On Fri, 2 May 1997, Nathan J Kurz wrote:
> I don't use identd, so I don't want to patch it, but it should be
> simple to substitute write_with_timeout() and read_with_timeout() for
> the write() and read() calls. Then you could just strip out all the
> alarm handling and not muck around with resetting it.
You also need a connect_with_timeout, which requires alarm(). In fact
the most common place for this specific to timeout is during connect()
I would bet, since that's where it'll get stuck if there's a firewall
just dropping the packets.
> Then again maybe keeping the alarms for 1.2 would be good. Although
> there were a lot of things being changed in Dean's fix, so maybe this
> would be cleaner.
There wasn't much being changed actually, all I did was record the current
"global" timeout length and restore it at the end of the rfc1413 module.
Not perfect... but...
Dean
|