Open samsrabin opened 9 months ago
Note, this starts showing up in ctsm5.3.003 and after for SMS_Lm3_D_Mmpi-serial.1x1_brazil.I2000Clm50FatesCruRsGs.izumi_intel.clm-FatesColdHydro because of a similar addition to shell_commands for it. It worked for @glemieux for ctsm5.3.003 because of his startup shell options.
Brief summary of bug
The Izumi
FatesColdTwoStream
andFatesColdTwoStreamNoCompFixedBioGeo
tests from aux_clm fail, if called withrun_sys_tests
, unless Python is loaded in a user's.bash_profile
(or presumably the respective profile files for other shells). The failure happens during theCREATE_NEWCASE
step, during the call ofshell_commands
.See this error reported by @slevis-lmwg during testing of during #2334, as well as subsequent messages from me.
General bug information
CTSM version you are using: ctsm5.1.dev164
Does this bug cause significantly incorrect results in the model's science? No
Configurations affected: n/a
Details of bug
The user can solve this by adding
module load lang/python/3.7.0
to their.bash_profile
, but it would be better to solve this permanently rather than relying on the user figuring this out.The problem is that
shell_commands
in these tests calls the script$FATESDIR/tools/modify_fates_paramfile.py
, which depends on numpy, scipy, etc. It would be more appropriate for that script to be called in therun_phase
of a new SystemTest—call itMODIFYFATESPARAMF
. This would make it so that, as long as the user callsrun_sys_tests
withctsm_pylib
loaded, everything necessary will be available. @glemieux, what do you think?Important details of your setup / configuration so we can reproduce the bug
On Izumi, in bash, with no
module load
commands in my.bash_profile
, I set up the failing test like so:Important output or errors that show the problem
TestStatus.log: