harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mark Hindess <mark.hind...@googlemail.com>
Subject Re: [java6][M1] Test failures on the java 6 branch
Date Wed, 17 Feb 2010 14:43:48 GMT

In message <4B7BEE13.70001@googlemail.com>, Oliver Deakin writes:
> Hi all,
> Looking forward to the next attempt at getting java6 M1 out the door, I 
> ran the full test suite on Windows XP x86 and see 5 failures and 10 
> errors. I'll raise JIRAs for the ones below I think will need looking at 
> for java6 M1.
> I don't think these failures should block M1:
> javax.swing.plaf.basic.BasicScrollBarUITest - this has been failing for 
> a while now and passes standalone, also fails/passes on java5 in the 
> same way so I don't think it should block M1.
> org.apache.harmony.jpda.tests.jdwp.Events.FieldModification002Test and 
> org.apache.harmony.jpda.tests.jdwp.VirtualMachine.AllThreadsTest - both 
> timeouts. We see intermittent timeouts in the java5 branch also (even 
> though the JDWP implementation is somewhat different there) so I think 
> these are caused by timing issues in the test cases.
> org.apache.harmony.luni.tests.java.net.URLConnectionTest - fails for me 
> intermittently and on java5 also, I don't think this is a blocker.
> I think these could block M1:
> org.apache.harmony.luni.tests.java.util.TreeMapTest - 8 errors here, all 
> NullPointerExceptions. These are more worrying and I think they need to 
> be investigated for M1. [1]
> org.apache.harmony.tests.java.lang.instrument.HelloWorldTest - 3 
> failures, two with "expected:<0> but was:<3>" and the last one with 
> "AssertionFailedError: null". These fail on java5 also for me and have 
> been discussed on the dev list [2]. I think they need a JIRA to track 
> them so I've opened [3].
> Does anyone else see similar failures (particularly the TreeMap and 
> instrument ones)?

Yes.  I've reverted the change that causes the TreeMap ones in r910980.


for more information.

I've committed r910685 yesterday evening so now you should be able to
look at the output for these tests to get a better idea of the source of
the problem.

I don't think we should block on the instrument failures as they are
new tests and not regressions as such but still it would be nice to fix
them.  Oddly, the tests pass when I (accidentally) run an x86 build on
x86_64 even though the same build fails on x86.


View raw message