On 30/07/2012, at 7:34 AM, Adam Murdoch wrote: > > On 27/07/2012, at 9:05 PM, Luke Daley wrote: > >> >> On 27/07/2012, at 11:59 AM, Szczepan Faber wrote: >> >>> >>> I'm wondering if we might have a problem where two daemons can be started >>> to run a build under certain circumstances. I don't have another >>> explanation for the failure, other than the test fixture being bugged. >>> >>> There might be a bug so that we start a daemon but for whatever reason it >>> didn't except the build request (for example: issue with remoting, daemon >>> says I'm busy, etc.). So the implementation starts a new daemon and try to >>> pass him the build request. I've seen this issue with this test before. >>> This symptom can also be caused by our daemon test harness. >> >> That doesn't explain it. If you look at the output you see that both daemons >> went busy, and both stopped when the stop sentinel is activated. This >> implies that two daemons were running the build concurrently. > > Could be interference from another test, where we're accidentally sharing a > home dir with another test.
That's the most plausible explanation, but I'd expect that this happened more often in that case. We had some weirdness with daemons due to recent changes that has been cleared up. Let's hope that makes this problem go away. -- Luke Daley Principal Engineer, Gradleware http://gradleware.com
