[wtp-releng] YEA! We finally had a unit test hang!

[David M Williams <david_williams@xxxxxxxxxx>]

Mostly kidding, but with reason. 

We finally did have a DNF unit test
after moving, last week, to new functionality in the "test runner"
we use from the Eclipse Platform that will capture a thread dump, a couple
of times, a few minutes before the maximum "timeout" value for
a test suite. Automatically. And writes it to the error log. No more waiting
for me to capture one! See bug 343448.  https://bugs.eclipse.org/bugs/show_bug.cgi?id=343448.
Or the platform's original bug 342053. https://bugs.eclipse.org/bugs/show_bug.cgi?id=342053.

So, the good news, the reason I'm happy
about this one hang, is that it proved that it works! Should greatly help
track down many of these intermittent hangs we get. 



Its not linked in main table, but fairly
easy to navigate to. The steps, from this evening's test hang: 

In the summary table, you'll see
the familiar DNF
http://build.eclipse.org/webtools/committers/wtp-R3.3.0-I/20110520234832/I-3.3.0-20110520234832/testResults.php

At the bottoms of the unit test summary
page, you'll see the usual "logs" link:

That takes you to the types of logs
available:

Click on "testSysErrorLogs"
to see what's available there ... and in that list, you can find the one
from the test that timed out.

Follow that link, and hopefully you
can make sense of the state of the "worker threads", etc., to
see what might be hanging or looping. In this case, the link is 

http://build.eclipse.org/webtools/committers/wtp-R3.3.0-I/20110520234832/I-3.3.0-20110520234832/testResults/consolelogs/testSysErrorLogs/org.eclipse.wst.common.frameworks.tests.bvt.AutomatedBVT.error.txt


Since many of these builds "disappear"
after a while, I suggest teams capture the logs in a bugzilla entry, so
if the problem is not obvious, maybe after it happens a few times it will
become more obvious or at least can be studied later after the actual build
is deleted.  

Hope it helps!