[Buildbot-devel] Problem setting slave environment property

[sergio borghese]

Update on the issue reported: it seems that restarting the slave solved the
issue.
Not sure why, but the behavior was like some sort of config cache was in
place, so the changed configuration did not have effects...
Does it make any sense?

Best Regards
Sergio



On Sat, Aug 9, 2014 at 4:27 PM, sergio borghese <sergio.borghese at gmail.com>
wrote:

> Hi Everyone,
>
> I'm facing a really weird issue on buildbot, setting the env property.
> Strange thing is that I have some configuration where the env
> configuration is working as expected.
>
> The scenario that is showing that misbehavior is the following:
>
> buildmaster v0.8.8 running on gentoo with python 2.7
> buildlslave v.0.8.9 running on CentOs 5.10 with python 2.6
>
> the slave has the following shell command:
>
> qmake = ShellCommand(
>
> command = ["qmake"],
>
> warnOnWarnings = True,
>
> haltOnFailure = True,
>
> workdir = nr.iqac_platform_build_root,
>
> name = "qmake",
>
> env = nr.*centos510_64bit_env_dict* )
>
>
> where *centos510_64bit_env_dict *is { 'PATH':
> ["/usr/local/Trolltech/Qt-4.8.5/bin", "${PATH}"] }
>
>
> Problem is that the environment is not properly set and the shell command
> fails.
>
> From the master log I can seen:
>
>
> 2014-08-09 16:14:05+0200 [-]
> ShellCommand.startCommand(cmd=<RemoteShellCommand '['qmake']'>)
>
> 2014-08-09 16:14:05+0200 [-]   cmd.args = {'workdir':
> 'build/nr_libs/uniqlogger/lib', 'timeout': 1200, *'env': None*,
> 'want_stdout': 1, 'usePTY': 'slave-config', 'maxTime': None, 'logEnviron':
> True, 'initial_stdin': None, 'want_stderr': 1, 'logfiles': {}}
>
> 2014-08-09 16:14:05+0200 [-] <RemoteShellCommand '['qmake']'>:
> RemoteCommand.run [6]
>
> 2014-08-09 16:14:05+0200 [-] command '['qmake']' in dir
> 'build/nr_libs/uniqlogger/lib'
>
> 2014-08-09 16:14:05+0200 [Broker,0,192.168.23.230] <RemoteShellCommand
> '['qmake']'> rc=1
>
> 2014-08-09 16:14:05+0200 [-] closing log <buildbot.status.logfile.LogFile
> instance at 0x20ba488>
>
> 2014-08-09 16:14:05+0200 [-]
> releaseLocks(<buildbot.steps.shell.ShellCommand object at 0x20943d0>): []
>
> 2014-08-09 16:14:05+0200 [-]  step 'qmake' complete: failure
>
> 2014-08-09 16:14:05+0200 [-]  <Build iqac_platform_centos510_64b>: build
> finished
>
> 2014-08-09 16:14:05+0200 [-] sending mail (780 bytes) to ['
> s.borghese at netresults.it']
>
> 2014-08-09 16:14:05+0200 [-] Starting factory
> <twisted.mail.smtp.ESMTPSenderFactory instance at 0x20ba368>
>
> 2014-08-09 16:14:05+0200 [-] releaseLocks(<BuildSlave 'centos510_64bit'>):
> []
>
> 2014-08-09 16:14:05+0200 [ESMTPSender,client] Stopping factory
> <twisted.mail.smtp.ESMTPSenderFactory instance at 0x20ba368>
>
> To me the above log means that the env is not set (as if the veriable has
> not been set in the slave configuration) Of course I might be wrong, that's
> why I'm asking for some advice.
> As stated before, I already used the env configuration property with
> another slave running on MacOSX (cannot retrieve the slave version as the
> host is down right now)
>
> Any idea of what I'm doing wrong ?
> Do you think the slave version > master version can be an issue?
> The slave has been installed using easy_install and that's the version
> retrieved by the setuptools fro python 2.6
>
> Best Regards
> Sergio
>
> --
> preferisco ammazzare il tempo,
> preferisco sparare cazzate,
> preferisco fare esplodere una moda,
> preferisco morire d'amore.
> (Caparezza)
>
>


-- 
preferisco ammazzare il tempo,
preferisco sparare cazzate,
preferisco fare esplodere una moda,
preferisco morire d'amore.
(Caparezza)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://buildbot.net/pipermail/devel/attachments/20140809/d41c3fed/attachment.html>