Jonas 'Sortie' Termansen <sortie@maxsi.org> wrote:

> uid_t can be 64-bit per POSIX. That's the case on my Sortix OS. Yeah I'm
> aware that it causes breakage like this every now and then, but those
> cases are easy to fix.

Where does it say this?  If true the number of portability issues this
will create, adjacent to security, is assuredly non-zero.  I don't believe
this to be true nor practical.  (The Austin group is making a shocking
number of unwise decisions recently, fighting reality is making them less
relevant).
 
> mktemp was removed in POSIX 2008. mkstemp is the safer standard
> replacement, although this pattern used here is correct for unix socket
> creation. It is easy to use mkstemp instead though.

No.   mkstemp creates a  file, not a socket.

Your diff simply swaps one race for another race, by using unlink and then
re-creating it.

> caddr_t is not a standard type. It's cleaner to use just char* instead.
> Although in this case msg_control is standardized as void* so no cast is
> needed at all, now that I think about it.

That seems like a holy war with one participant -- you.  That war is to
remove all defacto non-standardized things, right?  Good luck with that!