On 14/09/2012 02:35, Eric Wang wrote:
Hi Naoto,
Thank you to be my sponsor. I have checked the nightly results, this
error also occurred on Linux as well.
http://aurora.ru.oracle.com/functional/faces/RunDetails.xhtml?names=98081.CORELIBS-JDK8-NIGHTLY-JTREG-16
These links are Oracle internal links and so aren't accessible here. It
might be better to just inline the .jtr file in the mail so that others
can see the test failure too.
Anyway, from a quick read of the bug reports then this looks like an
issue in the vectorization support that was added to C2 in hs24-b21. It
looks like it doesn't reproduce with hs24-b22 but more investigation is
required. As hs24-b22 is in jdk8/jdk8 then I should we get jdk8/tl
sync'ed up to see if this issue goes away (leaving 7196199 open of
course as cause is not clear yet).
Lana - jdk8/tl is currently at jdk8-b54, do you plan to sync it up to
b56 soon?
In any case, you were right to propose adding the test to the exclude
list. I see Naoto questions "generic-all" and that is a good topic to
discuss. In this case the issue seems to be specific to C2 on x86 but
that is too fine grained for the platform matching that jtreg supports.
As I understand it, jtreg can only exclude tests that match the
operating system name and architecture. For such cases I would tend to
just add the test with "generic-all", otherwise you are into multi-lines
excludes trying to exclude everywhere except sparc in this case.
-Alan
