input_subdir doesn't propagate from default to global_time_series

[tangq]

When input_subdir is specified to a directory other than archive/*, the global_time_series bash file still uses the archive directory as input_subdir and causes failures.

An example cfg file is at: /home/ac.qtang/E3SMv2/scripts/post.v2.LR.historical_0101_bonus.cfg

To reproduce the error, only need to remove the input_subdir setting in the global_time_series session from this cfg file.

[forsyth2]

Note: In #248, it was mentioned that "adding input_subdir = "run" in the global_time_series session should bypass the error reported" in this issue. However, this issue is for including any arbitrary input_subdir.

[forsyth2]

@tangq I'm not sure I'm clear on the problem. If you're using input_subdir = "run", that works. What directory are you trying to use besides archive/ or run/?

If it differs from the input path, you can also override input, but I see that might be complicated by https://github.com/E3SM-Project/zppy/blob/main/zppy/templates/ocean_month.py#L40 also using input.

[tangq]

The problem is that input_subdir = "run" is specified in the "default" session, but global_time_series.bash doesn't pick up the "run" directory when "input_subdir" is omitted in the global_time_series session.

I created an example without specifying "input_subdir" in the global_time_series session at /lcrc/group/e3sm/ac.qtang/E3SMv2/v2.LR.historical_0101_bonus/post/scripts/tst

It seems that the ocn input path in the global_time_series bash file will use input_subdir="archive/ocn/hist" when not specified in the global_time_series session and default has input_subdir="run".

[ac.qtang@chrlogin1 scripts]$ pwd
/lcrc/group/e3sm/ac.qtang/E3SMv2/v2.LR.historical_0101_bonus/post/scripts
[ac.qtang@chrlogin1 scripts]$ diff global_time_series_1975-2014.bash tst/global_time_series_1975-2014.bash 
71c71
<     input=/lcrc/group/e3sm/ac.qtang/E3SMv2/v2.LR.historical_0101_bonus/run
---
>     input=/lcrc/group/e3sm/ac.qtang/E3SMv2/v2.LR.historical_0101_bonus/archive/ocn/hist

[forsyth2]

@tangq Oh I see. This is a known issue and is resultant from #212. A fix for that would be a major change to zppy -- one we're still evaluating if it would make sense to implement. We even have a test that checks this inheritance behavior -- https://github.com/E3SM-Project/zppy/blob/main/tests/integration/test_defaults.cfg#L8.

Is specifying input_subdir in the global_time_series task sufficient for now? If so, I can close this issue as a duplicate of #212.

[tangq]

Yes, specifying input_subdir is good for now. I am closing this issue.