This is the mail archive of the
mailing list for the glibc project.
Re: 2.0.110: [/tmp/glibc/usr/share/zoneinfo/Australia/Darwin] Illegal instruction
- To: John Kennedy <firstname.lastname@example.org>
- Subject: Re: 2.0.110: [/tmp/glibc/usr/share/zoneinfo/Australia/Darwin] Illegal instruction
- From: Philip Blundell <email@example.com>
- Date: Tue, 19 Jan 1999 10:01:11 +0100
- cc: firstname.lastname@example.org, glibc <email@example.com>
> I really miss the sockaddr_union in <bits/sockunion.h>. I don't follow
>such things closely, but since it was commented as "somewhat evil" I'm
>wondering if something has replaced it or if it was just purged to keep
>the rest of glibc cleaner.
It was only added in the first place because the IPv6 API drafts required it.
The newer versions of these specifications have dropped the requirement so I
pulled it from libc again. It was never a good idea to have <sys/socket.h>
pull in all the address-family-specific code, and there were a few corner
cases where you could trick it into multiply defining things or not defining
them at all by including things in the right order. To be honest I didn't
expect anybody to be using sockaddr_union; sorry.
I think the later API specifications do include something with similar
intent (a block of memory large enough to store any sockaddr structure) but
this isn't implemented yet. If you'd like to have a crack feel free to send
me a patch.