A while ago I built Xsun for Debian sparc. This is the Xsun collection of
packages currently on Anders' ftp site (for exact distribution info, please
read the mailing list archives - it's in there). There are two major bugs:
those Xsun packages do not work (at all) with TCX framebuffers (like the
framebuffer on my SS4, for instance) and they do not handle gzipped fonts.
This means that you need to make an X font server, as described in several
previous postings, to make everything work with the existing font packages.
This much is widely known.
What is less widely known is that I have been working on a fix for both
of these problems. I _think_ I have a workable patch. Unfortunately, I
can't seem to compile anything (including xlib) that uses sys/socket.h, so I
can't test it. This is why I haven't been as forthcoming with fresh, working
X packages as I'd like; I just haven't had the time to track this down.
In libc6-dev, version 2.0.93-980414-1 (from a mirror of Incoming),
<sys/socket.h> includes <bits/sockunion.h>. sockunion.h defines this massive
union of all possible sockaddr structs for some reason (used for IPv6?).
This union includes sockaddr structs for rose (<netrose/rose.h>, which in turn
includes <linux/rose.h>, only found in the unstable kernels) and for
AppleTalk (normally found in the libatalk-dev package, <netatalk/at.h>).
Getting rose.h is easy, just download the unstable kernels. Getting netatalk
is a bit trickier, since netatalk won't compile (it needs a working
<sys/socket.h> to build).
I finally have time to work around this (possibly by downgrading
libc6-dev) and I hope to have something compiling soon. But is this really
the way libc6-dev should be? Should it use headers in development kernels
and other packages like this? And, if so, could someone please upload a
working version of libatalk and libatalk-dev to master?
--
Mike "No .sig for you" Shuey
--
To UNSUBSCRIBE, email to debian-sparc-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org