[Buildbot-commits] [Buildbot] #1001: start_instance on a latent BuildSlave called before stop_instance finished

Buildbot buildbot-devel at lists.sourceforge.net
Mon Oct 25 22:04:16 UTC 2010


#1001: start_instance on a latent BuildSlave called before stop_instance finished
--------------------+-------------------------------------------------------
Reporter:  moschny  |       Owner:                
    Type:  defect   |      Status:  new           
Priority:  major    |   Milestone:  0.8.3         
 Version:  master   |    Keywords:  virtualization
--------------------+-------------------------------------------------------
Changes (by dustin):

  * milestone:  0.8.2 => 0.8.3


Comment:

 I think this is the central problem:

 {{{
     def buildFinished(self, build, sb, bids):
         """This is called when the Build has finished (either success or
         failure). Any exceptions during the build are reported with
         results=FAILURE, not with an errback."""

         # by the time we get here, the Build has already released the
 slave
         # (which queues a call to maybeStartBuild)
 }}}

 which is to say, by the time buildFinished is called, maybeStartBuild has
 already been called.  I'm assuming from your description that your
 subclass is not somehow preventing this maybeStartBuild invocation from
 triggering a build?

 What if you adjusted the implementation so that it only performed one
 build per instantiation, and refused to start anything after that?  Then
 nothing will successfully start until the insubstantiate and subsequent
 substantiate calls complete.

 Bumping this to 0.8.3 since it appears to be specific to your case, rather
 than a general blocker.

-- 
Ticket URL: <http://buildbot.net/trac/ticket/1001#comment:4>
Buildbot <http://buildbot.net/>
Buildbot: build/test automation


More information about the Commits mailing list