search

NOAA-EMC / RDASApp

Regional DAS
GNU Lesser General Public License v2.1
0 stars 5 forks source link

meet the data ownership requirement by the mpas-jedi tests with change on mpas-jedi-data to be copied into RDASapp/bundle dir #21

Closed TingLei-NOAA closed 1 month ago

TingLei-NOAA commented 1 month ago

MPAS-jedi apps require write permission to some template files, which are from mpas-jedi-data. This PR changes the symlink of mpas-jedi-data from a specified machine-dependent location (which might not be owned by the user and cause all mpas-jedi tests fail) to "copy" of that dir to the current RDASapp/bundle dir to resolve this issue. With this PR, all mpas-jedi tests passed except for :

The following tests FAILED:
    1375 - test_mpasjedi_3denvar_amsua_allsky (Failed)
    1376 - test_mpasjedi_3denvar_amsua_bc (Failed)
    1381 - test_mpasjedi_4denvar_VarBC (Failed)
    1382 - test_mpasjedi_4denvar_VarBC_nonpar (Failed)

which were attributed to the use of different crtm data .

delippi commented 1 month ago

Hi @TingLei-NOAA, so before mpas-jedi-data/ would have been under build/ and then symlinked to bundle/. What you are proposing is to just put mpas-jedi-data/ directly in bundle/ due to some instance of failures due to file permissions if the user doesn't own the copy of build/mpas-jedi-data/. Is that correct? Shouldn't each user just download their own test data and wouldn't that fix the problem? If the problem is different from that, then I still think we should keep the "build/mpas-jedi-data/ and symlinked to bundle/" configuration for consistency and find a cleaner solution for this particular use case. Perhaps just manually make the copy yourself?

TingLei-NOAA commented 1 month ago

@delippi before, the mpas-jedi-data in bundle dir would be the symlink of the one in Cory's dir, which we (except for Cory) have no write permission and this PR would copy Cory's mpas-jedi-data to under bundle ( it would still be linked to that under build). Yes, allowing users to clone mpas-jedi-data would also be a solution. but now, I 'd like to stick to the practice of using local version to reduce the downloading burden.

delippi commented 1 month ago

@TingLei-NOAA Okay, I see what the issue is now. I think that seems reasonable. Should we do the same with the other *-data directories?

TingLei-NOAA commented 1 month ago

@delippi Seems only mpas-jedi app needs write permission to its "input" data (some template files, i think). For being now, this PR would allow all ctests using various data work.