mpi-serial nag case fails with pio2 when defining h1 history restart file #1030
Labels
bug
something is working incorrectly
priority: high
High priority to fix/merge soon, e.g., because it is a problem in important configurations
Brief summary of bug
The test
ERS_Ly5_Mmpi-serial.1x1_numaIA.I2000Clm50BgcCropQianGs.izumi_nag.clm-monthly
fails when using pio2 when defining the h1 history restart file.General bug information
CTSM version you are using:
ctsm1.0.dev093-17-ge55f7451
; also suspected to be an issue on ctsm master if cime is switched to a tag on master (cime5.8.23
) rather than pointing to the cime branch that we're currently using (which was created to get around this issue) (cime master uses pio2 for this test, whereas the cime branch uses pio1).Does this bug cause significantly incorrect results in the model's science? No
Configurations affected: Only observed for this one test; unknown if it is a more general issue
Details of bug
The above-referenced test dies when writing the restart files, with:
The job output file indicates a segmentation fault.
I introduced the following diffs to get a better sense of where it's dying:
From this I got:
implying that the death occurs somewhere in:
I tried running a similar case in DEBUG mode. It died in the same place (i.e., after
htape_create : Successfully defined netcdf restart history file 2
; I didn't have all the extra write statements from the above diff, so I'm not sure exactly where it died, but the output upon death was the same as in the non-debug case prior to adding the additional write statements), and didn't give any additional useful information.The text was updated successfully, but these errors were encountered: