A probable problem is in finalizer mechanism, see HARMONY-3590<https://issues.apache.org/jira/browse/HARMONY-3590>. This issue fixes finalization work balance subsystem. It's affected stability very hard.
[1] https://issues.apache.org/jira/browse/HARMONY-3590 On 06/04/07, Pavel Rebriy <[EMAIL PROTECTED]> wrote:
I'm checking Linux x86_64 now. On 06/04/07, Pavel Rebriy <[EMAIL PROTECTED]> wrote: > > I checked gc.SynchronizedFinilazersTest test on Linux ia32 after commit > HARMONY-3288 revision 520478 and it passes 1000 times. > > On 06/04/07, Vladimir Ivanov <[EMAIL PROTECTED]> wrote: > > > > On 4/5/07, Alexey Varlamov < [EMAIL PROTECTED]> wrote: > > > 2007/4/5, Vladimir Ivanov < [EMAIL PROTECTED]>: > > > > OK, today I run smoke test gc.SynchronizedFinilazersTest on SUSE 9 > > > > Linux x86_64 in cycle. > > > > I found: > > > > - on current DRLVM this test hang on 13 iteration; > > > > - on r522000 test passed for 100 iteration; > > > > - on r522100 test hang on 53 iteration. > > > > > > > > The differences between 522100 and 522000 are: > > > > U vm/include/open/hythread_ext.h > > > > U vm/interpreter/src/interpreter.cpp > > > > U vm/thread/src/thread_java_monitors.c > > > > U vm/thread/src/hythr.exp > > > > U vm/thread/src/hythr.def > > > > U vm/thread/src/thread_native_fat_monitor.c > > > > U vm/thread/src/thread_native_basic.c > > > > U vm/thread/src/thread_ti_monitors.c > > > > U vm/thread/src/thread_native_tls.c > > > > U vm/thread/src/thread_private.h > > > > U vm/thread/src/thread_java_basic.c > > > > U vm/thread/src/thread_native_suspend.c > > > > U vm/thread/src/thread_init.c > > > > U vm/vmcore/src/thread/thread_manager.cpp > > > > U vm/vmcore/src/jvmti/jvmti_pop_frame.cpp > > > > U vm/vmcore/src/jvmti/jvmti_step.cpp > > > > U vm/vmcore/src/jvmti/jvmti_break.cpp > > > > U vm/jitrino/src/jet/compiler.cpp > > > > U vm/jitrino/src/vm/drl/DrlVMInterface.cpp > > > > > > > > Now I run this test on r522000 for 3000 iteration (400 already > > > > passed). If it passed I'll try to narrow the guilty changes. > > > > > > > The test passed on r522000 for 3000 iteration. > > > > > AFAIS in this interval was the sole drlvm commit: > > > [r522033] H3413, refactor, cleanup of exception safepoint callback > > build. > > > > Thanks. The result is: > > - the test is hang on the r522033 (iteration 143) > > - the test is passed already for 670 iteration on r522032. > > > > Could somebody from our thread guru evaluate it? > > > > thanks, Vladimir > > > > > > > > > > > > > > > thanks, Vladimir > > > > > > > > > > > > On 4/4/07, Vladimir Ivanov <[EMAIL PROTECTED] > wrote: > > > > > The test gc.SynchronizedFinilazersTest still hangs up on SUSE 9 > > Linux x86_64. > > > > > OK, I'll exclude it locally. > > > > > > > > > > thanks, Vladimir > > > > > > > > > > > > > > > On 3/30/07, Pavel Rebriy < [EMAIL PROTECTED]> wrote: > > > > > > I had a similar problem with WinX64 (see > > "[drlvm][stability][winx64] > > > > > > finalization tests irregular failure" branch for details). > > > > > > I had a problem with HARMONY-3485 commit. I rolled back it, > > and it was > > > > > > successful for tests passing :) > > > > > > I could solve the problem with smoke tests. > > > > > > > > > > > > On 30/03/07, Elena Semukhina <[EMAIL PROTECTED]> > > wrote: > > > > > > > > > > > > > > On 3/29/07, Peter Novodvorsky < [EMAIL PROTECTED]> > > wrote: > > > > > > > > > > > > > > > > HARMONY-3495 shows that gc.NPE fails in the JVM shutdown > > stage. Do you > > > > > > > > think that all other tests fail in the same situation? Can > > you > > > > > > > > explain, what makes you think that all these tests fail > > for the same > > > > > > > > reason? > > > > > > > > > > > > > > > > > > > > > It looks like the gc.PhantomReferenceQueueTest hangs