mpirun comes from config_machines.xml rather than case env_mach_specific.xml

[ekluzek]

This is related to #643. As a user I can't easily customize the mpirun command in a case since the The mpirun command to use for a case comes from the cime cime_config/$MODEL/machines/config_machines.xml rather than the env_mach_specific.xml that is in my $CASEDIR. However, env_mach_specific.xml has a copy of the mpirun settings so it appears to the user that you can change the settings, even though you can't. So the current configuration is at least deceptive, since you can change and add environment variables and module settings in env_mach_specific.xml and those changes do happen.

So I'd like to see that the mpirun settings comes from env_mach_specific.xml rather than from config_machines.xml. If there is a reason for this behavior, then at least the mpirun settings should be removed from the local env_mach_specific.xml.

There are many reasons a user might want to customize the mpirun command, to run through debuggers, or profilers, or add special options to mpirun itself, or a whole host of things. And there are a whole host of reasons that you might want this to apply to one case, but not to all.

This is at least with cime5.1.1.

[ekluzek]

As of cime5.2.0.alpha.9 in get_mpirun_cmd in case.py the mpirun command is fetched from env_mach_specific.xml so it now works as I'd like it to. This also true in the very latest version on master right now. So I'm going to close this issue.