[1.40.0] Freeze for release candidate
As far as I can tell, all the branches/release merges we were waiting for have now been completed. The plan is to go ahead with the release candidate as soon as all the release branch tests have cycled for revision 55757. Please commit no more changes to branches/release without explicit permission from me. Thanks, --Beman
On Mon, 24 Aug 2009, Beman Dawes wrote:
As far as I can tell, all the branches/release merges we were waiting for have now been completed. The plan is to go ahead with the release candidate as soon as all the release branch tests have cycled for revision 55757.
Please commit no more changes to branches/release without explicit permission from me.
One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before the release as well. What are other peoples' opinions on that? -- Jeremiah Willcock
Jeremiah Willcock wrote:
On Mon, 24 Aug 2009, Beman Dawes wrote:
As far as I can tell, all the branches/release merges we were waiting for have now been completed. The plan is to go ahead with the release candidate as soon as all the release branch tests have cycled for revision 55757.
Please commit no more changes to branches/release without explicit permission from me.
One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before the release as well. What are other peoples' opinions on that?
Someone would need to step forward to run them. If someone did that Soon (in the next few days), I'm pretty sure we would be happy to hold the release for a test cycle. Do we have any volunteers? -- Eric Niebler BoostPro Computing http://www.boostpro.com
On Mon, 24 Aug 2009, Eric Niebler wrote:
Jeremiah Willcock wrote:
On Mon, 24 Aug 2009, Beman Dawes wrote:
As far as I can tell, all the branches/release merges we were waiting for have now been completed. The plan is to go ahead with the release candidate as soon as all the release branch tests have cycled for revision 55757.
Please commit no more changes to branches/release without explicit permission from me.
One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before the release as well. What are other peoples' opinions on that?
Someone would need to step forward to run them. If someone did that Soon (in the next few days), I'm pretty sure we would be happy to hold the release for a test cycle. Do we have any volunteers?
Steve Robbins was running them (and probably still is), but the results were not being reported to the regression testing site. There was a discussion on boost-testing about the issue: "[Boost-testing] test results with mpi vanish", but I do not believe that the bug in the Web site was ever resolved to allow the tests to appear. -- Jeremiah Willcock
Jeremiah Willcock wrote:
On Mon, 24 Aug 2009, Eric Niebler wrote:
Jeremiah Willcock wrote:
On Mon, 24 Aug 2009, Beman Dawes wrote:
As far as I can tell, all the branches/release merges we were waiting for have now been completed. The plan is to go ahead with the release candidate as soon as all the release branch tests have cycled for revision 55757.
Please commit no more changes to branches/release without explicit permission from me.
One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before the release as well. What are other peoples' opinions on that?
Someone would need to step forward to run them. If someone did that Soon (in the next few days), I'm pretty sure we would be happy to hold the release for a test cycle. Do we have any volunteers?
Steve Robbins was running them (and probably still is), but the results were not being reported to the regression testing site. There was a discussion on boost-testing about the issue: "[Boost-testing] test results with mpi vanish", but I do not believe that the bug in the Web site was ever resolved to allow the tests to appear.
I believe it was resolved, and there was at least one fairly green run from Steven. The current results are unwell: http://www.boost.org/development/tests/trunk/developer/output/Debian-Sid-boo... But it seems due to genuine code-level problem. - Volodya
-- Jeremiah Willcock _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
On Mon, 24 Aug 2009, Vladimir Prus wrote:
Jeremiah Willcock wrote:
On Mon, 24 Aug 2009, Eric Niebler wrote:
Jeremiah Willcock wrote:
On Mon, 24 Aug 2009, Beman Dawes wrote:
As far as I can tell, all the branches/release merges we were waiting for have now been completed. The plan is to go ahead with the release candidate as soon as all the release branch tests have cycled for revision 55757.
Please commit no more changes to branches/release without explicit permission from me.
One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before the release as well. What are other peoples' opinions on that?
Someone would need to step forward to run them. If someone did that Soon (in the next few days), I'm pretty sure we would be happy to hold the release for a test cycle. Do we have any volunteers?
Steve Robbins was running them (and probably still is), but the results were not being reported to the regression testing site. There was a discussion on boost-testing about the issue: "[Boost-testing] test results with mpi vanish", but I do not believe that the bug in the Web site was ever resolved to allow the tests to appear.
