cdat using old libnetcdf about e3sm-unified HOT 8 CLOSED

@doutriaux1, I'm hoping you can help me with this. I suspect that the issue is that the current build of cdat and its dependencies is from 4 months ago and it is therefore using an outdated libnetcdf (not 4.6.1, as we require for the latest e3sm-unified). Would is be easy for you to rebuild the cdat packages with the latest conda-forge bindings?

from e3sm-unified.

@doutriaux1, hold off on this for a moment. Everything seems to be fine on Edison and Cori, which suggests that I may just have messed things up on my laptop when I tested just now.

from e3sm-unified.

@doutriaux1, I am seeing the same issue on acme1 and aims4, so it really does seem to be an issue. I'm not sure why it works on edison and cori, but my guess would be that libnetcdf.so.11 is in my library path somewhere, but not because of the anaconda environment.

from e3sm-unified.

I think I may have found a solution. If I use cdms2 v. 3.0.1 instead of 3.0, it comes from conda-forge instead of cdat and it seems to work (at least the following works):

> conda create -n test_vcs -c cdat -c conda-forge python=2 vcs mesalib libnetcdf=4.6.1 hdf5=1.10.2
...
The following NEW packages will be INSTALLED:
...
    cdms2:                3.0.1-py27h6091dcd_1                  conda-forge
...

> conda activate test_vcs
> python -c "import vcs"

from e3sm-unified.

@doutriaux1, does that seem like the right solution to you?

from e3sm-unified.

@xylar yes this is the way to go, we went through great pains to recompile everything on conda-forge with the newest compilers. You will need cdms2 3.0.1 and for the cdtime, libsdrs, libcdms, etc... Make sure you get the latest build of the latest version. We have nightly builds that tests these things so I'm confident it should work.
@zshaheen can help too, he's got a specific E3SM-DIAGS yml that dealt with tihs a week or so ago.

from e3sm-unified.

Thanks! Let's move the discussion to #19.

from e3sm-unified.

addressed by #19

from e3sm-unified.