> I would prefer that unexpected error codes cause glibc to crash. This > immediately alerts kernel and glibc developers of a mismatch in their > expectations before this ever gets out of experimental distributions. That's my inclination as well.