The patch sent lastly proves to be functional. I know it is a bit violent method and a hack, but it works perfectly. We have about every other day SIGKILL in gdm debug log.
I put it here for reference for others who might experience the same problem as _temporary_ solution. The cause (gdm daemon waiting possibly infinitely for its child to exit after multiple SIGTERMs, being blocked for logins for that whole time) remains: sleep(10) in the code. Regards, Vaclav -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]