Garrett Rooney wrote:
The test is trying to send a nonexistant UID/GID (it hardcodes 9999999 as the value, FWIW), that seems like "faulty data" to me. If you aren't allowed to just throw random crap at these functions that's also fine, but given that we actually have a test for that behavior it sure implies that it's allowed. If not, the test should go away and we should probably document that these values are platform specific, and that the only portable way to get one is via an APR function that goes from name to UID or name to GID.
I guess the point/question comes down to this, is 9999999 a valid apr_uid_t / apr_gid_t? If they are then yes, I concur with your patch. But nothing on the win32 side could return such a value. Those types are pointers to data. Now could we have an invalid user/group on win32? Sure, but it wouldn't be a numeric constant on win32 like this example. It would be a pointer to a SID that was unrecognized. Bill