I believe it was resolved, and there was at least one fairly green run from Steven. The current results are unwell:
http://www.boost.org/development/tests/trunk/developer/output/Debian-Sid-boo...
But it seems due to genuine code-level problem.
Those results date from August 3 (three weeks ago), however; I think Steve hacked up the result file manually to make the Website work correctly (default runs still do not upload, I believe). There also does not appear to be anything similar for the release branch. The page for the release branch exists if you know the URL, but it is not listed in the library list. -- Jeremiah Willcock
Jeremiah Willcock wrote:
On Mon, 24 Aug 2009, Vladimir Prus wrote:
Jeremiah Willcock wrote:
On Mon, 24 Aug 2009, Eric Niebler wrote:
Jeremiah Willcock wrote:
On Mon, 24 Aug 2009, Beman Dawes wrote:
As far as I can tell, all the branches/release merges we were waiting for have now been completed. The plan is to go ahead with the release candidate as soon as all the release branch tests have cycled for revision 55757.
Please commit no more changes to branches/release without explicit permission from me.
One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before the release as well. What are other peoples' opinions on that?
Someone would need to step forward to run them. If someone did that Soon (in the next few days), I'm pretty sure we would be happy to hold the release for a test cycle. Do we have any volunteers?
Steve Robbins was running them (and probably still is), but the results were not being reported to the regression testing site. There was a discussion on boost-testing about the issue: "[Boost-testing] test results with mpi vanish", but I do not believe that the bug in the Web site was ever resolved to allow the tests to appear.
I believe it was resolved, and there was at least one fairly green run from Steven. The current results are unwell:
http://www.boost.org/development/tests/trunk/developer/output/Debian-Sid-boo...
But it seems due to genuine code-level problem.
Those results date from August 3 (three weeks ago), however; I think Steve hacked up the result file manually to make the Website work correctly (default runs still do not upload, I believe).
I have personally fixed the test reporting, and saw the MPI tests appear shortly thereafter, with no hacked results.
There also does not appear to be anything similar for the release branch.
Because Steven is not running tests for release branch.
The page for the release branch exists if you know the URL, but it is not listed in the library list.
Can you look at: http://www.boost.org/development/tests/trunk/developer/summary.html It does list mpi. - Volodya
On Mon, 24 Aug 2009, Vladimir Prus wrote:
Can you look at:
http://www.boost.org/development/tests/trunk/developer/summary.html
It does list mpi.
I have seen that; it has MPI and Graph.Parallel. However, it is still for the trunk. If you are correct that nobody is running the MPI tests for the release branch, that would explain them not showing up on the result list. It would also be good if someone with Expat set up their Boost.Build configuration to use that and ensure that the Graph tests that use it also work. -- Jeremiah Willcock
Vladimir Prus wrote:
Jeremiah Willcock wrote:
On Mon, 24 Aug 2009, Eric Niebler wrote:
On Mon, 24 Aug 2009, Beman Dawes wrote:
As far as I can tell, all the branches/release merges we were waiting for have now been completed. The plan is to go ahead with the release candidate as soon as all the release branch tests have cycled for revision 55757.
Please commit no more changes to branches/release without explicit permission from me. One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before the release as well. What are other peoples' opinions on that? Someone would need to step forward to run them. If someone did that Soon (in
Jeremiah Willcock wrote: the next few days), I'm pretty sure we would be happy to hold the release for a test cycle. Do we have any volunteers? Steve Robbins was running them (and probably still is), but the results were not being reported to the regression testing site. There was a discussion on boost-testing about the issue: "[Boost-testing] test results with mpi vanish", but I do not believe that the bug in the Web site was ever resolved to allow the tests to appear.
I believe it was resolved, and there was at least one fairly green run from Steven. The current results are unwell:
http://www.boost.org/development/tests/trunk/developer/output/Debian-Sid-boo...
But it seems due to genuine code-level problem.
On the release branch, boost/archive/detail/iserializer.hpp does not include boost/archive/detail/check.hpp, which seems to indicate to me that this tester is either testing old code or else is not testing the release branch at all. Can someone comment? Steve? -- Eric Niebler BoostPro Computing http://www.boostpro.com
Eric Niebler wrote:
http://www.boost.org/development/tests/trunk/...developer/output/Debian-Sid-...
But it seems due to genuine code-level problem.
