Skip to content

Conversation

davegill
Copy link
Owner

No description provided.

This brings in the new GNU compiler. If the standard GNU compiler
on Linux is too old to build WRF, this is what is required. The
side effect is that the "yum install netcdf" commands are now
generating inconsistent Fortran *.mod files. The "yum install"
command is defaulting to an assumed GNU version (4.8.5, OLD).
With some source changes in WRF, there is no longer a need for
the Fortran MODULE "use netcdf" statment, we can get away with
"include netcdf.inc".
Dockerfile has GNU 8 (fixes RRTMG compile fails with -d)
Dockerfile builds netcdf, hdf5, openmpi
Dockerfile fixes env and various LD_LIB and PATHs

Dockerfile and .travis.yml have option for regtest arg

Includes small data option (No DA, no HWRF, No KPP)
Includes better script with nml test number on pass
Includes better script with defined $d on run failures
Includes better script with stacksize unlimited

Takes a while to build the image now, 40+ minutes, likely
due to the building of those !@#$%^&* libraries from source

Changes to be committed:
	modified:   .travis.yml
	modified:   Dockerfile
	modified:   Dockerfile-template
… make install

Stick files in /wrf/wrfoutput. The output was enormous, and travis
complained. If the builds work, who looks at that stuff anyways, right?

Changes to be committed:
	modified:   Dockerfile
	modified:   Dockerfile-template
Changes to be committed:
	modified:   Dockerfile
	modified:   Dockerfile-template
…top the build

This is due to the travis requirement that the std out and err output from the
configure and make of the external libs (netcdf, hdf5, openmpi) was too large.
So, you stick the std out and err in a file to be a nice guy. Is travis happy?
No. Now it sees no output for a long time (since the three libs take a while) and says
"Hey, I'll bet this process is just hung".

Travis wants to see output every 10 minutes. Instead of modifying the .travis.yml
file, I tried to manufacture output after each of the configures (4 of them) and
after each of the makes (4 of them). Fingers crossed.

Sheesh.

Changes to be committed:
	modified:   Dockerfile
	modified:   Dockerfile-template
Trying to dump piles of print for openmpi make

Changes to be committed:
	modified:   Dockerfile
	modified:   Dockerfile-template
module_ra_rrtmg_swf.f90:3419:0:
       use rrsw_kg21_f, only : absa, ka, absb, kb, forref, selfref, &
internal compiler error: in gfc_trans_use_stmts, at fortran/trans-decl.c:5111
Please submit a full bug report,

Removing -d, seems to mostly be OK then with RRTMG fast sw then.
But on Darwin and cheyenne, the GNU 8 fixed this.

Also, adding -j 3 for the parallel build.

Changes to be committed:
	modified:   .travis.yml
Modified the associated script.csh to always bypass the
RRTMG FAST build

Changes to be committed:
	modified:   .travis.yml
davegill added 20 commits April 20, 2019 15:00
1. Has build WRF and repeat if necessary script.csh
 Changes to be committed:
	modified:   .travis.yml
1. cenv gets {}, not required, but conventional for script
2. zap entire "addons" with the "apt" "ubuntu" packages" and "gcc" stuff
3. remove language=fortran, never used sine we are doing containers
4. make branches of interest: master, release-v*, and develop
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant