[Nut-upsuser] Loosing comms with UPS

[Arjen de Korte]

Alain Williams wrote:

> I ran strace on the upsd process and what I get is below.

Usually, running strace is only used as a last resort, if adding the
-DDDDD option doesn't provide sufficient output. Since this is a
communications problem between the server and the client, we probably
need them to run both with the -DDDDD option in order to find out which
one is the culprit.

> It appears that upsd gets a connection (from somewhere), reads a GET VAR,
> writes a VAR reply, read EOF - and so generate the 'read failure'.

What is might be missing from the server, is that we still FD_SET the
client file descriptor, even if we already know it is no longer
connected. This is probably not what we want. See the trunk (> r1027).

Best regards, Arjen