Removing a non-empty directory results in EIO about fscq HOT 3 CLOSED

Thanks; you are absolutely correct that this is a bug and it would be good to fix. At the moment, neither fscq's implementation nor theorems attempt to return the right errno values (e.g., see our SOSP paper's eval section); almost every system call likely returns the wrong errnos.

from fscq.

Most of the errnos look correct/reasonable, actually. The only ones that we picked up are related to directory removal and renaming. Would you like me to file the other errno issue or leave it here?

I also have an nlink issue and a number of known unsupported feature issues. Would you like them filed or sent somewhere else or dropped?

from fscq.

This issue turns out to be fixed in FSCQ, which now propagates error codes at least in some cases.

from fscq.