On the release branch, boost/archive/detail/iserializer.hpp does not include boost/archive/detail/check.hpp, which seems to indicate to me that this tester is either testing old code or else is not testing the release branch at all. Can someone comment? Steve?
The above are trunk results (as the URL say). - Volodya
On Mon, Aug 24, 2009 at 08:55:25PM +0400, Vladimir Prus wrote:
Jeremiah Willcock wrote:
On Mon, 24 Aug 2009, Eric Niebler wrote:
Jeremiah Willcock wrote:
On Mon, 24 Aug 2009, Beman Dawes wrote:
As far as I can tell, all the branches/release merges we were waiting for have now been completed. The plan is to go ahead with the release candidate as soon as all the release branch tests have cycled for revision 55757.
Please commit no more changes to branches/release without explicit permission from me.
One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before the release as well. What are other peoples' opinions on that?
Someone would need to step forward to run them. If someone did that Soon (in the next few days), I'm pretty sure we would be happy to hold the release for a test cycle. Do we have any volunteers?
Steve Robbins was running them (and probably still is), but the results were not being reported to the regression testing site. There was a discussion on boost-testing about the issue: "[Boost-testing] test results with mpi vanish", but I do not believe that the bug in the Web site was ever resolved to allow the tests to appear.
I believe it was resolved,
It was resolved for trunk only, not for release. I'm building "Debian-Sid" for both trunk and release -- with MPI configured -- for weeks now, but the results only show on trunk. Anyone interested can obtain the results by ftp, or I can send in email. Release has at least one failure, btw. -Steve
On Mon, Aug 24, 2009 at 08:20:10PM -0500, Steve M. Robbins wrote:
It was resolved for trunk only, not for release. I'm building "Debian-Sid" for both trunk and release -- with MPI configured -- for weeks now, but the results only show on trunk. Anyone interested can obtain the results by ftp, or I can send in email.
I took today's results, manually changed mpi_run --> run, and uploaded Debian-Sid. If you look at the release summary page before it gets overwritten by tomorrow's results, you can see the three failures: mpi, optional, and python. Cheers, -Steve
Steve M. Robbins wrote:
Steve Robbins was running them (and probably still is), but the results were not being reported to the regression testing site. There was a discussion on boost-testing about the issue: "[Boost-testing] test results with mpi vanish", but I do not believe that the bug in the Web site was ever resolved to allow the tests to appear.
I believe it was resolved,
It was resolved for trunk only, not for release. I'm building "Debian-Sid" for both trunk and release -- with MPI configured -- for weeks now, but the results only show on trunk.
Oh, that's fun. For all I can tell, everything should be using trunk version of XSL scripts. Who is running the reporting script for release, and how exactly? - Volodya
On Aug 24, 2009, at 11:22 PM, Vladimir Prus wrote:
Steve M. Robbins wrote:
Steve Robbins was running them (and probably still is), but the results were not being reported to the regression testing site. There was a discussion on boost-testing about the issue: "[Boost-testing] test results with mpi vanish", but I do not believe that the bug in the Web site was ever resolved to allow the tests to appear.
I believe it was resolved,
It was resolved for trunk only, not for release. I'm building "Debian-Sid" for both trunk and release -- with MPI configured -- for weeks now, but the results only show on trunk.
Oh, that's fun. For all I can tell, everything should be using trunk version of XSL scripts. Who is running the reporting script for release, and how exactly?
That would be me. But it sure looks like there's graph_parallel and mpi results from Debian-Sid on Release. http://www.boost.org/development/tests/release/developer/summary.html -- Noel
Belcourt, Kenneth wrote:
On Aug 24, 2009, at 11:22 PM, Vladimir Prus wrote:
Steve M. Robbins wrote:
Steve Robbins was running them (and probably still is), but the results were not being reported to the regression testing site. There was a discussion on boost-testing about the issue: "[Boost-testing] test results with mpi vanish", but I do not believe that the bug in the Web site was ever resolved to allow the tests to appear.
I believe it was resolved,
It was resolved for trunk only, not for release. I'm building "Debian-Sid" for both trunk and release -- with MPI configured -- for weeks now, but the results only show on trunk.
Oh, that's fun. For all I can tell, everything should be using trunk version of XSL scripts. Who is running the reporting script for release, and how exactly?
