test results with mpi vanish

[smr99]

Hi,

I enabled mpi on my "Debian-Sid" testing box, for both trunk and
release.  The test script runs and the results are uploaded but
nothing appears on the testing pages; e.g.
http://beta.boost.org/development/tests/release/developer/summary.html

If I remove mpi, the results are shown as usual.  Perhaps something
has to be updated on the testing server for MPI?

Cheers,
-Steve



_______________________________________________
Boost-Testing mailing list
Boost-Testing@...
http://lists.boost.org/mailman/listinfo.cgi/boost-testing

[Beman Dawes]

On Sat, Jul 25, 2009 at 4:31 AM, Steve M. Robbins<steve@...> wrote:
> Hi,
>
> I enabled mpi on my "Debian-Sid" testing box, for both trunk and
> release.  The test script runs and the results are uploaded but
> nothing appears on the testing pages; e.g.
> http://beta.boost.org/development/tests/release/developer/summary.html
>
> If I remove mpi, the results are shown as usual.  Perhaps something
> has to be updated on the testing server for MPI?

I can't think of any reason the server reporting would be affected by MPI.

Does a manually run test of Boost.MPI work outside of the regression
test harness?

Does the uploaded xml file (inside the .zip) look OK, particularly the
Boost.MPI tests?

Is anyone else running regressions with MPI enabled? I'm not seeing
any test under that name on either the trunk or release.

--Beman
_______________________________________________
Boost-Testing mailing list
Boost-Testing@...
http://lists.boost.org/mailman/listinfo.cgi/boost-testing

[Jeremiah Willcock]

On Sat, 25 Jul 2009, Beman Dawes wrote:

> On Sat, Jul 25, 2009 at 4:31 AM, Steve M. Robbins<steve@...> wrote:
>> Hi,
>>
>> I enabled mpi on my "Debian-Sid" testing box, for both trunk and
>> release.  The test script runs and the results are uploaded but
>> nothing appears on the testing pages; e.g.
>> http://beta.boost.org/development/tests/release/developer/summary.html
>>
>> If I remove mpi, the results are shown as usual.  Perhaps something
>> has to be updated on the testing server for MPI?
>
> I can't think of any reason the server reporting would be affected by MPI.
>
> Does a manually run test of Boost.MPI work outside of the regression
> test harness?
>
> Does the uploaded xml file (inside the .zip) look OK, particularly the
> Boost.MPI tests?
>
> Is anyone else running regressions with MPI enabled? I'm not seeing
> any test under that name on either the trunk or release.

Is there anything in the normal test rules in Boost.Build that handles
updating the XML results?  Boost.MPI (and the parallel parts of
Boost.Graph) have custom test running rules because they need to call a
wrapper program that is part of the MPI implementation to run their tests.
It could be that this custom rule (mpi-test from
tools/build/v2/tools/mpi.jam) does not have all of the features of the
normal test rule.  From my looking through testing.jam and mpi.jam,
testing.jam produces XML output around line 272 in the SVN HEAD, while
mpi.jam does not mention XML at all.

-- Jeremiah Willcock
_______________________________________________
Boost-Testing mailing list
Boost-Testing@...
http://lists.boost.org/mailman/listinfo.cgi/boost-testing

[smr99]

Hi Beman,

I was away for a week, so apologies for the delay in response.

The "trunk" summary page still does not display my results.

The "release" summary page, however, does display a column for
the machine, but no MPI results are reported.  There are, indeed,
mpi results inside the xml file.


On Sat, Jul 25, 2009 at 11:13:29AM -0400, Beman Dawes wrote:

> On Sat, Jul 25, 2009 at 4:31 AM, Steve M. Robbins<steve@...> wrote:
> > Hi,
> >
> > I enabled mpi on my "Debian-Sid" testing box, for both trunk and
> > release. ?The test script runs and the results are uploaded but
> > nothing appears on the testing pages; e.g.
> > http://beta.boost.org/development/tests/release/developer/summary.html
> >
> > If I remove mpi, the results are shown as usual. ?Perhaps something
> > has to be updated on the testing server for MPI?
>
> I can't think of any reason the server reporting would be affected by MPI.

I can't either, but as I say, merely enabling MPI causes these
mysterious problems.


> Does the uploaded xml file (inside the .zip) look OK, particularly the
> Boost.MPI tests?

Looks OK to me, but I'm no expert on what it should look like.  I expect you
can ftp the file yourself to have a look; or I can email you a copy.

