attempts to write to stderr from openmpi are causing sys_write to go through the pipe code path. Since kitten expects a valid mutex for pipes I am getting a kernel fault when the uninitialized mutex is locked.
I've hacked around it in the short term (redirecting stderr to stdout) but I will fix it after MPI is working multinode.
attempts to write to stderr from openmpi are causing sys_write to go through the pipe code path. Since kitten expects a valid mutex for pipes I am getting a kernel fault when the uninitialized mutex is locked.
I've hacked around it in the short term (redirecting stderr to stdout) but I will fix it after MPI is working multinode.