This is a new problem with the latest CIME, that ensures that an expected file isn't a directory. When running single point compsets we commonly turn off MOSART, but use a compset with it in place. So this doesn't actually matter, but it's now causing tests to die with the new cime.
With the latest cime master, there was a bug fix in check_inputdata for testing if a inputdata file is a file. So there is a failure in mosart.input_data_list where
frivinp_rtm = /glade/p/cesmdata/cseg/inputdata/
It looks like mosart isn't setting a file for single column clm.
This is a new problem with the latest CIME, that ensures that an expected file isn't a directory. When running single point compsets we commonly turn off MOSART, but use a compset with it in place. So this doesn't actually matter, but it's now causing tests to die with the new cime.
Here's the message from @fischer-ncar....