That would be me. But it sure looks like there's graph_parallel and mpi results from Debian-Sid on Release.
http://www.boost.org/development/tests/release/developer/summary.html
Presumably because Steven's has manually tweaked the results again :-/ Is the boost_wide_report.py script run from release checkout? This might explain the situation, as it's apparently the only script that does not fetch XSL things from trunk. - Volodya
On Tue, Aug 25, 2009 at 09:40:52AM +0400, Vladimir Prus wrote:
Belcourt, Kenneth wrote:
That would be me. But it sure looks like there's graph_parallel and mpi results from Debian-Sid on Release.
http://www.boost.org/development/tests/release/developer/summary.html
Presumably because Steven's has manually tweaked the results again :-/
Yes, I did. -Steve
Steve M. Robbins wrote:
On Tue, Aug 25, 2009 at 09:40:52AM +0400, Vladimir Prus wrote:
Belcourt, Kenneth wrote:
That would be me. But it sure looks like there's graph_parallel and mpi results from Debian-Sid on Release.
http://www.boost.org/development/tests/release/developer/summary.html
Presumably because Steven's has manually tweaked the results again :-/
Yes, I did.
So, let me summarize the options we have at this point: 1. Decide that apparently, MPI tests were never run in past as part of regression testing framework, and therefore can wait till 1.41. 2. Have Noel try running with trunk version of XSL scripts, either using "svn switch", or by merging to release branch. Then, fix MPI failures. Can the release management team decide? Incidentally, I wonder if (2) is necessary for fixing MPI failures. Jeremiah, do you get clean MPI test run locally? If not, maybe you can start to fix failures pro-actively? Thanks, Volodya
On Tue, 25 Aug 2009, Vladimir Prus wrote:
Steve M. Robbins wrote:
On Tue, Aug 25, 2009 at 09:40:52AM +0400, Vladimir Prus wrote:
Belcourt, Kenneth wrote:
That would be me. But it sure looks like there's graph_parallel and mpi results from Debian-Sid on Release.
http://www.boost.org/development/tests/release/developer/summary.html
Presumably because Steven's has manually tweaked the results again :-/
Yes, I did.
So, let me summarize the options we have at this point:
1. Decide that apparently, MPI tests were never run in past as part of regression testing framework, and therefore can wait till 1.41.
2. Have Noel try running with trunk version of XSL scripts, either using "svn switch", or by merging to release branch. Then, fix MPI failures.
Can the release management team decide?
Incidentally, I wonder if (2) is necessary for fixing MPI failures. Jeremiah, do you get clean MPI test run locally? If not, maybe you can start to fix failures pro-actively?
I run on Darwin, and do get a clean run locally. I am actually not the maintainer of Boost.MPI, so I am not sure where to start to look for the bug (there is no output from the failed runs). I will try to get everything set up on Linux, but it might be easier if someone (Steve?) could see if there is a crash or otherwise where the failure is coming from and get a backtrace. -- Jeremiah Willcock
Vladimir Prus wrote:
Steve M. Robbins wrote:
On Tue, Aug 25, 2009 at 09:40:52AM +0400, Vladimir Prus wrote:
Belcourt, Kenneth wrote:
That would be me. But it sure looks like there's graph_parallel and mpi results from Debian-Sid on Release.
http://www.boost.org/development/tests/release/developer/summary.html Presumably because Steven's has manually tweaked the results again :-/ Yes, I did.
The results aren't there, now.
So, let me summarize the options we have at this point:
1. Decide that apparently, MPI tests were never run in past as part of regression testing framework,
Is this really an accurate characterization? I can't believe that we've *never* had MPI test results.
and therefore can wait till 1.41.
I don't think we should ship untested code.
2. Have Noel try running with trunk version of XSL scripts, either using "svn switch", or by merging to release branch. Then, fix MPI failures.
Can the release management team decide?
My vote goes for (2). We have some MPI test results now. Matthias should have a look and report on the failures there. Then we can decide whether there are any showstoppers. Marshal, can you make sure Matthias has access to your test results? -- Eric Niebler BoostPro Computing http://www.boostpro.com
Eric Niebler wrote:
Vladimir Prus wrote:
Steve M. Robbins wrote:
On Tue, Aug 25, 2009 at 09:40:52AM +0400, Vladimir Prus wrote:
Belcourt, Kenneth wrote:
That would be me. But it sure looks like there's graph_parallel and mpi results from Debian-Sid on Release.
http://www.boost.org/development/tests/release/developer/summary.html Presumably because Steven's has manually tweaked the results again :-/ Yes, I did.
