|
Boost Testing : |
From: Jason Sankey (jason_at_[hidden])
Date: 2007-09-27 03:02:20
Rene Rivera wrote:
> Jason Sankey wrote:
>> Hi all,
>>
>> The latest couple of builds on the demo Pulse server have failed to
>> complete due to what appears to be a hanging interprocess test. The
>> last output lines can be seen here:
>>
>> http://pulse.zutubi.com/tailRecipeLog.action?id=1081629&buildId=1015903
>>
>> Changes in the first hung build are here:
>>
>> http://pulse.zutubi.com/viewChanges.action?id=1015903
>>
>> I guess either change 39545 or 39546 is the cause.
>>
>> What is the most appropriate way to communicate such a failure to
>> someone who can resolve the cause? I hope in time the build server will
>> become useful enough that people will sign up for notifications and fix
>> these issues as they happen, but until then I may be the first to notice
>> a failure.
>
> We've noticed that and similar hangs recently. Which is why Noel spent
> time last week implementing the bjam time limit feature on Unix we've
> had on Windows for some time. Just add a "-l300", or more than 300
> seconds if you like, to the bjam invocation that runs the tests.
I see, and thanks for the tip! Will using this option allow other tests
to be run and their results captured, or will any tests that normally
come after interprocess be skipped? I can also set up a build timeout
in Pulse that will stop things hanging indefinitely.
There is still the question of how to get the issue resolved? This
failure seems fairly reliable: no occurrences over a couple of weeks
(several builds/day) and now I have three hung builds in a row.
Cheers,
Jason