search

CDAT / cdtime

climate calendar manipulation tools
0 stars 2 forks source link

Fixes extension config #51

Closed jasonb5 closed 3 years ago

jasonb5 commented 3 years ago

Cleans up the extension config.

chengzhuzhang commented 3 years ago

Hey @jasonb5, it was bought up by @gleckler1 that, PMP's CI tests experienced issue with this change. https://app.circleci.com/pipelines/github/PCMDI/pcmdi_metrics/479/workflows/03633821-dbaa-47ab-94ba-249399d832e6/jobs/10025?invite=true#step-106-52 I know this PR is waiting for a review from @downiec, but the conda-forge version has merged a patch of this fix. I'm wondering if this is related to PMP's CI problem?

It also looks like we need a new cdms2 build with new libnetcdf4.8.0. I will follow up with a separate email on this...

jasonb5 commented 3 years ago

@chengzhuzhang I can take a look at that in the next couple days. It's possible this change might have something to do with it.

jasonb5 commented 3 years ago

@chengzhuzhang @gleckler1 You can try rebuilding pcmdi_metrics again looks like cdms2 needed to be rebuild for libnetcdf 4.8.0.

chengzhuzhang commented 3 years ago

@jasonb5 Thanks a lot!
@gleckler1 Hope this can fix the PMP CI test.

lee1043 commented 3 years ago

@jasonb5 thanks for checking on this. Should I wait for this PR to be merged before rebuilding pcmdi_metrics?

gleckler1 commented 3 years ago

@jasonb5 @lee1043 @chengzhuzhang Thanks for looking at this. We are still experiencing the one failure ci/circleci: test-linux-3.6-nompi — Your tests failed on CircleCI

but maybe we need to wait until your PR51 is merged to expect this to work?

jasonb5 commented 3 years ago

@lee1043 @gleckler1 You can rebuild now the new cdms2 package came out earlier today. I'll follow up first thing tomorrow with an email.

jasonb5 commented 3 years ago

@downiec Sounds good i'll merge. Might have some more fixes coming out soon for these libraries, i'll be making a PR for cdms2 tomorrow.

chengzhuzhang commented 3 years ago

Hey @jasonb5, It looks like both E3sm-diags and ARM-diags have similar issues with CDMS2 file IO. This might be relevant. Let me know if it is ready to try re-build. Thank you for taking care of this at this moment.

lee1043 commented 3 years ago

Many thanks to @jasonb5 and all for following up on this issue.

It looks like we still have same issue with the PMP's circle ci test, related to CDMS2 file IO, as similar to @chengzhuzhang pointed. I see there is libnetcdf 4.8.0 included in the circle ci testing environment. It only fails with linux-python3.6 test envs, as @gleckler1 pointed.

Any insight would be much appreciated. Thank you.