explore clashing reqs in environment.yml

[ilaflott]

TLDR summary

nccmp is still non-resolvable with cmor>=3.14.2 in the same environment.
cdo==2.5.0 is now resolvable with cmor>=3.14.2 thanks to additional re-builds with libnetcdf and hdf5

what's this for

double-checking incompatibility that was discovered between cmor>=3.14.2 and cdo/python-cdo, and/or nccmp.

results

currently on main

the environment resolves

first commit

the environment resolution gets stuck when updating cmor>=3.14.0 to cmor>=3.14.2, failing to resolve after about an hour.

second commit

keeping cmor>=3.14.2, nccmp, cdo, and python-cdo are all commented out. in this scenario (in progress as of this edit), we expect that the environment will resolve, but that tests will fail due to missing requirements.
update: result is as expected, environment resolution, failing tests due to missing requirements.

third commit

cmor>=3.14.2+nccmp, but no cdo/python-cdo --> no environment resolution expected, no tests run
update: in-progress, takes awhile --> check later
update: result is as expected

fourth commit

cmor>=3.14.2+cdo/python-cdo, but no nccmp
UPDATE: Environment resolution worked! tests run and fail, which is expected after environment resolution

[ilaflott]

@ceblanton @singhd789 @bensonr @chanwilson

good news! cdo/python-cdo must have been updated, now resolves with cmor>=3.14.2 which is better than it was.

nccmp still blocks cmor>=3.14.2 unfortunately.

if we can work out the last snag, we can begin integrating fremor into fre-cli ala the CatalogBuilder model for the click bindings

[ilaflott]

delightfully not crazy, the screenshots show that `2.5.0` is getting updated in-place on `conda-forge`

https://github.com/NOAA-GFDL/fre-cli/actions/runs/26179000175/job/77016580306?pr=896#step:6:25 shows that we are using 2.5.0

obvious question- why is cdo not bumping their version #'s?

[ilaflott]

related to Try2Code/cdo-bindings#59 which is just where the python-cdo bindings are

we're still using python-cdo==1.6.1, https://github.com/NOAA-GFDL/fre-cli/actions/runs/26179000175/job/77016580306?pr=896#step:6:4613

python-cdo has NOT been updated (in-place or otherwise) in six months, see screenshot

[ilaflott]

nccmp has NOT been updated (in-place or otherwise) in six months. see screenshot

[ilaflott]

TLDR cdo==2.5.0 is not the same package from six months ago.

note that dependency pinning does not defend against something like that alone.

[ilaflott]

cdo is still updating to versions beyond 2.5.0: https://code.mpimet.mpg.de/projects/cdo/news

they are on version 2.6.*. I now suspect that cdo has a bug in their conda-forge/cdo-feedstock, potentially

[ilaflott]

... lightbulb moment... cdo doesn't have a github but they must have a feedstock! i can ping them on here!

[ilaflott]

cdo's in-place update of 2.5.0 is by-design and intentional, see conda-forge/cdo-feedstock#174