On 12/25/06, Gregory Shimansky <[EMAIL PROTECTED]> wrote:
Elena Semukhina wrote: > On 12/22/06, Elena Semukhina <[EMAIL PROTECTED]> wrote: >> >> >> >> On 12/22/06, Gregory Shimansky <[EMAIL PROTECTED]> wrote: >> > >> > On Thursday 21 December 2006 17:23 Elena Semukhina wrote: >> > > On 12/21/06, Gregory Shimansky < [EMAIL PROTECTED]> wrote: >> > > > Evgueni Brevnov wrote: >> > > > > On 12/21/06, Elena Semukhina <[EMAIL PROTECTED]> wrote: >> > > > >> On 12/21/06, Evgueni Brevnov < [EMAIL PROTECTED]> wrote: >> > > > >> > Naveen, >> > > > >> > >> > > > >> > It is true, HARMONY-2817 is not intended to fix stress.Mix. >> > There >> > > > >> > are two different issues with stress.Sync & stress.Mix. I'm >> > looking >> > > > >> > at stress.Sync problem only. It will be good if someone else >> > takes >> > > > >> > care about stress.Mix. >> > > > >> >> > > > >> stress.Mix always passes for me on SUSE9 linux ia32. >> > > > >> On SUSE9 linux em64t it started to fail with Segmentation fault >> > two >> > > > >> > > > days >> > > > >> > > > >> ago. HARMONY-2772 has been filed to track this issue. >> > > > > >> > > > > Hi Elena, >> > > > > >> > > > > Thank you for the patch evaluation. I've updated HARMONY-2817 >> with >> > a >> > > > > new patch. Could you please test it and let us know how it is >> > > > > doing.... >> > > > >> > > > I've tried the patch and left comment in JIRA. It looks like it >> > still >> > > > needs more work. >> > > >> > > While the issue fixing is in progress, I suggest that we exclude >> > > stress.Syncfrom testing. Otherwise CC could report it failure as a >> > > false alarm. >> > > >> > > I created >> > > https://issues.apache.org/jira/browse/HARMONY-2838 to exclude this >> > tests on >> > > JIT. >> > > >> > > The patch also excludes 2 more test whose hangs I observe rarely for >> > quite >> > > a long time. >> > >> > It looks like Evgueni has finally created a fix in HARMONY-2817 which >> > doesn't >> > break other things. The tests are running ok for me and I'll probably >> > commit >> > the patch when the tests are done. So I would rather not exclude any >> > tests >> > right now. >> >> >> >> OK, let's see how other two tests will behave after HARMONY-2817 is >> applied. Possibly this patch will fix their problems as well. >> > > > Gregory, > > HARMONY-2817 seems to fix stress.Sync but I still observe rare gc.LOSand > gc.MultiThreads hangs on linux JIT. Could you apply HARMONY-2838 to exclude > these tests? This is strange because for me tests on linux run ok. And I run them a lot because I test every commit. Anyway it could be the configuration difference, like the number of CPUs. I'll exclude the tests now, but I think the idea of having a local CC exclude list in the other thread is a good one. This way CC will be stable while we'll still see test failures and will work on fixing them.
Gregory, I caught the stress.Sync hang in the interpreter mode and attached the stack traces to HARMONY-2817. Elena --
Gregory
-- Thanks, Elena
