You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
/discover/swdev/gmao_SIteam/MPI/openmpi/5.0.2-SLES15/gcc-13.2.0/bin/mpirun -np 96 /discover/nobackup/mathomp4/Experiments/mapldev-2024Mar21-1day-c24-GCC13-Debug/scratch/GEOSgcm.x --logging_config logging.yaml
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.
Backtrace for this error:
with no usable traceback.
However, if you do not pass in --logging_config logging.yaml, it works:
/discover/swdev/gmao_SIteam/MPI/openmpi/5.0.2-SLES15/gcc-13.2.0/bin/mpirun -np 96 /discover/nobackup/mathomp4/Experiments/mapldev-2024Mar21-1day-c24-GCC13-Debug/scratch/GEOSgcm.x
MAPL: No configure file specified for logging layer. Using defaults.
Starting pFIO input server on Clients
Starting pFIO output server on Clients
MAPL: Running with MOAB library for ESMF Mesh: F
SHMEM: NumCores per Node = 96
...
This can be duplicated with ExtDataDriver.x and bringing in the logging.yaml from GEOSgcm. No flag:
$ ./ExtDataDriver.x
MAPL: No configure file specified for logging layer. Using defaults.
Starting pFIO input server on Clients
Starting pFIO output server on Clients
Running new case
SHMEM: NumCores per Node = 1
...
with flag:
$ ./ExtDataDriver.x --logging_config logging.yaml
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.
Backtrace for this error:
#0 0x148e69f41dbf in ???
#1 0x0 in ???
Segmentation fault
So just adding that option, boom.
Note, though that the -h flag doesn't cause this:
$ ./ExtDataDriver.x -h
usage: executable [-h][--root_dso ROOT_DSO][--esmf_logtype ESMF_LOGTYPE][--egress_file EGRESS_FILE][--cap_rc CAP_RC][--npes_model NPES_MODEL][--n_members N_MEMBERS][--use_sub_comm][--comm_model COMM_MODEL][--prefix PREFIX][--npes_input_server][--npes_output_server][--nodes_input_server][--nodes_output_server][--logging_config LOGGING_CONFIG][--oserver_type OSERVER_TYPE][--npes_backend_pernode NPES_BACKEND_PERNODE][--compress_nodes][--fast_oclient][--one_node_output][--with_io_profiler][--with_esmf_moab]
optional arguments:
-h, --help This message.
--root_dso name of root dso to use
--esmf_logtype ESMF Logging type (allowed: none, single, multi, multi_on_error)
...
Nor does, say, --esmf_logtype:
$ ./ExtDataDriver.x --esmf_logtype multi
MAPL: No configure file specified for logging layer. Using defaults.
Starting pFIO input server on Clients
Starting pFIO output server on Clients
Running new case
SHMEM: NumCores per Node = 1
...
So it seems to be an interaction with fArgParse. I know @bena-nasa and myself traced it (with prints) to a certain call. I just need to (re-)add the prints and get to it.
The text was updated successfully, but these errors were encountered:
mathomp4
changed the title
MAPL/Fargparse Fails with GCC 13 with logging command line option
MAPL/Fargparse/pFlogger Fails with GCC 13 with logging command line option
Mar 21, 2024
GEOSgcm + GCC 13 dies instantly in a default run:
with no usable traceback.
However, if you do not pass in
--logging_config logging.yaml
, it works:This can be duplicated with
ExtDataDriver.x
and bringing in thelogging.yaml
from GEOSgcm. No flag:with flag:
So just adding that option, boom.
Note, though that the
-h
flag doesn't cause this:Nor does, say,
--esmf_logtype
:So it seems to be an interaction with fArgParse. I know @bena-nasa and myself traced it (with prints) to a certain call. I just need to (re-)add the prints and get to it.
The text was updated successfully, but these errors were encountered: