> > If you mean checking the return from g_path_get_dirname(), I realized
> after sending my message that we don't check the return from that
> anywhere else in the code, and that it can't fail - it will always
> return something non-null (if nothing else it will return "."), so the
> proper change there was just to remove the check. And in the callers of
> the new function, as long as virPCIDeviceIsOnRootBus() is logging an
> error in all cases when it returns failure, we don't need to (and
> shouldn't) log anything in the caller (since it's already been logged).
> Is there something else I've missed?
I think that makes sense. As you said, the real critical path is trying to
resolve the symlink to an actual device and that seems covered.
