On Tuesday 30 May 2006 23:39, Michael Stefaniuc wrote: > it looks like the function returns on the error path with the lock held. > Is it supposed to do that? If not the attached patch fixes it. No, that's a bug. Thanks. -Hans