The spec for those get*_r calls says it fills in with NULL on errors or not found (which is a non-error return value), so it is not necessary to check the return value unless you want to distinguish "some error reading the passwd info" from "lookup found nothing". But here perhaps we do, to return different error codes to the caller for those two cases.
I haven't really considered your changes deeply beyond that. addauth with numeric IDs that are not in passwd/group/shadow should still work for root. Does it? When it fails for a nonroot user it should diagnose a more useful error than EINVAL, at least EPERM. In fact, really addauth should just handle numeric IDs and try it, and when the auth server refuses for nonroot the EPERM return from auth_makeauth should propagate all the way back. (For example, imagine an auth server decides to implement a policy where a UID has control over a private range of other UIDs. The client side should not be constraining the auth server's process of deciding who can do what.) Thanks, Roland _______________________________________________ Bug-hurd mailing list [EMAIL PROTECTED] http://lists.gnu.org/mailman/listinfo/bug-hurd
