Andrew John Hughes
gnu_andrew at member.fsf.org
Wed Oct 7 02:20:16 PDT 2009
2009/10/7 Joe Darcy <Joe.Darcy at sun.com>:
> Andrew John Hughes wrote:
>> 2009/10/5 Joseph D. Darcy <Joe.Darcy at sun.com>:
> [big snip]
>>> Any results for the jdk/test regression tests?
>> JDK: Test results: passed: 3,269; failed: 168; error: 3
>> The failed seems high, the number of errors matches our usual JTreg
>> runs. I'll have to take a closer look but I don't usually do runs on
>> this box. What were your own results like for OpenJDK6 b17?
> My results for building OpenJDK 6 on Ubuntu 9.04 x64 and running tests there
> Test results: passed: 24
> Test results: passed: 1354
> Test results: passed: 3,100; failed: 43; error: 1
> All test suites were run with "-a -ignore:quiet"; in addition, langtools was
> run with "-s".
> I've uploaded my summary.txt files to:
> (The jdk tests were run a bit differently than my usual test setup where I
> point the graphical tests into a virtual framefer. In this case, I just ran
> the tests without redirecting the graphics.)
> I haven't looked at the IcedTea patch set recently; if the Gervill sound
> tests are still being duplicated, I think the IcedTea copies can be removed
> at this point since Karl is keeping the OpenJDK 6 copy of the code and tests
Ok, the majority of JDK failures I'm seeing are not an issue; they are
either compile failures from the optional NIO2 backport in IcedTea6 or
headless exceptions for the AWT/Swing stuff. The former are known to
fail, as I just haven't had time to backport the tests. This isn't a
major issue as most IcedTea6 builds don't enable the backport anyway.
The headless exceptions are also expected given the tests were run in
screen without any virtual framebuffer. Perhaps we should have the
check-jdk target use Xvfb by default?
Two Gervill tests fail:
javax/sound/midi/Gervill/SoftReceiver/GetMidiDevice.java (compile failure)
javax/sound/midi/Gervill/SoftChannel/NoteOverFlowTest2.java : IndexOutOfBounds
The rest seem to be related to java.beans (can't find java.awt.Color
-- possibly a headless issue again) or crypto (having ECC enabled
seems to cause additional failures).
I don't think there's anything that's a stopper for b17, especially as
your results seem pretty good (and the failures seem the same as we've
seen). We just need to make some decisions on how to resolve the
differences between the IcedTea6 and plain OpenJDK6 builds.
What do you think?
Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)
Support Free Java!
Contribute to GNU Classpath and the OpenJDK
PGP Key: 94EFD9D8 (http://subkeys.pgp.net)
Fingerprint: F8EF F1EA 401E 2E60 15FA 7927 142C 2591 94EF D9D8
More information about the jdk6-dev