Closed lqxyz closed 3 years ago
@lqxyz Thanks for finding this bug! May I ask why the CI tests failed? For inclusion of this PR into the master branch, these tests must pass.
The following is the error message from CI test, and it seems due to the NetCDF library for ifort
.
test (ifort):
Build NetCDF FORTRAN library
checking for a thread-safe mkdir -p... /bin/mkdir -p
Run source /opt/intel/inteloneapi/setvars.sh || true
/home/runner/work/_temp/ac9c91f0-2733-494f-aa2a-3a9712fbcd5a.sh: line 1: /opt/intel/inteloneapi/setvars.sh: No such file or directory
I've compared the outputs of the ifort step with a previous successful test, and it looks like the CI test retrieves the latest ifort version, and therefore we can be affected by changes in the ifort configuration (e.g. /opt/intel/inteloneapi/setvars.sh doesn't exist in the latest ifort or it's in a different place). This is a problem of robustness of the CI test that we need to fix.
I have looked at the Intel OneAPI documentation, and it seems that setvars.sh is now located in /opt/intel/oneapi. However, changing this doesn't fix the ifort test, it still fails with an error in configure:
configure: error: Could not link conftestf.o and conftest.o
I have no idea why this happens. I'll create an issueto document this.
@lqxyz I've fixed the CI tests in #55 . You should be able to continue with the development/testing of this change.
Thank you very much @alejandrobodas. I wonder how could I restart the CI test for this P/R?
Merge the changes from branch master
onto your branch. This will require a commit, which will fire up another round of CI.
Thank you @RobertPincus. I have merged my branch with the master
as you suggested, and now all the CI tests have passed. It looks like this P/R can be merged into master
after some reviews.
@dustinswales It looks good to me but maybe you can check and formally approve? This is really in the guts of the thing...
I approve of these changes.
In
cosp_config.F90
, the liquidReff
related variables should useLIQ
rather thanICE
, right? I notice thatnReffICE
andnReffLIQ
are equal, so the misuse of them won't cause any problems. However,reffICE_binCenters
andreffLIQ_binCenters
are different, which needs to be corrected.