In D26376#586472, @kib wrote:

I do not object against returning error if subcommand is not specified, or if no path is given, but why ?

The usage message implies that at least one path must be given. I think it is also more conventional to return an error instead of silently doing nothing, and it helps catch bugs (e.g., someone writes "posixshmcontrol create $foo" in a shell script but forgets to assign to foo). Indeed, I noticed because in some manual testing I made a similar mistake.