On Fri, Sep 21, 2007 at 03:42:25PM +0200, Emmanuel Venisse wrote: > you use the 1.1-beta-3 sent to the vote, right? not a previeous snapshot?
Dooh, my bad. I mixed up the numbers. Please replace beta-3 by beta-2 below. I was originally using beta-1 when I reported the issue, then upgraded to beta-2. You guys are releasing faster than I can upgrade ;) I can try to test beta-3 but as the bug only happens about once a week now, it will take about two weeks to be confident that it is gone for good. Is there a place where I can download it or should I build it from SVN ? > I'll try to reproduce it a new time :( Don't bother yet. Give me a couple of weeks to try beta-3. I'll ring you back if the issue it still present. -- Julien > Emmanuel > > Julien Stern a écrit : > >On Fri, Sep 21, 2007 at 03:13:55PM +0200, Emmanuel Venisse wrote: > >>I fixed it 3 days ago, before to start the vote of 1.1-beta-3 release. > > > >Yes, I saw that. It took me some time to actually test beta-3. > >I originally though the bug was fixed too, as it did not happen > >for 4 or 5 days. It really seem to happen rarely now. > > > >What is mostly annoying with this bug is when it happens on a "stable" > >project, or worse on a "parent pom". Because the project needs to > >be rebuilt manually, it forces the rebuild of many projects. > > > >What is really weird is that you could not reproduce the bug even > >with very short schedule. Maybe the fact that we are using CVS > >(and not SVN) as the SCM could be the difference ? > > > >>For 1.1-beta-3, do you use a db from 1.1-beta-2, a new db or a db created > >>by the migration tool? > > > >A new DB. > > > >-- > >Julien > > > >>Emmanuel > >> > >>Julien Stern a écrit : > >>>Emmanuel, > >>> > >>>just to let you know, that unfortunately, the issue > >>>http://jira.codehaus.org/browse/CONTINUUM-1384 > >>>still happens with beta-3, albeit much less often than with beta-2. > >>> > >>>It happened twice in the last two weeks. > >>>In both cases there was nothing in the logs except the fact that > >>>the cvs update for the failing project and the schedule activator > >>>were essentially triggered at the same time. > >>> > >>>-- > >>>Julien > >>> > >>>On Thu, Aug 30, 2007 at 02:38:00PM +0200, Emmanuel Venisse wrote: > >>>>Julien Stern a écrit : > >>>>>On Thu, Aug 30, 2007 at 01:36:38PM +0200, Emmanuel Venisse wrote: > >>>>>>Julien Stern a écrit : > >>>>>>>On Thu, Aug 30, 2007 at 03:23:13AM -0700, Ionut S wrote: > >>>>>>>>Hi, > >>>>>>>>I don't know for what reason one of our projects (shown when you > >>>>>>>>hit "Show projects" link) show as "Build in error" although the > >>>>>>>>last 10 or so builds are successful. > >>>>>>>> > >>>>>>>>Did anybody had this problem ? > >>>>>>>Yes, this happens to us about twice a day with 1.1 > >>>>>>>It may be related to http://jira.codehaus.org/browse/CONTINUUM-1384 > >>>>>>> > >>>>>>>You may want to check your logs to see if this is the same issue > >>>>>>>or a different one. > >>>>>Emmanuel, > >>>>> > >>>>>>Do you have it with beta-2 too? > >>>>>sorry, I haven't installed beta-2 yet. I will install it and let > >>>>>you know. > >>>>> > >>>>>>What is your default schedule? > >>>>>Our default schedule is short (15 minutes) with many projects (> 200). > >>>>>That's probably why we trigger this issue often. > >>>>yes. > >>>> > >>>>>>What is the max path length for your files? > >>>>>I'm not sure I understand this question. > >>>>>If you mean the max length of the absolute unix file path from the cvs > >>>>>root, I'd hazard a guess as between 128 and 256. > >>>>it's what I meant. > >>>> > >>>>>I've taken a very quick look at the continuum code. I would tend to > >>>>>think that this issue may be related to a project state that is not > >>>>>set back > >>>>>to the correct value because two concurrent threads try to modify it > >>>>>(the thread that performs the scm update and the thread that queues the > >>>>>builds ?) > >>>>> > >>>>>Maybe (just maybe, really :) ) in the class > >>>>>UpdateWorkingDirectoryFromScmContinuumAction.java the latest > >>>>>store.updateProject( project ) could fail, leaving an inconsistent > >>>>>state ? > >>>>if the problem always exist in beta-2, yes, it should be the right > >>>>place to fix it. > >>>> > >>>>>Again, this is a wild guess that could be totally wrong, the continuum > >>>>>code is too complex to be understood after a quick look. > >>>>The code isn't too complex as you find where we can have an error about > >>>>project state ;) > >>>> > >>>>Emmanuel > >>>> > >>>> > >>> > > > > >
