[ET Trac] [Einstein Toolkit] #1075: Test cases with specific number of processes
Einstein Toolkit
trac-noreply at einsteintoolkit.org
Thu Apr 25 09:54:47 CDT 2013
#1075: Test cases with specific number of processes
--------------------------+-------------------------------------------------
Reporter: eschnett | Owner:
Type: enhancement | Status: reopened
Priority: major | Milestone: ET_2013_11
Component: Cactus | Version:
Resolution: | Keywords:
--------------------------+-------------------------------------------------
Comment (by knarf):
Moving the release target again is unfortunate, I agree. It is not
critical, however. The release tag should only indicate that this "should
be fixed by that release".
It would be nice if issues with the test system would get more attention
though. I would think they should be critical. The past showed that while
most probably agree with that statement, reality tells differently. As for
the next release: I don't think whatever solution we might come up with
would make it in the release. There isn't enough time for it. I still
think this is important enough to keep reminding ourselves.
On this specific issue: the underlying root issue is that most testsuites
that set a specific number of processes run just fine on other numbers,
but their output can currently not be successfully compared by Cactus.
This could be fixed by either Cactus being able to compare the files, or
by having processor-number-independend output files. The former was
started by a student at LSU, but died after the student decided to stop
working on it (undergraduate...). The latter would be nice, but didn't
happen so far and especially performance implications are unclear. So, in
order to pick up one or the other again in the future we should keep this
ticket open - and it is important enough to keep reminding us about this
embarrassing issue.
--
Ticket URL: <https://trac.einsteintoolkit.org/ticket/1075#comment:35>
Einstein Toolkit <http://einsteintoolkit.org>
The Einstein Toolkit
More information about the Trac
mailing list