You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@harmony.apache.org by "Geir Magnusson Jr." <ge...@pobox.com> on 2006/11/21 17:55:49 UTC
[drlvm][build] exclude files for kernel tests
I just checked in a change to the drlvm test system that uses external
exclude files, delineated by platform. (I just realized that I need to
also separate by jit/opt/int as well, as we have that distinction on
x86_64 linux. I'll do that in a sec)
The idea is that in make/excludes are files for each os/arch and soon
jit. It uses the same nomenclature as in classlib, and in fact the
build now 'reaches over' and imports the properties.xml file from
classlib build so we don't have more than open copy of that code around.
Note that these tokens are different than what we'd using in DRLVM
traditionally, but we need to converge and this is a first, harmless
step. I'll be slowly trying to switch DRLVM over to that token set for
arch and OS.
I've tested on linux x86 and now testing on x86_64. if someone could
simply give it a whirl on windows, I'd be mighty obliged
geir
Re: [drlvm][build] exclude files for kernel tests
Posted by Elena Semukhina <el...@gmail.com>.
On 11/22/06, Geir Magnusson Jr. <ge...@pobox.com> wrote:
>
> ok - I'm fixed this to have a set of excludes for opt, jet and int - on
> x86_64, they turn out to be different.
Geir,
x86_64 exclude lists differ from ones for other platforms: there are some
more excluded tests there. Are there JIRA issues which could be matched to
those failures? I'd like to add this info to the
http://wiki.apache.org/harmony/DRLVMInternalTests page.
Thanks,
Elena
So now, you can run linux x86_64 in the no-fork mode, and it runs to
> completion.
>
> On to "no regressions" :)
>
> I'm going to look at the smoke tests now and try to do the same thing.
>
> geir
>
>
> Geir Magnusson Jr. wrote:
> > I just checked in a change to the drlvm test system that uses external
> > exclude files, delineated by platform. (I just realized that I need to
> > also separate by jit/opt/int as well, as we have that distinction on
> > x86_64 linux. I'll do that in a sec)
> >
> > The idea is that in make/excludes are files for each os/arch and soon
> > jit. It uses the same nomenclature as in classlib, and in fact the
> > build now 'reaches over' and imports the properties.xml file from
> > classlib build so we don't have more than open copy of that code around.
> >
> > Note that these tokens are different than what we'd using in DRLVM
> > traditionally, but we need to converge and this is a first, harmless
> > step. I'll be slowly trying to switch DRLVM over to that token set for
> > arch and OS.
> >
> > I've tested on linux x86 and now testing on x86_64. if someone could
> > simply give it a whirl on windows, I'd be mighty obliged
> >
> > geir
>
--
Thanks,
Elena
Re: [drlvm][build] exclude files for kernel tests
Posted by "Geir Magnusson Jr." <ge...@pobox.com>.
ok - I'm fixed this to have a set of excludes for opt, jet and int - on
x86_64, they turn out to be different.
So now, you can run linux x86_64 in the no-fork mode, and it runs to
completion.
On to "no regressions" :)
I'm going to look at the smoke tests now and try to do the same thing.
geir
Geir Magnusson Jr. wrote:
> I just checked in a change to the drlvm test system that uses external
> exclude files, delineated by platform. (I just realized that I need to
> also separate by jit/opt/int as well, as we have that distinction on
> x86_64 linux. I'll do that in a sec)
>
> The idea is that in make/excludes are files for each os/arch and soon
> jit. It uses the same nomenclature as in classlib, and in fact the
> build now 'reaches over' and imports the properties.xml file from
> classlib build so we don't have more than open copy of that code around.
>
> Note that these tokens are different than what we'd using in DRLVM
> traditionally, but we need to converge and this is a first, harmless
> step. I'll be slowly trying to switch DRLVM over to that token set for
> arch and OS.
>
> I've tested on linux x86 and now testing on x86_64. if someone could
> simply give it a whirl on windows, I'd be mighty obliged
>
> geir
Re: [drlvm][build] exclude files for kernel tests
Posted by Elena Semukhina <el...@gmail.com>.
On 11/24/06, Elena Semukhina <el...@gmail.com> wrote:
>
>
>
> On 11/23/06, Elena Semukhina <el...@gmail.com> wrote:
> >
> >
> >
> > On 11/22/06, Elena Semukhina < elena.semukhina@gmail.com> wrote:
> > >
> > >
> > >
> > > On 11/21/06, Geir Magnusson Jr. < geir@pobox.com > wrote:
> > > >
> > > > I just checked in a change to the drlvm test system that uses
> > > > external
> > > > exclude files, delineated by platform. (I just realized that I need
> > > > to
> > > > also separate by jit/opt/int as well, as we have that distinction on
> > > > x86_64 linux. I'll do that in a sec)
> > > >
> > > > The idea is that in make/excludes are files for each os/arch and
> > > > soon
> > > > jit. It uses the same nomenclature as in classlib, and in fact the
> > > > build now 'reaches over' and imports the properties.xml file from
> > > > classlib build so we don't have more than open copy of that code
> > > > around.
