On Fri, Nov 08, 2013 at 12:03:41PM +0000, Frank Shearar wrote:
> On 6 November 2013 21:44, Frank Shearar <frank.shearar at gmail.com> wrote:
> > On 6 November 2013 20:15, Frank Shearar <frank.shearar at gmail.com> wrote:
> >> On 5 November 2013 07:01, Frank Shearar <frank.shearar at gmail.com> wrote:
> >>> On 4 November 2013 18:40,  <commits at source.squeak.org> wrote:
> >>>> Frank Shearar uploaded a new version of Tests to project The Trunk:
> >>>> http://source.squeak.org/trunk/Tests-fbs.261.mcz
> >>>>
> >>>> ==================== Summary ====================
> >>>>
> >>>> Name: Tests-fbs.261
> >>>> Author: fbs
> >>>> Time: 4 November 2013, 6:40:42.33 pm
> >>>> UUID: f5961117-ed1d-7743-ad50-ef8c0cca4ab7
> >>>> Ancestors: Tests-fbs.260
> >>>>
> >>>> Delete testInfoProxy as per Chris Muller's request, to see if it unbreaks CI.
> >>>>
> >>>> =============== Diff against Tests-fbs.260 ===============
> >>>
> >>>
> >>> This didn't work - CI was still broken - but Chris' idea of logging
> >>> out every test makes me suspect it's
> >>> MCMethodDefinitionTest>>#testRevertOverrideMethod. If you look at
> >>> http://build.squeak.org/job/SqueakTrunk/lastSuccessfulBuild/console
> >>> it's the very last test before the script commits suidice:
> >>>
> >>> 2013-11-04T22:41:48.317+01:00:
> >>> MCMethodDefinitionTest>>#testPartiallyRevertOverrideMethod
> >>> 2013-11-04T22:41:53.698+01:00: MCMethodDefinitionTest>>#testRevertOldMethod
> >>> 2013-11-04T22:41:59.078+01:00: MCMethodDefinitionTest>>#testRevertOverrideMethod
> >>> !!! Killed command 0 for exceeding allotted time: nice
> >>> /home/jenkins/workspace/SqueakTrunk/target/cog.r2776/coglinux/bin/squeak
> >>> -vm-sound-null -vm-display-null
> >>> "/home/jenkins/workspace/SqueakTrunk/target/PostTestTrunkImage.image"
> >>> ../tests.st.
> >>> /home/jenkins/workspace/SqueakTrunk/benchmarks.st
> >>> spawning command 1 with timeout 600 seconds: nice
> >>> /home/jenkins/workspace/SqueakTrunk/target/cog.r2776/coglinux/bin/squeak
> >>> -vm-sound-null -vm-display-null
> >>> "/home/jenkins/workspace/SqueakTrunk/target/PostTestTrunkImage.image"
> >>> ../benchmarks.st
> >>> 2013-11-04T23:22:12.32+01:00: MCOrganizationTest>>#testReordering
> >>> 2013-11-04T23:22:12.325+01:00:
> >>> MCOrganizationTest>>#testReorderingWithNoCategoriesInVersion
> >>
> >> Interpreter VM, no hang:
> >> http://build.squeak.org/job/SqueakTrunkOnInterpreter/314/console
> >> Cog r.2776 VM, hang: http://build.squeak.org/job/SqueakTrunk/578/consoleFull
> >>
> >> Same machine, a 32 bit Linux (Ubuntu). Is it possible that we're
> >> seeing some Cog-specific problem? We're _not_ seeing something like a
> >> segfaulting image. The image just seems to get stuck.
> >
> > Another datum. Interpreter VM on 64 bit Linux (Ubuntu?), completes
> > test suite: https://travis-ci.org/squeak-smalltalk/squeak-ci/builds/13602324
> 
> http://build.squeak.org/job/SqueakTrunk/595/console shows a build,
> using Cog, _succeeding_ on a Linux machine, build.squeak.org itself. I
> can't run jobs at the moment on the node showing the failure, but I'll
> rerun the jobs tonight.
> 

Regarding the Jenkins CogVM job, I'm happy to see that it moved to "green"
for a couple of days. Green is good, but I do want to note that red is not
necessarily anything to worry about on these VM code generator jobs (CogVM
and InterpreterVM). Green means that the respective VMMaker and Subversion
platform sources are nicely in sync and that you can compile a possibly
working VM from the generated sources.

I also put this important note in the Jenkins CogVM job:

  The Cog VM is in a period of active development for Spur. This Jenkins
  project is not a good indication of the state of that development. Please
  refer to Eliot's blog and to the vm-dev list for updates.

Right now both the InterpreterVM and CogVM jobs are failing due to some
issue in loading VMMaker into the headless image. It's probably just
some sort of warning notifier or request for author intials popping up
in the headless image. I won't be able to look into it for another week
or two, but if anyone is interested in taking a look at it, all you need
to do is start with a fresh image and file in the scripts from the
Jenkins workspaces.

For Cog:

  http://build.squeak.org/job/CogVM/ws/VMCogUnixBuild.st

For Interpreter VM:

  http://build.squeak.org/job/InterpreterVM/ws/VMUnixBuild.st

Have fun,

Dave

> (Turns out it's a bad idea to nice(1) the running of the tests because
> then you lose the PID of the Squeak process: Ruby's spawn method
> returns the PID of the nice(1) process. That's why I could see no
> output, presumably.)
> 
> frank