Thanks @hawartens for your PR.

I'm seeing that your code change is actually adding this little "magic" and also adding maxrc as new clush.conf option. This is probably worthing adding in the commit or make 2 commits of that.

I'm not against adding this option in config file is you think this is useful, but it seems you did this to propagate the maxrc flag value down to the Worker code. If this is just for this, you can simply add your echo command all the time whatever the maxrc value is.
The problem is I don't understand why maxrc is not working for Rsh. It looks to me rsh is a simple command working like any command runnable with an ExecWorker or like a SshWorker.
Could you explain to me (and maybe in the commit message) what was the problem you are solving with this code?

Checking the code, that would means ExecClient._close() is not working as expected for rsh. Could you clarify what you are seeing?

Thanks @degremont. Correct I added the maxrc option so that I could propagate it nicely down to the worker code a little more cleanly. I wanted the code to behave more like pdsh in this respect and not add the magic unless it was necessary. It is possible to grab maxrc in the Worker by looking at self.worker.eh._display.maxrc but it felt that would be violating your intent. If you prefer, I can package this as two commits (I did it together for the exact reason you surmised).

Understood for the maxrc config change. But what about the original problem?
Could you explain in details what is not working with RshWorker which is working with ssh or exec ?

Understood for the maxrc config change. But what about the original problem?
Could you explain in details what is not working with RshWorker which is working with ssh or exec ?

Sure. rsh does not correctly set return codes. As far as I know, it never has. See example below:

> cat /tmp/foo.sh
#!/bin/bash
exit 1

> ssh localhost /tmp/foo.sh
> echo $?
1

> rsh localhost /tmp/foo.sh
> echo $?
0

This manage shows the reason:
https://docs.oracle.com/cd/E36784_01/html/E36870/rsh-1.html

rsh does not return the exit status code of command.

Oh! I understood now how this trick makes sense.
But, in this case, why not simply apply this trick all the time and just update the local rc for each rsh command, as this is done for ssh or exec?
that way, the -S will be automatically handled.

Oh! I understood now how this trick makes sense.
But, in this case, why not simply apply this trick all the time and just update the local rc for each rsh command, as this is done for ssh or exec?
that way, the -S will be automatically handled.

I patterned it after the way it works in Pdsh. We only do the "magic" if the user requests it. This way we do not do any extra work if it is not necessary. If you would prefer to have it done all the time, that is okay with me, was just trying to avoid doing it unless it was specifically asked for.

I add some coding guidelines for this feature. I'm inline with @thiell and limit this PR to the rc feature. Don't include maxrc change here.