Present were: Frank, Philipp, Ian, Erik, Peter, Christian Ott, Josh, Bruno, Roland
ET release: * there were still open tickets in review state. Decision made: ** will make timelevels parameters steerable ** will hold GetComponents changes since only "unusual" thorn lists trigger the bug ** slope limiters: leave as they are for the release ** hold patch to handle optional dependencies since it does not affect the distributed thorn lists ** all tickets cleared out now * please look at release notes once they are available * mention failing tests in release notes, warn about broken slope limiters * advertise release to same set of mailing list as last time * make sure to link full release notes on website
Testsuites: * there are a number of tests failing. It seems as if they fail on systems using Intel12 based compilers. Need to collect exact build number of compiler (icpc -V) to compare. * failing Exact/KS-tilted test is very likely due to RELTOL=0. Increase RELTOL to 1-12 * ADAxiOdd test failures are in a thorn that is known to produce different results each time it is called (tolerance is 1e-7) * many tests fail on Pandora, Peter is the goto person for this machine * SphericalHarmonicRecon/regression tests fail with 2MPI processes on some machines, mention in release notes * no tests for trestles are present since the machine was down during the last week
Misc: * grant Béla Szilágyi write access to PITTNULL code
Roland
On 28 May 2012, at 19:04, Roland Haas wrote:
Testsuites:
- there are a number of tests failing. It seems as if they fail on
systems using Intel12 based compilers. Need to collect exact build number of compiler (icpc -V) to compare.
Just to clarify: all tests pass on most machines, it's just that some fail on some machines.
- ADAxiOdd test failures are in a thorn that is known to produce
different results each time it is called (tolerance is 1e-7)
What I said in the call was slightly wrong -- it's not that IDAxiOddBrillBH produces different results each time it is run, it produces different results if the executable is different, even if the differences are unrelated to this thorn. See https://trac.einsteintoolkit.org/ticket/727.
On Mon, May 28, 2012 at 10:04:08AM -0700, Roland Haas wrote:
- please look at release notes once they are available
Hi,
Here they come:
Short: https://docs.einsteintoolkit.org/et-docs/Release_Announcement
Detailed: https://docs.einsteintoolkit.org/et-docs/Detailed_Release_Announcement
Please take a look and feel free to edit, append ect.
Frank
Hello all,
Detailed: https://docs.einsteintoolkit.org/et-docs/Detailed_Release_Announcement
* removed AEILocalInterp OpenMP claim (was reverted in r 56) * added option of reducing the number of active timelevels as second workaround for the timelevels-during-checkout issue (aka "AHFinderDirect fails during recovery" though it is not limited to AHFinderDirect) * we have known bugs in "Handle requirements recursively". Should we mention them? (ticket https://trac.einsteintoolkit.org/ticket/911) * we have (possible) known bugs in "CarpetMask: Keep track of the volume that is masked out": https://trac.einsteintoolkit.org/ticket/434 * we have known bugs (I claim) in "CarpetIOASCII: Add new 'compact' output format ": https://trac.einsteintoolkit.org/ticket/902
* we mention Pandora in the notes but I cannot find a simfactory entry for it (grep -ir pandora simfactory returns nothing).
Looks ok otherwise.
Yours, Roland
Hi,
On Mon, 28 May 2012, Roland Haas wrote:
Hello all,
Detailed: https://docs.einsteintoolkit.org/et-docs/Detailed_Release_Announcement
- removed AEILocalInterp OpenMP claim (was reverted in r 56)
- added option of reducing the number of active timelevels as second
workaround for the timelevels-during-checkout issue (aka "AHFinderDirect fails during recovery" though it is not limited to AHFinderDirect)
- we have known bugs in "Handle requirements recursively". Should we
mention them? (ticket https://trac.einsteintoolkit.org/ticket/911)
- we have (possible) known bugs in "CarpetMask: Keep track of the volume
that is masked out": https://trac.einsteintoolkit.org/ticket/434
- we have known bugs (I claim) in "CarpetIOASCII: Add new 'compact'
output format ": https://trac.einsteintoolkit.org/ticket/902
- we mention Pandora in the notes but I cannot find a simfactory entry
for it (grep -ir pandora simfactory returns nothing).
That's because there is no simfactory entry for it yet. I have an as yet incompete entry (misses tested submission and run scripts) that I have used for the tests. Once the machine becomes stable and useful, we'll add the entry.
Looks ok otherwise.
Yours, Roland
Cheers,
Peter
I'd like to commit the changes for ticket #768 (see patch 11 on that ticket).
This would change the way Cactus builds to use a per thorn include mechanism rather than a per thorn define mechanism. It also removes the "dependency fixer" from the build system.
Frank suggested that this would be a good time to do the commit, and that I should ask if anyone objects. Does anyone?
Cheers, Steve
Hello all,
I'd like to commit the changes for ticket #768 (see patch 11 on that ticket).
The ticket was in "new" state. I changed it to review. Did anyone actually review it so far? Erik seems to have looked at it by there is no "Please apply" yet.
Yours, Roland
On 05/31/2012 02:57 PM, Steven R. Brandt wrote:
I'd like to commit the changes for ticket #768 (see patch 11 on that ticket).
This would change the way Cactus builds to use a per thorn include mechanism rather than a per thorn define mechanism. It also removes the "dependency fixer" from the build system.
Frank suggested that this would be a good time to do the commit, and that I should ask if anyone objects. Does anyone?
Cheers, Steve
I've heard no objections, and I've addressed the issues raised by Erik. Shall I go ahead?
Cheers, Steve
users@lists.einsteintoolkit.org
-
Frank Loeffler -
Ian Hinder -
Peter Diener -
Roland Haas -
Steven R. Brandt