#!/bin/bash # # Run from test/unit/python/ as: # ./scripts/splitmpidebug.sh # # Configure number of repeats and number of mpi processes like this: # REPEAT=2 PROCS='3 4' ./scripts/splitmpidebug.sh fem book # # This script runs pytest in with mpi such that output from each process # is visible in separate xterm windows, and on a crash you can abort # and exit into gdb to inspect the stacktrace in each xterm separately. # # You can pass a list of test files or test modules as arguments to this script. # These tests are executed in separate pytest instances. # See also mpidebug.sh. # # ... FILES = list of test files or modules to try separately (default .) FILES=${@:-.} # ... Max number of times to run each test file REPEAT=${REPEAT:-1} # ... List of number of mpi processes to try PROCS=${PROCS:-3} echo echo Running $REPEAT times with $PROCS processes each of echo $FILES echo for f in $FILES do for p in $PROCS do n=1 # Loop at most $m times until file fails while [ $? -eq 0 -a $n -le $REPEAT ] do echo === Take $n, $p processes, `date`: $f n=$((n+1)) # Clean before each run to remove error sources from file system ./scripts/clean.sh # Run! mpirun -n $p xterm -e gdb -ex r -ex q -args python -B -m pytest -sv $f done done done