at the > > shutdown stage > > > > > > > as well: it prints PASSED and then hangs with the same stack > > traces. > > > > > > > gc.SynchronizedFinilazersTest ( > > > > > > > https://issues.apache.org/jira/browse/HARMONY-3517) shows up > > another stack > > > > > > > traces. > > > > > > > > > > > > > > I'm not sure that the cause is the same for both issues. I'm > > only almost > > > > > > > sure that we have these troubles since last weekend. > > > > > > > > > > > > > > Thanks, > > > > > > > Elena > > > > > > > > > > > > > > Thanks, > > > > > > > > Peter. > > > > > > > > > > > > > > > > > > > > > > > > On 3/29/07, Elena Semukhina <[EMAIL PROTECTED] > > > wrote: > > > > > > > > > I managed to reproduce gc.PhantomReferenceQueueTest hang > > on SUSE 9 > > > > > > > Linux > > > > > > > > > x86_64 as well. > > > > > > > > > > > > > > > > > > It seems that we have new serious problem in threading > > system since > > > > > > > last > > > > > > > > > weekend. A number of test can hang intermittently. Each > > day we see new > > > > > > > > > test's hang. > > > > > > > > > > > > > > > > > > Shouldn't we focus efforts on investigating this issue? > > From my > > > > > > > > experiens > > > > > > > > > gc.NPE hangs more often than others and could be a good > > candidate for > > > > > > > > issue > > > > > > > > > reproducing (https://issues.apache.org/jira/browse/HARMONY-3495 > > ) > > > > > > > > > > > > > > > > > > Elena > > > > > > > > > > > > > > > > > > On 3/29/07, Vladimir Ivanov < [EMAIL PROTECTED]> > > wrote: > > > > > > > > > > > > > > > > > > > > Actually not Linux x86 only :( The test gc.NPE hangs > > on Linux x86_64 > > > > > > > > also. > > > > > > > > > > Today the test gc.PhantomReferenceQueueTest hangs on > > SUSE 9 Linux > > > > > > > > x86_64. > > > > > > > > > > Note, all these hangs are intermittent. > > > > > > > > > > > > > > > > > > > > Thanks, Vladimir > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > On 3/28/07, Elena Semukhina <[EMAIL PROTECTED]> > > wrote: > > > > > > > > > > > This week I noticed that some smoke tests hung in > > iterative runs > > > > > > > on > > > > > > > > > > linux > > > > > > > > > > > x86. I did not see this before. Possibly they detect > > some new > > > > > > > > issues. > > > > > > > > > > > > > > > > > > > > > > Yesterday I filed > > > > > > > > https://issues.apache.org/jira/browse/HARMONY-3495about > > > > > > > > > > > gc.NPE because CC also detected the hang. We > > excluded gc.NPE. > > > > > > > > > > > > > > > > > > > > > > Today I saw that gc.SynchronizedFinilazersTest hung > > and filed > > > > > > > > > > > https://issues.apache.org/jira/browse/HARMONY-3517. > > > > > > > > > > > > > > > > > > > > > > I wouldn't like to exclude each test which hangs for > > now. Possibly > > > > > > > > we > > > > > > > > > > can > > > > > > > > > > > fix the issue quickly. > > > > > > > > > > > > > > > > > > > > > > I managed to reproduce gc.SynchronizedFinilazersTesthang after > > > > > > > 450 > > > > > > > > runs > > > > > > > > > > in > > > > > > > > > > > JET and after 50 runs in OPT. I provided te stack > > traces in JIRAs. > > > > > > > > > > Possibly > > > > > > > > > > > they could be useful. > > > > > > > > > > > > > > > > > > > > > > -- > > > > > > > > > > > Thanks, > > > > > > > > > > > Elena > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > -- > > > > > > > > > Thanks, > > > > > > > > > Elena > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > -- > > > > > > > Thanks, > > > > > > > Elena > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > -- > > > > > > Best regards, > > > > > > Pavel Rebriy > > > > > > > > > > > > > > > > > > > > > > > > -- > Best regards, > Pavel Rebriy > -- Best regards, Pavel Rebriy
-- Best regards, Pavel Rebriy
