes from them, not from some arbitrary points in time? (2) Add a (cronjob?) suspend-all option to the smoke harness: basically, send a SIGSTOP signal for all the participating machines and process) that can be sent at 07:30am? (Coupled with a continue-all, SIGCONT, when the next 22:45 arrives). The stopped processes will consume virtual memory, not much else. (3) Some extensions of (2) would be kill-all, SIGKILL dealt over the whole smoke cluster, and shutdown-all (SIGHUP with graceful cleanup handler). -- $jhi++; # http://www.iki.fi/jhi/ # There is this special biologist word we use for 'stable'. # It is 'dead'. -- Jack Cohen