On Tue, Oct 01, 2002, Robert Collins wrote:
> > Now I'm trying a possible implementation of FD_CLOSE_METHOD on nt branch,
> > but at this time I'm still evaluating the benefits, if any.
>
> I think what Adrian is saying is that rather than use FD_CLOSE_METHOD,
> all close() calls should be removed and replaced with
> comm_close()/file_close(). Then there is only one close() call that
> needs replacing for the NT port.
There's two views:
* extra verification is good - you should know what "type" the FD is
beforehand
* why should I care if my helper is talking over tcp/ip or a local pipe?
close() should close()!
Its annoying, but its one of the things I'm planning on tidying up with
the comm code - nothing outside of comm*/disk* should _ever_ call
a direct fd call (read, write, recvfrom, sendto, pipe..)
Adrian
Received on Thu Oct 03 2002 - 19:04:13 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:16:52 MST