[OpenAFS] OpenAFS on OpenBSD

[Benjamin Kaduk]

On Sun, Aug 27, 2017 at 04:09:22PM -0400, Adam Jensen wrote:
> On Sun, 27 Aug 2017 14:59:43 -0400
> Adam Jensen <hanzer@riseup.net> wrote:
> 
> > What would be involved in building the most recent OpenAFS release (1.6.21) on the latest release of OpenBSD (6.1, x86-64)? The README file suggests this platform might have been abandoned long ago.
> > 
> > sysname
> > -------
> >       i386_obsd31, i386_obsd32, i386_obsd33, i386_obsd34, i386_obsd35,
> >          i386_obsd36, i386_obsd37, i386_obsd38, i386_obsd39, i386_obsd40,
> >          i386_obsd41
> > -------
> 
> I neglected to mention that I am interested in a complete OpenAFS solution on OpenBSD - client + server + tools. That platform is such a hassle to deal with, a half-measure probably isn't worth the effort, IMO.

It's worth mentioning, yes.  The server is a pretty portable POSIX application
and should build without much trouble, but the client (kernel module) is
where the difficulty is likely to lie.  The last commit in the tree working
on OpenBSD client support is e1d0342326d11a14e1fb0075fb62cc6be9389b97, from
2014, which added support for OpenBSD 5.4, which is quite a few releases
behind.  So, someone would need to look at the VFS- and VM-layer changes
in OpenBSD between 5.4 and 6.1, and make the necessary (conditional!)
adjustments in the OpenAFS source tree to match up.

In FreeBSD, there are usually not too many such changes in a given point
release (though major releases can have large changes), and usually the
points of breakage become visible at compile time (but not always).  There's
also enough runtime assertions in debug kernels that finding the other places
that need changing is also not too hard.  Unfortunately, I don't follow
OpenBSD closely enough to know whether the situation is similar there.

-Ben