#2 bind-listen-accept-loop needs handler

[Brian D. Carlstrom]

From: jeer.1997Jun11@mockery.yrek.com (Quowong P Liu)
To: scsh-news@martigny.ai.mit.edu
Subject: bind-listen-accept-loop
Date: 11 Jun 1997 15:52:20 GMT

I'm using a copy of bind-listen-accept-loop with an
error handler around the call to accept-connection.

Here's why. I've got a little SMTP server to test out
ideas for detecting junk mail. The problem is, every
few hours, I get these connections that causes the
bind-listen-accept-loop to throw an error, which would
bring down the server. I suspect it's due to
connection
attempts from sites in netblocks that my ISP has
rerouted
to oblivion due their high spew volume.

Putting an error handler outside bind-listen-accept-
loop
isn't too useful for recovering from these kinds of
errors, so I think bind-listen-accept-loop should have
some mechanism for dealing with them. I'm using scsh
0.5,
in case it's since been changed.

With the error caught and logged, this is what it looks
like:

Wed, 11 Jun 1997 06:02:06 UCT [237]: listen-accept
error: (syscall-error 113 No\ route to host #{Procedure 9599 %accept} 4 2 ????????)
Wed, 11 Jun 1997 06:02:30 UCT [237]: listen-accept
error: (syscall-error 113 No\ route to host #{Procedure 9599 %accept} 4 2 ????????)
Wed, 11 Jun 1997 07:57:37 UCT [237]: listen-accept
error: (syscall-error 113 No\ route to host #{Procedure 9599 %accept} 4 2 ????????)
Wed, 11 Jun 1997 07:58:01 UCT [237]: listen-accept
error: (syscall-error 113 No\ route to host #{Procedure 9599 %accept} 4 2 ????????)
Wed, 11 Jun 1997 11:06:21 UCT [237]: listen-accept
error: (syscall-error 113 No\ route to host #{Procedure 9599 %accept} 4 2 ????????)
Wed, 11 Jun 1997 11:19:22 UCT [237]: listen-accept
error: (syscall-error 113 No\ route to host #{Procedure 9599 %accept} 4 2 ????????)
Wed, 11 Jun 1997 11:19:35 UCT [237]: listen-accept
error: (syscall-error 113 No\ route to host #{Procedure 9599 %accept} 4 2 ????????)

--
qpliu@mockery.yrek.com
550 No junkmail

[Martin Gasbichler]

Logged In: YES
user_id=17553

This is not a bug but a feature request. One that we should
consider I think.