geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From David Blevins <david.blev...@visi.com>
Subject Re: Unit/Stress Tests (Bite Sized Task)
Date Wed, 04 Aug 2004 22:11:36 GMT
On Wed, Aug 04, 2004 at 05:01:57PM -0500, Dain Sundstrom wrote:
> On Aug 3, 2004, at 2:09 PM, Dain Sundstrom wrote:
> 
> >>>That's basically a 20 minute difference with only 4.5 minutes
> >>>reportedly spend on our tests.  What is taking up the other 15
> >>>minutes?  Is setUp/tearDown not counted in the test time?
> >>
> >>If not, the zillions of kernels we create in setUp methods could 
> >>easily explain the difference.
> >
> >I would wager that the 15 minutes is spent launching the java vms.  It 
> >looks like on my machine that there is one vm launched for each unit 
> >test.  On a very small module, my test time with fork on was 30s and 
> >with fork off was 10s.
> 
> It occurs to me this could be just like the setContextClassLoader 
> permission issue in that we may be forking the test vm for no reason.
> 
> Can someone try to run the unit tests with "maven.junit.fork=true" 
> commented out in the etc/global.properties file?  I'm interested if 
> only a few tests are failing, or is most tests fail for a single simple 
> reason.
> 
> If we can fix this, we can all be more productive.

Gave it a shot yesterday.  The
org.apache.geronimo.system.url.GeronimoURLFactoryTest tests error out
and I couldn't get the build to continue even with the
maven.test.failure.ignore=true flag.

Maven2 uses separate classloaders for running each testcase so forking
is largely unneeded.  Things that mess with the VM itself though
(URLFactories, SecurityManagers, etc.) would probably still need to be
forked.

For now, we could go module by module and see if we can get find some
that don't need to be forked.

-David

Mime
View raw message