The results aren't there, now.
Apparently, new tests results were uploaded and now: - we no longer see any results on release branch - the trunk results are all green
So, let me summarize the options we have at this point:
1. Decide that apparently, MPI tests were never run in past as part of regression testing framework,
Is this really an accurate characterization? I can't believe that we've *never* had MPI test results.
I don't see how we could ever had them. The XSL scripts were fixed to accept run_mpi test type just recently, with previous change in 2007-12-13. mpi.jam uses RUN_MPI type since 2006-12-18. So, it seems like no MPI results could appear in the matrix for all that time.
and therefore can wait till 1.41.
I don't think we should ship untested code.
2. Have Noel try running with trunk version of XSL scripts, either using "svn switch", or by merging to release branch. Then, fix MPI failures.
Can the release management team decide?
My vote goes for (2).
And which of alternatives should we go with? E.g. switch or merge? - Volodya
Vladimir Prus wrote:
Eric Niebler wrote:
Vladimir Prus wrote:
2. Have Noel try running with trunk version of XSL scripts, either using "svn switch", or by merging to release branch. Then, fix MPI failures.
Can the release management team decide?
My vote goes for (2).
And which of alternatives should we go with? E.g. switch or merge?
Marshall has results that Matthias can inspect by hand. I don't think we need to make any changes to the regression testing system to feel good enough to deliver a release. After the release is out, we'll address the problems in the regression testing system. -- Eric Niebler BoostPro Computing http://www.boostpro.com
On Tue, 25 Aug 2009, Vladimir Prus wrote:
Steve M. Robbins wrote:
On Tue, Aug 25, 2009 at 09:40:52AM +0400, Vladimir Prus wrote:
Belcourt, Kenneth wrote:
That would be me. But it sure looks like there's graph_parallel and mpi results from Debian-Sid on Release.
http://www.boost.org/development/tests/release/developer/summary.html
Presumably because Steven's has manually tweaked the results again :-/
Yes, I did.
So, let me summarize the options we have at this point:
1. Decide that apparently, MPI tests were never run in past as part of regression testing framework, and therefore can wait till 1.41.
2. Have Noel try running with trunk version of XSL scripts, either using "svn switch", or by merging to release branch. Then, fix MPI failures.
Can the release management team decide?
Incidentally, I wonder if (2) is necessary for fixing MPI failures. Jeremiah, do you get clean MPI test run locally? If not, maybe you can start to fix failures pro-actively?
Using the Intel compiler version 10 on Linux, I get a static assertion failure in save_enum() line 65 in boost/mpi/detail/mpi_datatype_oarchive.hpp when trying to compile ring_test. However, when I use a different method of checking the equality of the two sides of the comparison, it works just fine. Is MPL_STATIC_ASSERT broken on that platform? Can anyone else with more experience find out what is going on there? Thank you. -- Jeremiah Willcock
On Aug 25, 2009, at 4:45 PM, Jeremiah Willcock wrote:
On Tue, 25 Aug 2009, Vladimir Prus wrote:
Steve M. Robbins wrote:
On Tue, Aug 25, 2009 at 09:40:52AM +0400, Vladimir Prus wrote:
Belcourt, Kenneth wrote:
That would be me. But it sure looks like there's graph_parallel and mpi results from Debian-Sid on Release.
http://www.boost.org/development/tests/release/developer/summary.html
Presumably because Steven's has manually tweaked the results again :-/
Yes, I did.
So, let me summarize the options we have at this point:
1. Decide that apparently, MPI tests were never run in past as part of regression testing framework, and therefore can wait till 1.41.
2. Have Noel try running with trunk version of XSL scripts, either using "svn switch", or by merging to release branch. Then, fix MPI failures.
Can the release management team decide?
Incidentally, I wonder if (2) is necessary for fixing MPI failures. Jeremiah, do you get clean MPI test run locally? If not, maybe you can start to fix failures pro-actively?
Using the Intel compiler version 10 on Linux, I get a static assertion failure in save_enum() line 65 in boost/mpi/detail/ mpi_datatype_oarchive.hpp when trying to compile ring_test. However, when I use a different method of checking the equality of the two sides of the comparison, it works just fine. Is MPL_STATIC_ASSERT broken on that platform? Can anyone else with more experience find out what is going on there? Thank you.
