[Box-Admins] Re: [Build server] Problem with Universes test?

[Frank Shearar]

I mean, obviously it's my fault. In this case, Universes depends on
XML-Parser and Nebraska, and I've just stripped those out of the Trunk
core image. Sigh. The fix is to construct an Installer script for
Universes. The proper job is to make a SqueakMap entry with this
Installer script (or go the Metacello route). But one step at a
time...

frank

On 1 April 2013 21:26, Frank Shearar <frank.shearar at gmail.com> wrote:
> On 1 April 2013 21:21, Ken Causey <ken at kencausey.com> wrote:
>> On 04/01/2013 03:13 PM, Frank Shearar wrote:
>>>
>>> Hi Ken,
>>>
>>> Thanks. I'm hoping that once I've figured out the latest issue with CI
>>> this will go away. Jenkins aborts the build, but since the build kicks
>>> off the actual tests in separate child processes, it looks like those
>>> aren't killed when the build process is killed.
>>>
>>> frank
>>>
>>> On 1 April 2013 21:04, Ken Causey<ken at kencausey.com>  wrote:
>>>>
>>>> There were 4 stuck build processes that all appear to be related to
>>>> testing
>>>> Universes, I think.  Here is an example:
>>>>
>>>> jenkins  27839  0.0  0.0   1748   492 ?        S    Mar30   0:00 sh -c
>>>> nice
>>>>
>>>> /var/lib/jenkins/workspace/SqueakTrunk/target/cog.r2701/coglinux/bin/squeak
>>>> -vm-sound-null -vm-display-null
>>>> "/var/lib/jenkins/workspace/SqueakTrunk/target/Universes.image"
>>>> ../package-load-tests/Universes.st
>>>> jenkins  27841  2.2  2.3 1054528 24096 ?       RN   Mar30  60:37
>>>>
>>>> /var/lib/jenkins/workspace/SqueakTrunk/target/cog.r2701/coglinux/bin/../lib/squeak/4.0-2701/squeak
>>>> -vm-sound-null -vm-display-null
>>>> /var/lib/jenkins/workspace/SqueakTrunk/target/Universes.image
>>>> ../package-load-tests/Universes.st
>>>>
>>>> I killed them.
>>>>
>>>> Ken
>>>
>>>
>>
>> I spoke too soon.  There are 2 more from today.  They are about 22 hours old
>> so I'm going to go ahead and kill them.  I wondered at first why there were
>> 2 that started at almost the same time (4 minutes apart) but it appears
>> these tests are being done both with an interpreter VM and Cog.  Correct?
>
> Yes, that's exactly right. So until I figure out the problem I guess
> we'll be looking at two hung processes per SqueakTrunk run.
>
> frank
>
>> Ken