Open glemieux opened 3 years ago
It seems to me that the primary reason we had non-FATES tests in the fates testlist was to make sure that non-FATES tests continue to work fine. However, now normally the fates test list is run in conjunction with the aux_clm tests so maybe we don't need non-FATES tests in the fates test list anymore?
I think it makes sense to still keep a clm-only test in the fates list for when we're testing things that don't require an api update, that way we catch potential issues that a fates-only update might screw up something in ctsm. Our current procedure is to only run the aux_clm
tests when we have an api update iirc.
That said, I think having coverage for clm-only in the fates suite is probably only necessary for production compilers (i.e. not nag) since aux_clm
on izumi covers that.
I have only skimmed this, but a couple of notes:
Brief summary of bug
ERS_D_Ld5.f19_g17.I2000Clm50BgcCru.izumi_nag.clm-default
will not complete its run due to hitting the wall clock limit.General bug information
CTSM version you are using: ctsm5.1.dev057-59-g282920abd
Does this bug cause significantly incorrect results in the model's science? No
Configurations affected:
Details of bug
In the course of updating PR #1275 to provide nag coverage on izumi for all debug tests in the
fates
category, I included the one non-fates testmod as part of the list via https://github.com/ESCOMP/CTSM/pull/1275/commits/5f3a8228a35caf6b340706caaf1e43804e763294. Although this test is short, it never completes. Only the first set of logs are available and are not compressed. Examining the logs it appears that either the clm or mosart files might be getting hung up during one of the final writes, although all the expected output files do appear to be present when comparing the list to a successful run of the same test on cheyenne.For the time being this test has been dropped from the fates category on izumi for the nag compiler since we have coverage elsewhere for a different production compiler.