I have no knowledge of MPL_STATIC_ASSERT, but would suggest that one can just add a workaround for that compiler in ring_test' Matthias
On Aug 24, 2009, at 9:20 PM, Steve M. Robbins wrote:
On Mon, Aug 24, 2009 at 08:55:25PM +0400, Vladimir Prus wrote:
Jeremiah Willcock wrote:
On Mon, 24 Aug 2009, Eric Niebler wrote:
Jeremiah Willcock wrote:
On Mon, 24 Aug 2009, Beman Dawes wrote:
As far as I can tell, all the branches/release merges we were waiting for have now been completed. The plan is to go ahead with the release candidate as soon as all the release branch tests have cycled for revision 55757.
Please commit no more changes to branches/release without explicit permission from me.
One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before the release as well. What are other peoples' opinions on that?
Someone would need to step forward to run them. If someone did that Soon (in the next few days), I'm pretty sure we would be happy to hold the release for a test cycle. Do we have any volunteers?
Steve Robbins was running them (and probably still is), but the results were not being reported to the regression testing site. There was a discussion on boost-testing about the issue: "[Boost-testing] test results with mpi vanish", but I do not believe that the bug in the Web site was ever resolved to allow the tests to appear.
I believe it was resolved,
It was resolved for trunk only, not for release. I'm building "Debian-Sid" for both trunk and release -- with MPI configured -- for weeks now, but the results only show on trunk. Anyone interested can obtain the results by ftp, or I can send in email.
Release has at least one failure, btw.
Can you please send me the failure? Matthias
On Aug 24, 2009, at 9:28 AM, Eric Niebler wrote:
Jeremiah Willcock wrote:
As far as I can tell, all the branches/release merges we were waiting for have now been completed. The plan is to go ahead with the release candidate as soon as all the release branch tests have cycled for revision 55757.
Please commit no more changes to branches/release without explicit permission from me. One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before
On Mon, 24 Aug 2009, Beman Dawes wrote: the release as well. What are other peoples' opinions on that?
Someone would need to step forward to run them. If someone did that Soon (in the next few days), I'm pretty sure we would be happy to hold the release for a test cycle. Do we have any volunteers?
Eric - I can run a darwin test cycle tonight, if that helps. -- Marshall
Marshall Clow wrote:
On Aug 24, 2009, at 9:28 AM, Eric Niebler wrote:
Jeremiah Willcock wrote:
One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before the release as well. What are other peoples' opinions on that?
Someone would need to step forward to run them. If someone did that Soon (in the next few days), I'm pretty sure we would be happy to hold the release for a test cycle. Do we have any volunteers?
Eric - I can run a darwin test cycle tonight, if that helps.
If you can get results for Boost.MPI and Boost.Graph.Parallel, then yes, it helps. Thanks, Marshall. -- Eric Niebler BoostPro Computing http://www.boostpro.com
On Mon, 24 Aug 2009, Eric Niebler wrote:
Marshall Clow wrote:
On Aug 24, 2009, at 9:28 AM, Eric Niebler wrote:
Jeremiah Willcock wrote:
One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before the release as well. What are other peoples' opinions on that?
Someone would need to step forward to run them. If someone did that Soon (in the next few days), I'm pretty sure we would be happy to hold the release for a test cycle. Do we have any volunteers?
Eric - I can run a darwin test cycle tonight, if that helps.
If you can get results for Boost.MPI and Boost.Graph.Parallel, then yes, it helps. Thanks, Marshall.
On a new enough Mac OS X version (I believe 10.5 is the first one), you should only need to add "using mpi ;" to your user-config.jam and do not need to install any additional software. If you could also try "using expat ;", that would also be nice; I believe that is installed by default as well. Thank you for running these tests. -- Jeremiah Willcock
Jeremiah Willcock wrote:
On Mon, 24 Aug 2009, Eric Niebler wrote:
Marshall Clow wrote:
On Aug 24, 2009, at 9:28 AM, Eric Niebler wrote:
Jeremiah Willcock wrote:
One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before the release as well. What are other peoples' opinions on that?
Someone would need to step forward to run them. If someone did that Soon (in the next few days), I'm pretty sure we would be happy to hold the release for a test cycle. Do we have any volunteers?
Eric - I can run a darwin test cycle tonight, if that helps.
