On Tue, 31 Dec 1996, Piete Brooks wrote:
> Hmm -- as I understood it, PH10 ripped the code off from another system.
I copied the code from Pine.
> I assume that at least some other system has the same problem,
Pine would have the same problem.
> and I kind of assumed that return codes were checked -- I take it they are not
The same code is found in procmail, which originally used exactly the
same algorithm. You don't check the return code of the link(), you do a
stat and look for 2 links afterwards. This covers you in the case when
the link is made, but the NFS server crashes and reboots before it can
send a reply to the link(), which then fails (times out). However,
procmail's author has subsequently updated the algorithm to check the
return from link() and only if it fails do you go on to do the stat(), in
case there has been a crash/reboot of the server.
> Well, we know that Linux is being "fixed", and as I understand it PH10 is
> re-doing exim as well, so whhich is "in the wrong" isn't important.
The next release of Exim, which I am currently working on, has been
updated to use the new algorithm.
--
Philip Hazel University Computing Service,
ph10@??? New Museums Site, Cambridge CB2 3QG,
P.Hazel@??? England. Phone: +44 1223 334714