Cheers,
-Steve


_______________________________________________
Boost-Testing mailing list
Boost-Testing@...
http://lists.boost.org/mailman/listinfo.cgi/boost-testing

[Jeremiah Willcock]

On Sun, 2 Aug 2009, Steve M. Robbins wrote:

> Hi Beman,
>
> I was away for a week, so apologies for the delay in response.
>
> The "trunk" summary page still does not display my results.
>
> The "release" summary page, however, does display a column for
> the machine, but no MPI results are reported.  There are, indeed,
> mpi results inside the xml file.

One thing that I see is that the meta:is_test_log_a_test_case function in
tools/regression/xsl_reports/xsl/v2/common.xsl does not recognize run_mpi
as a type of test case; all Boost.MPI and Boost.Graph.Parallel tests use
this type and so probably would not show up in the test result Web pages.

-- Jeremiah Willcock
_______________________________________________
Boost-Testing mailing list
Boost-Testing@...
http://lists.boost.org/mailman/listinfo.cgi/boost-testing

[smr99]

On Mon, Aug 03, 2009 at 12:00:38AM -0400, Jeremiah Willcock wrote:

> On Sun, 2 Aug 2009, Steve M. Robbins wrote:
>
> >Hi Beman,
> >
> >I was away for a week, so apologies for the delay in response.
> >
> >The "trunk" summary page still does not display my results.
> >
> >The "release" summary page, however, does display a column for
> >the machine, but no MPI results are reported.  There are, indeed,
> >mpi results inside the xml file.
>
> One thing that I see is that the meta:is_test_log_a_test_case
> function in tools/regression/xsl_reports/xsl/v2/common.xsl does not
> recognize run_mpi as a type of test case; all Boost.MPI and
> Boost.Graph.Parallel tests use this type and so probably would not
> show up in the test result Web pages.

Thanks.  This indeed seems to be the culprit: I manually changed
"run_mpi" to "run" in my testing results and re-uploaded them.
The testing results website now includes mpi.  

I trust that either common.xsl or the generator of "run_mpi" will be
modified.

All MPI tests fail to build, by the way.  Here's the error shown in
the first few:

In file included from ../boost/mpi/detail/mpi_datatype_cache.hpp:13,
                 from ../boost/mpi/datatype.hpp:27,
                 from ../boost/mpi/communicator.hpp:20,
                 from ../boost/mpi/collectives.hpp:21,
                 from ../boost/mpi/collectives_fwd.hpp:19,
                 from ../boost/mpi/collectives/broadcast.hpp:11,
                 from ../boost/mpi/collectives/all_reduce.hpp:16,
                 from ../libs/mpi/test/all_reduce_test.cpp:8:
../boost/mpi/detail/mpi_datatype_oarchive.hpp:45: error: expected constructor, destructor, or type conversion before 'BOOST_SERIALIZATION_REGISTER_ARCHIVE'


Regards,
-Steve


_______________________________________________
Boost-Testing mailing list
Boost-Testing@...
http://lists.boost.org/mailman/listinfo.cgi/boost-testing

[smr99]

On Mon, Aug 03, 2009 at 12:00:38AM -0400, Jeremiah Willcock wrote:

> One thing that I see is that the meta:is_test_log_a_test_case
> function in tools/regression/xsl_reports/xsl/v2/common.xsl does not
> recognize run_mpi as a type of test case; all Boost.MPI and
> Boost.Graph.Parallel tests use this type and so probably would not
> show up in the test result Web pages.

Here's a trivial patch to fix this.  Who can get this installed
on the results server?

Index: tools/regression/xsl_reports/xsl/v2/common.xsl
===================================================================
--- tools/regression/xsl_reports/xsl/v2/common.xsl (revision 55395)
+++ tools/regression/xsl_reports/xsl/v2/common.xsl (working copy)
@@ -151,7 +151,7 @@
         <xsl:param name="test_log"/>      
         <xsl:variable name="type" select="$test_log/@test-type"/>
         <func:result select="$type='compile' or $type='compile_fail' or $type='link' or $type='link_fail'
-                             or $type='run' or $type='run_fail' or $type='run_pyd'"/>
+                             or $type='run' or $type='run_fail' or $type='run_pyd' or $type='run_mpi'"/>
     </func:function>
 

Thanks,
-Steve


_______________________________________________
Boost-Testing mailing list
Boost-Testing@...
http://lists.boost.org/mailman/listinfo.cgi/boost-testing

[Rene Rivera-2]

Steve M. Robbins wrote:

> On Mon, Aug 03, 2009 at 12:00:38AM -0400, Jeremiah Willcock wrote:
>
>> One thing that I see is that the meta:is_test_log_a_test_case
>> function in tools/regression/xsl_reports/xsl/v2/common.xsl does not
>> recognize run_mpi as a type of test case; all Boost.MPI and
>> Boost.Graph.Parallel tests use this type and so probably would not
>> show up in the test result Web pages.
>
> Here's a trivial patch to fix this.  Who can get this installed
> on the results server?

