Re: anita, qemu, xen, building qemu0

[Andreas Gustafsson <gson%gson.org@localhost>]

Greg Troxel wrote:
> I've used anita a bit, and have a few related questions.
> 
> 1) It depends on qemu, and defaults to 0.  Is qemu1 generally broken,
> even on netbsd-6?

Sadly, most 1.x releases of qemu have been broken even on Linux hosts.
I have run into and reported a total of five regressions in 1.x that
have impacted running NetBSD/i386 as a guest:

   https://bugs.launchpad.net/qemu/+bug/897771
   https://bugs.launchpad.net/qemu/+bug/1089996
   https://bugs.launchpad.net/qemu/+bug/1091241
   https://bugs.launchpad.net/qemu/+bug/1127369
   https://bugs.launchpad.net/qemu/+bug/1154328

The last couple of these were fixed only recently, and I believe they
are now all fixed in the 1.5.0 release (as tested on a Linux host).
Now all we need is for someone to import 1.5.0 to pkgsrc...

I am also successfully using qemu 1.2.0 from pkgsrc under NetBSD 6.x
for emulating i386 and amd64, but there seem to be some issues with
emulating sparc (these may or may not also affect 1.5.0 - I have not
tested that).

> If so, perhaps DESCR should say that.  Or a comment
> in the py-anita Makefile?  Or perhaps there's a better place.

The anita(1) manpage already discusses this.  Hopefully we can get the
problem resolved by importing qemu 1.5.0 to pkgsrc.

> 2) Building py-anita brings in truly vast amounts of packages.   I
> wonder if it shouldn't depend on qemu, since it can be used with xen.

It needs to depend on qemu so that the default settings work out of
the box.  I'd rather see some of qemu's default dependencies dropped
(e.g., pulseaudio).

> 3) Building qemu0 on netbsd-6 i386 failed, filling up /var.

I don't know about this issue; I have never run into it myself.

Regards,
-- 
Andreas Gustafsson, gson%gson.org@localhost