search

ESMCI / cime

Common Infrastructure for Modeling the Earth
http://esmci.github.io/cime
Other
161 stars 206 forks source link

Testing: Failed namelist build shows as SETUP PEND #4680

Closed samsrabin closed 2 weeks ago

samsrabin commented 2 weeks ago

I have 3 tests in /glade/derecho/scratch/samrabin/tests_0913-095333de that all have namelist build errors like this:

ERROR: Command /glade/work/samrabin/ctsm_hs-uii/bld/build-namelist failed rc=255
out=
err=Warning : CLM build-namelist::CLMBuildNamelist::setup_logic_initial_conditions() : WARNING: You have set use_hillslope while use_init_interp is TRUE.
 This means all hillslope columns in a gridcell will read identical values from initial conditions. If you are sure you want this behaviour:
 -- Add -ignore_warnings option to CLM_BLDNML_OPTS to ignore this warning'
  File "/glade/work/samrabin/ctsm_hs-uii/cime/CIME/test_scheduler.py", line 1125, in _run_catch_exceptions
    return run(test)
  File "/glade/work/samrabin/ctsm_hs-uii/cime/CIME/test_scheduler.py", line 1016, in _setup_phase
    "Fatal error in case.cmpgen_namelists: {}".format(output),
  File "/glade/work/samrabin/ctsm_hs-uii/cime/CIME/utils.py", line 176, in expect
    raise exc_type(msg)

However, cs.status makes it look like they're pending SHAREDLIB_BUILD:

0913-095333de_gnu: 1 test
  SMS_D_Mmpi-serial_Ld5.5x5_amazon.I1850Clm60Bgc.derecho_gnu.clm-HillslopeC (Overall: PEND) details:
    PASS SMS_D_Mmpi-serial_Ld5.5x5_amazon.I1850Clm60Bgc.derecho_gnu.clm-HillslopeC CREATE_NEWCASE
    PASS SMS_D_Mmpi-serial_Ld5.5x5_amazon.I1850Clm60Bgc.derecho_gnu.clm-HillslopeC XML
    PASS SMS_D_Mmpi-serial_Ld5.5x5_amazon.I1850Clm60Bgc.derecho_gnu.clm-HillslopeC SETUP
    PEND SMS_D_Mmpi-serial_Ld5.5x5_amazon.I1850Clm60Bgc.derecho_gnu.clm-HillslopeC SHAREDLIB_BUILD

0913-095333de_int: 2 tests
  ERP_D_P64x2_Ld10.f10_f10_mg37.I2000Clm60Bgc.derecho_intel.clm-Hillslope (Overall: PEND) details:
    PASS ERP_D_P64x2_Ld10.f10_f10_mg37.I2000Clm60Bgc.derecho_intel.clm-Hillslope CREATE_NEWCASE
    PASS ERP_D_P64x2_Ld10.f10_f10_mg37.I2000Clm60Bgc.derecho_intel.clm-Hillslope XML
    PASS ERP_D_P64x2_Ld10.f10_f10_mg37.I2000Clm60Bgc.derecho_intel.clm-Hillslope SETUP
    PEND ERP_D_P64x2_Ld10.f10_f10_mg37.I2000Clm60Bgc.derecho_intel.clm-Hillslope SHAREDLIB_BUILD
  SMS_D_Ld3.f10_f10_mg37.I1850Clm60Bgc.derecho_intel.clm-HillslopeFromFile (Overall: PEND) details:
    PASS SMS_D_Ld3.f10_f10_mg37.I1850Clm60Bgc.derecho_intel.clm-HillslopeFromFile CREATE_NEWCASE
    PASS SMS_D_Ld3.f10_f10_mg37.I1850Clm60Bgc.derecho_intel.clm-HillslopeFromFile XML
    PASS SMS_D_Ld3.f10_f10_mg37.I1850Clm60Bgc.derecho_intel.clm-HillslopeFromFile SETUP
    PEND SMS_D_Ld3.f10_f10_mg37.I1850Clm60Bgc.derecho_intel.clm-HillslopeFromFile SHAREDLIB_BUILD

It would be helpful for such tests to instead be marked as FAIL for SETUP, or maybe a different phase.

jedwards4b commented 2 weeks ago

I'm concerned that this will mark other namelist differences as fails - have you tested this possibility?

samsrabin commented 2 weeks ago

Good question, but it doesn't: see /glade/derecho/scratch/samrabin/tests_0913-145045de/SMS_D_Ld65.f10_f10_mg37.I2000Clm45BgcCropQianRs.derecho_intel.clm-FireLi2014Qian.C.0913-145045de_int. A namelist diff from baseline is correctly marked as SETUP PASS, NLCOMP FAIL.

samsrabin commented 2 weeks ago

(Assuming you meant to comment on #4681)