Can't build on edison

[mfwehner@lbl_gov]

My cam5 aquaplanet script (cesm1_2_2) that successfully built on edison.nersc.gov on February 14, 2015 now fails to compile on March 3, 2015. Probably due to some upgrade that I missed.The error message in the log file is /global/u1/m/mwehner/cesm1_2_2/models/drv/driver/ccsm_comp_mod.F90(46): error #7002: Error in opening the compiled module file.  Check INCLUDE paths.   [OCN_COMP_MCT]
   use ocn_comp_mct, only: ocn_init_mct, ocn_run_mct, ocn_final_mct
-------^
/global/u1/m/mwehner/cesm1_2_2/models/drv/driver/ccsm_comp_mod.F90(46): error #6580: Name in only-list does not exist.   [OCN_INIT_MCT]
   use ocn_comp_mct, only: ocn_init_mct, ocn_run_mct, ocn_final_mct
---------------------------^
/global/u1/m/mwehner/cesm1_2_2/models/drv/driver/ccsm_comp_mod.F90(46): error #6580: Name in only-list does not exist.   [OCN_RUN_MCT]
   use ocn_comp_mct, only: ocn_init_mct, ocn_run_mct, ocn_final_mct
-----------------------------------------^
/global/u1/m/mwehner/cesm1_2_2/models/drv/driver/ccsm_comp_mod.F90(46): error #6580: Name in only-list does not exist.   [OCN_FINAL_MCT]
   use ocn_comp_mct, only: ocn_init_mct, ocn_run_mct, ocn_final_mct
------------------------------------------------------^
/global/u1/m/mwehner/cesm1_2_2/models/drv/driver/ccsm_comp_mod.F90(1130): error #6406: Conflicting attributes or multiple declaration of name.   [OCN_INIT_MCT]

 Suggestions?Michael 

 

[jedwards]

Hi Michael,There must be an error earlier in the build of file ocn_comp_mct.   Check the atm and aquap buildlogs.

 

[mfwehner@lbl_gov]

JimIt does not appear to be in the atm log. And I see no aquap log. Here is the directory, permissions should be ok/scratch1/scratchdirs/mwehner/seCAM5v2_2_AQUAPLANET_plus2K/bld 

 

[jedwards]

Have you tried $CASE.clean_build and $CASE.build?The directory /scratch1/scratchdirs/mwehner/seCAM5v2_2_AQUAPLANET_plus2K/bld/ocn/obj appears to be corrupted, doing a clean_build before building should solve the problem.

 

[mfwehner@lbl_gov]

Jimfor me a clean build is rm -rf $CASEROOTand the error is repeatableM

 

[jedwards]

Can you fix file permissions in /scratch1/scratchdirs/mwehner/seCAM5v2_2_AQUAPLANET_plus2Kso that I can read them.

 

[eaton]

FYI, I'm able to build cam5/aquaplanet from cesm1_2_2 using a standalone script.  I'm using the default intel environment.  The only customization is "module load cray-netcdf".  My script is in ~eaton1/CASE/cesm1_2_2/xc30in-tests/f1.9c5aqP24T2-3s.sh.

 

[eaton]

Followup:  Although the build went fine I get seg faults when trying to run cesm1_2_2.  Haven't been able to track it down.  I am able to build and run the latest cam trunk code with cam5 physics in both aquaplanet and normal configs. 

 

[jedwards]

I am able to build and run ne16_ne16 FC5AQUAP using cesm1_2_2 on edison with the following change:

1. create the case
2. Replace env_mach_specific with the Edison file linked from https://bb.cgd.ucar.edu/list-envmachspecific-files-and-links-their-source-code
3. Edit the new env_mach_specific file and remove the xmlquery lines at the top.
4. Build and run.   

 

[eaton]

Thanks Jim.  I was testing FV and the indication from the log file was that the seg fault was coming from the dycore.  I am also able to run this configuration successfully at ne16.  In the FV runs when I turn debug on the run works.  So I suspect it's a compiler bug.