Doesn't the above just "work" assuming that the system in question is
using FCNTL_SERIALIZED_ACCEPT? At last look, there were a few systems that
didn't bother (or need to) use that... does this lock them out (no
pun intended)?
Well, at least one platform *did* have problems with that version of the
patch (see subsequent email) --- FWIW, FCNTL_SERIALIZED_ACCEPT is only
defined on those systems which need some kind of mutual exclusion around
accept() --- most (including A/UX, as of last I'd heard) appear not to;
see conf.h for the ugly details.
(SunOS is one of the systems which doesn't need locking, but it's what
I've got, so I forced the issue with an explicit -DFCNTL_... on the command
line to gcc to test the code out).
rst