> -----Original Message-----
> From: Ivan Zhakov [mailto:[email protected]]
> Sent: dinsdag 1 september 2015 11:44
> To: Bert Huijben <[email protected]>
> Cc: [email protected]; [email protected]
> Subject: Re: svn commit: r1700439 - in /serf/tools/buildbot/windows: serf-
> config.cmd.template serftest-build.cmd serftest-cleanup.cmd serftest-
> template.cmd serftest-test.cmd
>
> On 1 September 2015 at 12:39, Bert Huijben <[email protected]> wrote:
> >> -----Original Message-----
> >> From: Ivan Zhakov [mailto:[email protected]]
> >> Sent: dinsdag 1 september 2015 11:37
> >> To: [email protected]
> >> Cc: [email protected]
> >> Subject: Re: svn commit: r1700439 - in /serf/tools/buildbot/windows: serf-
> >> config.cmd.template serftest-build.cmd serftest-cleanup.cmd serftest-
> >> template.cmd serftest-test.cmd
> >
> >
> >> > +taskkill /im msbuild.exe /f 2> nul:
> >> > +taskkill /im svn.exe /f 2> nul:
> >> Is it really needed for serf tests to run?
> >
> > This buildbot runs on the same hardware and as the same user as the
> > svn-windows-ra buildbot, so to guarantee stable runs I think we should
> > just apply the same kills from both runs.
> >
> But test will fail anyway if serf test kill svn.exe while svn tests
> are running. Did I miss something?
As answered on IRC:
Yes
* All builds on this slave run sequential, so no other buildbot task will run
at the same time
* The other buildbot (svn-windows-local) runs as a different -not
administrator- user, so the taskkill commands won't apply to that bot.
The problem on the Windows bot was that sometimes the python scripts timeout
while some subversion command still hangs (and sometimes takes 100% of a CPU
core), these commands which are just copied from there just improve the chance
on a OK build after some severely broken previous build.
But please don't apply these scripts as generic good scripts on a separate
bot.... :-)
Bert
