Segmentation fault problem

[Ajay Bhadran]

Hi all,
I am contacting you regarding a segmentation fault issue I encountered while running the cice+icepack configuration.
Specifically, I have configured the model with 15 layers of sea ice thickness and one layer of snow. However, when attempting to write internal snow temperatures to the history file, I am facing a segmentation fault problem. I am unable to pinpoint the root cause of the problem. It would be beneficial for me if anyone could offer suggestions or insights into where the issue might lie.

I have attached the ice_in file and the error message I received to provide more context.

 

[Philippe Blain]

Hello Ajay,

I've looked at your runlog and namelist. I tried running CICE6.1.0 (the version indicated in your runlog) with your namelist. It does not work outside the box because you seem to have custom forcing (NICE_2015) in NetCDF format. You also have two additions to the zbgc_nml namelist (Limiting_factors_file and Bottom_turb_mix.

I ran the model with these namellist modifications with respect to yours:

diff --git 1/ice_in.txt 2/ice_in
index da16f6a..e9acb60 100644
--- 1/ice_in.txt
+++ 2/ice_in
@@ -179,12 +179,12 @@ &forcing_nml
     ice_data_type   = 'default'
     fyear_init      = 2015
     ycycle          = 1
-    atm_data_format = 'nc'
-    atm_data_dir    = '/home/ajaybhadran/cice-dirs/input/peduro_data/NICE_2015/'
-    bgc_data_dir    = '/home/ajaybhadran/cice-dirs/input/peduro_data/NICE_2015/'
-    ocn_data_format = 'nc'
-    ocn_data_dir    = '/home/ajaybhadran/cice-dirs/input/peduro_data/NICE_2015/'
-    oceanmixed_file = 'oceanmixed_daily.nc'
+    atm_data_format = 'bin'
+    atm_data_dir    = '/path/to/CICE_data/forcing/gx3/NCAR_bulk'
+    bgc_data_dir    = 'unknown_bgc_data_dir'
+    ocn_data_format = 'bin'
+    ocn_data_dir    = '/path/to/CICE_data/forcing/gx3/'
+    oceanmixed_file = 'unknown_oceanmixed_file'
 /
 
 &domain_nml
@@ -351,8 +351,6 @@ &zbgc_nml
     F_abs_chl_sp       = 4.0
     F_abs_chl_phaeo    = 5.0
     ratio_C2N_proteins = 7.0
-    Limiting_factors_file = .true.
-    Bottom_turb_mix = .false.
 /
 
 &icefields_nml

and it ran correctly.

To try to debug your segfault, you should first recompile with debug symbols (-g) and without optimization (-O0). With these flags, the runlog should contain a useful backtrace when it segfaults, instead of just memory adresses like in your runlog above. This will tell you in which subroutine the code segfaults.

Since you are running with a single MPI task, you could also try running the model without MPI (set up the case with ./cice.setup -p 1x1 ...), and run it in a debugger, i.e. gdb --args ./cice, see section "Running with a Debugger" in the doc.

 

[Ajay Bhadran]

Philippe, thank you for clearing up my uncertainty. I'll review based on your advice.