> > > >
> > > > Note that these tokens are different than what we'd using in DRLVM
> > > > traditionally, but we need to converge and this is a first, harmless
> > > > step. I'll be slowly trying to switch DRLVM over to that token set
> > > > for
> > > > arch and OS.
> > > >
> > > > I've tested on linux x86 and now testing on x86_64. if someone
> > > > could
> > > > simply give it a whirl on windows, I'd be mighty obliged
> > >
> > >
> > > I ran the tests on Windows 2003/XP. The tests passed on Windows 2003.
> > > On Windows XP they hung on java.lang.ThreadTest in opt mode. I'll try
> > > to locate the exact test case which causes the hang.
> > >
> > > I suspect that java.lang.Object test has been excluded because it
> > > intermittently failed on testWaitlongint() with too early exit from wait(). http://issues.apache.org/jira/browse/HARMONY-2204
> > > describes the issue. I plan to do some more investigation on that and
> > > compare to RI's behavior so that we could fix the test.
> > >
> >
> > I ran ObjectTest iteratively on RI and saw that it failed occasionally
> > in the same way as on DRLVM: it waited 499 ms instead of 500. I think we can
> > allow the test more to be more tolerant in waiting. I'll adjust the test and
> > see how it behaves.
> >
>
> I finished my investigation on ObjectTest. Actually there are two test
> cases in it which fail on both RI and DRLVM. I fixed the test so that it
> passes on both VM's now. I ran it iteratively on linux / Windows for a long
> time and have not seen a failure.
>
> I think it could be included back to test runs. Please see
> https://issues.apache.org/jira/browse/HARMONY-2294 for the suggested fix.
>
Alexey has already committed the fix so I've updated
http://wiki.apache.org/harmony/DRLVMInternalTests.
Now is the ThreadTest's turn. I plan to revise it because it still shows
occasional failures in iterative runs.
>
>
> geir
> > > >
> > >
> > >
> > >
> > > --
> > > Thanks,
> > > Elena
> >
> >
> >
> >
> > --
> > Thanks,
> > Elena
>
>
>
>
> --
> Thanks,
> Elena
--
Thanks,
Elena
Re: [drlvm][build] exclude files for kernel tests
Posted by "Geir Magnusson Jr." <ge...@pobox.com>.
Elena Semukhina wrote:
> On 11/23/06, Elena Semukhina <el...@gmail.com> wrote:
>>
>> I ran ObjectTest iteratively on RI and saw that it failed occasionally in
>> the same way as on DRLVM: it waited 499 ms instead of 500. I think we can
>> allow the test more to be more tolerant in waiting. I'll adjust the
>> test and
>> see how it behaves.
>>
>
> I finished my investigation on ObjectTest. Actually there are two test
> cases
> in it which fail on both RI and DRLVM. I fixed the test so that it
> passes on
> both VM's now. I ran it iteratively on linux / Windows for a long time and
> have not seen a failure.
>
> I think it could be included back to test runs. Please see
> https://issues.apache.org/jira/browse/HARMONY-2294 for the suggested fix.
Excellent! Thank you!
geir
Re: [drlvm][build] exclude files for kernel tests
Posted by Elena Semukhina <el...@gmail.com>.
On 11/23/06, Elena Semukhina <el...@gmail.com> wrote:
>
>
>
> On 11/22/06, Elena Semukhina <el...@gmail.com> wrote:
> >
> >
> >
> > On 11/21/06, Geir Magnusson Jr. < geir@pobox.com> wrote:
> > >
> > > I just checked in a change to the drlvm test system that uses external
> > > exclude files, delineated by platform. (I just realized that I need
> > > to
> > > also separate by jit/opt/int as well, as we have that distinction on
> > > x86_64 linux. I'll do that in a sec)
> > >
> > > The idea is that in make/excludes are files for each os/arch and soon
> > > jit. It uses the same nomenclature as in classlib, and in fact the
> > > build now 'reaches over' and imports the properties.xml file from
> > > classlib build so we don't have more than open copy of that code
> > > around.
> > >
> > > Note that these tokens are different than what we'd using in DRLVM
> > > traditionally, but we need to converge and this is a first, harmless
> > > step. I'll be slowly trying to switch DRLVM over to that token set
> > > for
> > > arch and OS.
> > >
> > > I've tested on linux x86 and now testing on x86_64. if someone could
> > > simply give it a whirl on windows, I'd be mighty obliged
> >
> >
> > I ran the tests on Windows 2003/XP. The tests passed on Windows 2003. On
> > Windows XP they hung on java.lang.ThreadTest in opt mode. I'll try to
> > locate the exact test case which causes the hang.
