Subject: Re: Use of sun_len in AF_UNIX socket addresses
To: SODA Noriyuki <soda@sra.co.jp>
From: Martin Husemann <martin@duskware.de>
List: tech-net
Date: 10/11/2006 09:35:23
On Wed, Oct 11, 2006 at 10:01:12AM +0900, SODA Noriyuki wrote:
> That's practically impossible, at least for third party packages,
> because only 4.4BSD derived OSes have the sun_len member, and Solaris,
> HP-UX and Linux don't have the sun_len member.
> (Even 4.3BSD-tahoe didn't have it, either.)

Should we consider removing it then?

Otherwise, could we rely on it being either 0 or correctly set? We could
change the kernel then to consistently use a new version of SUN_LEN() that
checks for this and does the right thing.

Martin