Ignore empty output bundle

[DusanJovic-NOAA]

Description

(Instructions: this, and all subsequent sections of text should be removed and filled in as appropriate.)
Fix model crash if one or the output bundles does not have any field.

What bug does it fix, or what feature does it add? #676
Is a change of answers expected from this PR? No

Issue(s) addressed

Link the issues to be closed with this PR, whether in this repository, or in another repository.
(Remember, issues should always be created before starting work on a PR branch!)

• fixes Model fails when output bundle contains no fields #676

Testing

How were these changes tested?
What compilers / HPCs was it tested with?
Are the changes covered by regression tests? (If not, why? Do new tests need to be added?)
Have the ufs-weather-model regression test been run? On what platform?

• Will the code updates change regression test baseline? If yes, why? Please show the baseline directory below.
• Please commit the regression test log files in your ufs-weather-model branch

Dependencies

N/A

[junwang-noaa]

Thanks for the fix, the code changes look good to me. So now if quilt is on and diag_table is empty, the model will not crash, right?

[DusanJovic-NOAA]

Thanks for the fix, the code changes look good to me. So now if quilt is on and diag_table is empty, the model will not crash, right?

diag_table file can not be empty (like zero size), it must have at least these two lines at the beginning, for example:

20210322.06Z.C96.64bit.non-mono
2021 03 22 06 0 0

but it does not need to have any 'files' or 'fields' specifications.

Also, post must be turned off, unless we add some test to detect that both dyn and phy bundles are not empty, and that both dyn and phy bundles have all necessary fields that post needs. But such a list does not exist, so I do not know how we can check that all post fields are present in both bundles.

[junwang-noaa]

Yes, what I mean is that diag_table does not have any output fields.
Can we set dopost to false if all the fieldCount from history bundles are zero? Other than that, I think at this time it's the developers' job to use correct post control file if the history files do not contain enough model fields to produce post products.

[DusanJovic-NOAA]

Yes, what I mean is that diag_table does not have any output fields. Can we set dopost to false if all the fieldCount from history bundles are zero? Other than that, I think at this time it's the developers' job to use correct post control file if the history files do not contain enough model fields to produce post products.

Where should we add such a test? Post gets the fields from output bundles in set_postvars_fv3 routine based on which grid_id it is processing at the moment and whether the bundle is a 'restart' bundle or non-restart bundle. Checking the number of fields in set_postvars_fv3 is probably too late. Adding the check somewhere in the write grid component will require non-trivial logic that loops over all bundles, skip all restart bundles, skip all native grid bundles (because post can nor read native grid fields), and finally skip all bundles on grids (domains) other than one currently being processed. I'm not sure adding all this logic is worth it. Especially since the criteria whether a bundle is empty or not is not enough to guarantee that the post will have access to all required fields.

[junwang-noaa]

OK. I agree with you.
@WenMeng-NOAA just let you know that if developers do not include output fields in the diag_table, they need to turn off inline post (write_dopost: .false.). Model will not check inline post setting from the content in diag_table, it will make model code too messy.

[zach1221]

Good morning, @BrianCurtis-NOAA . Testing is complete on #1825 . Can you please merge this PR for us?