If you can get results for Boost.MPI and Boost.Graph.Parallel, then yes, it helps. Thanks, Marshall.
On a new enough Mac OS X version (I believe 10.5 is the first one), you should only need to add "using mpi ;" to your user-config.jam and do not need to install any additional software. If you could also try "using expat ;", that would also be nice; I believe that is installed by default as well. Thank you for running these tests.
using expat ; will not work, since there's no tools/build/v2/tools/expat.jam, or expat.jam anywhere else. I think you need to set some variables with -s, but I don't exactly remember which and what documentation should describe that. - Volodya
On Aug 24, 2009, at 10:05 AM, Eric Niebler wrote:
Marshall Clow wrote:
On Aug 24, 2009, at 9:28 AM, Eric Niebler wrote:
Jeremiah Willcock wrote:
One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before the release as well. What are other peoples' opinions on that?
Someone would need to step forward to run them. If someone did that Soon (in the next few days), I'm pretty sure we would be happy to hold the release for a test cycle. Do we have any volunteers? Eric - I can run a darwin test cycle tonight, if that helps.
If you can get results for Boost.MPI and Boost.Graph.Parallel, then yes, it helps. Thanks, Marshall.
Sigh, I got bit by <https://svn.boost.org/trac/boost/ticket/2185> - hang while linking python tests. Trying again.... -- Marshall
On Aug 24, 2009, at 6:19 PM, Marshall Clow wrote:
On Aug 24, 2009, at 10:05 AM, Eric Niebler wrote:
Marshall Clow wrote:
On Aug 24, 2009, at 9:28 AM, Eric Niebler wrote:
Jeremiah Willcock wrote:
One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before the release as well. What are other peoples' opinions on that?
Someone would need to step forward to run them. If someone did that Soon (in the next few days), I'm pretty sure we would be happy to hold the release for a test cycle. Do we have any volunteers? Eric - I can run a darwin test cycle tonight, if that helps.
If you can get results for Boost.MPI and Boost.Graph.Parallel, then yes, it helps. Thanks, Marshall.
Sigh, I got bit by <https://svn.boost.org/trac/boost/ticket/2185> - hang while linking python tests. Trying again....
Turns out that it _finally_ finished (hours later). Results uploaded successfully.
Marshall Clow wrote:
On Aug 24, 2009, at 10:05 AM, Eric Niebler wrote:
Marshall Clow wrote:
On Aug 24, 2009, at 9:28 AM, Eric Niebler wrote:
Jeremiah Willcock wrote:
One issue that I am concerned about is that there are no regression test results for the release branch with MPI (and thus Boost.MPI and Boost.Graph.Parallel) turned on. Thus, neither of these components have been tested in a couple of weeks, and they were broken last time the regression tests ran for them. I would feel more comfortable if the regression tests passed for those before the release as well. What are other peoples' opinions on that?
Someone would need to step forward to run them. If someone did that Soon (in the next few days), I'm pretty sure we would be happy to hold the release for a test cycle. Do we have any volunteers? Eric - I can run a darwin test cycle tonight, if that helps.
If you can get results for Boost.MPI and Boost.Graph.Parallel, then yes, it helps. Thanks, Marshall.
Sigh, I got bit by <https://svn.boost.org/trac/boost/ticket/2185> - hang while linking python tests.
Got a backtrace? You might need to build bjam with debug information, using "build.sh --debug" for a useful backtrace. - Volodya
Vladimir Prus wrote:
Marshall Clow wrote:
Sigh, I got bit by <https://svn.boost.org/trac/boost/ticket/2185> - hang while linking python tests.
Got a backtrace? You might need to build bjam with debug information, using "build.sh --debug" for a useful backtrace.
It is better to get a strace output.
Ilya Sokolov wrote:
Vladimir Prus wrote:
Marshall Clow wrote:
Sigh, I got bit by <https://svn.boost.org/trac/boost/ticket/2185> - hang while linking python tests.
Got a backtrace? You might need to build bjam with debug information, using "build.sh --debug" for a useful backtrace.
It is better to get a strace output.
If OSX has strace, it would be helpful as well. - Volodya
participants (9)
-
Belcourt, Kenneth -
Beman Dawes -
Eric Niebler -
Ilya Sokolov -
Jeremiah Willcock -
Marshall Clow -
Matthias Troyer -
Steve M. Robbins -
Vladimir Prus