Theoretically, just checking in the change will get it to the results
server(s).


--
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software.com
-- rrivera/acm.org (msn) - grafik/redshift-software.com
-- 102708583/icq - grafikrobot/aim,yahoo,skype,efnet,gmail
_______________________________________________
Boost-Testing mailing list
Boost-Testing@...
http://lists.boost.org/mailman/listinfo.cgi/boost-testing

[Beman Dawes]

On Mon, Aug 3, 2009 at 11:54 PM, Steve M. Robbins<steve@...> wrote:

> On Mon, Aug 03, 2009 at 12:00:38AM -0400, Jeremiah Willcock wrote:
>
>> One thing that I see is that the meta:is_test_log_a_test_case
>> function in tools/regression/xsl_reports/xsl/v2/common.xsl does not
>> recognize run_mpi as a type of test case; all Boost.MPI and
>> Boost.Graph.Parallel tests use this type and so probably would not
>> show up in the test result Web pages.
>
> Here's a trivial patch to fix this.  Who can get this installed
> on the results server?

OK, committed to trunk.

Please keep an eye on this to make sure it really propagates to the
results servers, and has the intended effects.

Thanks,

--Beman
_______________________________________________
Boost-Testing mailing list
Boost-Testing@...
http://lists.boost.org/mailman/listinfo.cgi/boost-testing

[smr99]

On Wed, Aug 05, 2009 at 09:00:19AM -0400, Beman Dawes wrote:

> On Mon, Aug 3, 2009 at 11:54 PM, Steve M. Robbins<steve@...> wrote:
> > On Mon, Aug 03, 2009 at 12:00:38AM -0400, Jeremiah Willcock wrote:
> >
> >> One thing that I see is that the meta:is_test_log_a_test_case
> >> function in tools/regression/xsl_reports/xsl/v2/common.xsl does not
> >> recognize run_mpi as a type of test case; all Boost.MPI and
> >> Boost.Graph.Parallel tests use this type and so probably would not
> >> show up in the test result Web pages.
> >
> > Here's a trivial patch to fix this. ?Who can get this installed
> > on the results server?
>
> OK, committed to trunk.
>
> Please keep an eye on this to make sure it really propagates to the
> results servers, and has the intended effects.

I don't see any effect on the results server.

-Steve


_______________________________________________
Boost-Testing mailing list
Boost-Testing@...
http://lists.boost.org/mailman/listinfo.cgi/boost-testing

[Jeremiah Willcock]

On Thu, 6 Aug 2009, Steve M. Robbins wrote:

> On Wed, Aug 05, 2009 at 09:00:19AM -0400, Beman Dawes wrote:
>> On Mon, Aug 3, 2009 at 11:54 PM, Steve M. Robbins<steve@...> wrote:
>>> On Mon, Aug 03, 2009 at 12:00:38AM -0400, Jeremiah Willcock wrote:
>>>
>>>> One thing that I see is that the meta:is_test_log_a_test_case
>>>> function in tools/regression/xsl_reports/xsl/v2/common.xsl does not
>>>> recognize run_mpi as a type of test case; all Boost.MPI and
>>>> Boost.Graph.Parallel tests use this type and so probably would not
>>>> show up in the test result Web pages.
>>>
>>> Here's a trivial patch to fix this. ?Who can get this installed
>>> on the results server?
>>
>> OK, committed to trunk.
>>
>> Please keep an eye on this to make sure it really propagates to the
>> results servers, and has the intended effects.
>
> I don't see any effect on the results server.

I don't either.  Could someone please fix this soon because this library
is currently completely broken in the release branch and it needs to be
fixed up before the release.

-- Jeremiah Willcock
_______________________________________________
Boost-Testing mailing list
Boost-Testing@...
http://lists.boost.org/mailman/listinfo.cgi/boost-testing