From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <aae95359a010fad875ecc996ec656bb5@plan9.bell-labs.com>
To: 9fans@cse.psu.edu
From: Sape Mullender <sape@plan9.bell-labs.com>
In-Reply-To: <3afc1f4defb71eb39b20882e5243fb22@vitanuova.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Subject: [9fans] (Re:) Exception handling
Date: Mon, 12 May 2003 13:30:24 -0400
Topicbox-Message-UUID: a9a8d2ca-eacb-11e9-9e20-41e7f4b1d025

I have `ported' (it's so trivial, it doesn't deserve being calleda port) the kernel's
exception mechanism (error, waserror, nexterror, poperror)
to a largish application we're building for a high-speed wireless switch.
I did this after we already had a lot of conventional error checking in the code, e.g.,
	snprint(buf, sizeof(buf), "add %s 255.255.255.255 %s %d proxy",
		con->locaddr, con->remaddr, 2048);
	if(fprint(cfd, "%s", buf) < 0){
		close(cfd);
		return "can't set  addresses";
	}
	if(fprint(cfd, "iprouting 1") < 0){
		close(cfd);
		return "can't set  iprouting";
	}
(This routine used a nn-nil return as an error indication).
The caller of this routine would have a similar test:
	if (error = ipsetup(con)){
		cleanup();
		...
Converting the whole app from this error checking style to using waserror, etc.
took some time, but in the end it did reduce code size and it made the code more
legible.  It also made it harder to ignore unhandled errors deep in several
layers of subroutines.