> >
> > I suspect that java.lang.Object test has been excluded because it
> > intermittently failed on testWaitlongint() with too early exit from wait(). http://issues.apache.org/jira/browse/HARMONY-2204
> > describes the issue. I plan to do some more investigation on that and
> > compare to RI's behavior so that we could fix the test.
> >
>
> I ran ObjectTest iteratively on RI and saw that it failed occasionally in
> the same way as on DRLVM: it waited 499 ms instead of 500. I think we can
> allow the test more to be more tolerant in waiting. I'll adjust the test and
> see how it behaves.
>
I finished my investigation on ObjectTest. Actually there are two test cases
in it which fail on both RI and DRLVM. I fixed the test so that it passes on
both VM's now. I ran it iteratively on linux / Windows for a long time and
have not seen a failure.
I think it could be included back to test runs. Please see
https://issues.apache.org/jira/browse/HARMONY-2294 for the suggested fix.
geir
> > >
> >
> >
> >
> > --
> > Thanks,
> > Elena
>
>
>
>
> --
> Thanks,
> Elena
--
Thanks,
Elena
Re: [drlvm][build] exclude files for kernel tests
Posted by Elena Semukhina <el...@gmail.com>.
On 11/22/06, Elena Semukhina <el...@gmail.com> wrote:
>
>
>
> On 11/21/06, Geir Magnusson Jr. <ge...@pobox.com> wrote:
> >
> > I just checked in a change to the drlvm test system that uses external
> > exclude files, delineated by platform. (I just realized that I need to
> > also separate by jit/opt/int as well, as we have that distinction on
> > x86_64 linux. I'll do that in a sec)
> >
> > The idea is that in make/excludes are files for each os/arch and soon
> > jit. It uses the same nomenclature as in classlib, and in fact the
> > build now 'reaches over' and imports the properties.xml file from
> > classlib build so we don't have more than open copy of that code around.
> >
> > Note that these tokens are different than what we'd using in DRLVM
> > traditionally, but we need to converge and this is a first, harmless
> > step. I'll be slowly trying to switch DRLVM over to that token set for
> > arch and OS.
> >
> > I've tested on linux x86 and now testing on x86_64. if someone could
> > simply give it a whirl on windows, I'd be mighty obliged
>
>
> I ran the tests on Windows 2003/XP. The tests passed on Windows 2003. On
> Windows XP they hung on java.lang.ThreadTest in opt mode. I'll try to
> locate the exact test case which causes the hang.
>
> I suspect that java.lang.Object test has been excluded because it
> intermittently failed on testWaitlongint() with too early exit from wait(). http://issues.apache.org/jira/browse/HARMONY-2204
> describes the issue. I plan to do some more investigation on that and
> compare to RI's behavior so that we could fix the test.
>
I ran ObjectTest iteratively on RI and saw that it failed occasionally in
the same way as on DRLVM: it waited 499 ms instead of 500. I think we can
allow the test more to be more tolerant in waiting. I'll adjust the test and
see how it behaves.
geir
> >
>
>
>
> --
> Thanks,
> Elena
--
Thanks,
Elena
Re: [drlvm][build] exclude files for kernel tests
Posted by Elena Semukhina <el...@gmail.com>.
On 11/21/06, Geir Magnusson Jr. <ge...@pobox.com> wrote:
>
> I just checked in a change to the drlvm test system that uses external
> exclude files, delineated by platform. (I just realized that I need to
> also separate by jit/opt/int as well, as we have that distinction on
> x86_64 linux. I'll do that in a sec)
>
> The idea is that in make/excludes are files for each os/arch and soon
> jit. It uses the same nomenclature as in classlib, and in fact the
> build now 'reaches over' and imports the properties.xml file from
> classlib build so we don't have more than open copy of that code around.
>
> Note that these tokens are different than what we'd using in DRLVM
> traditionally, but we need to converge and this is a first, harmless
> step. I'll be slowly trying to switch DRLVM over to that token set for
> arch and OS.
>
> I've tested on linux x86 and now testing on x86_64. if someone could
> simply give it a whirl on windows, I'd be mighty obliged
I ran the tests on Windows 2003/XP. The tests passed on Windows 2003. On
Windows XP they hung on java.lang.ThreadTest in opt mode. I'll try to locate
the exact test case which causes the hang.
I suspect that java.lang.Object test has been excluded because it
intermittently failed on testWaitlongint() with too early exit from wait().
http://issues.apache.org/jira/browse/HARMONY-2204 describes the issue. I
plan to do some more investigation on that and compare to RI's behavior so
that we could fix the test.
geir
>
--
Thanks,
Elena