Re: [xen-unstable test] 27016: regressions - FAIL

[Anthony PERARD <anthony.perard@citrix.com>]

On Tue, Jun 10, 2014 at 11:55:09AM +0100, Ian Jackson wrote:
> Ian Campbell writes ("Re: [Xen-devel] [xen-unstable test] 27016: regressions - FAIL"):
> > On Tue, 2014-06-10 at 10:02 +0100, Jan Beulich wrote:
> > > From the looks of it this failure is persistent (seen in all three flights
> > > over the weekend), pointing at qemu and USB. Not sure who to
> > > ask to look into this, not the least because none of the rather
> > > many pending changes appear to be an immediate candidate.
> > > Perhaps we ought to wait for the bisector to get going on it...
> > 
> > The bisector is close to fingering the big upstream qemu merge. Stefano
> > and Anthony are investigating and Ian J was looking into perhaps doing
> > an adhoc bisect of the upstream branch, but last I heard picking a
> > baseline was looking problematic.
> 
> Our options at this stage are:
> 
>  * Continue to try to debug it, blocking other trees in the meantime.
> 
>  * Force push xen-unstable (and other affected trees of which there
>    will probably be quite a few).  That will turn this failure from a
>    blocking regression into an allowable failure.  This will allow
>    other work to continue but may take the pressure off trying to fix
>    it.
> 
>  * Revert the merge in qemu-upstream-unstable.git.  This will return
>    us to qemu 1.6.0.
> 
> 11:37 <anthonyper> Diziet, I did not find any obvious bug with qemu
>                    2.0. All I found is the guest that take about 1 or
>                    2 min to respond again
> 
> Stefano says that this is a decision for the release manager, and I'm
> inclined to concur.
> 
> (If a force push is called for I will do it.  I can also help with
> reverting the merge if that's what's decided.)

The first qemu upstream commit to fail the test:
  6eefccc0bb9c34051b1e21880fc3a1c1c8686edd
  roms: update vgabios binaries
  This also switches from lgplvgabios to seavgabios.

A more recent version of the seavgabios binary fix the issue with
freebsd, so I asked for a backport in an other email.


But there is still a minor issue related to qemu 2.0, it concernes VNC.
The screen does not properly get updated (when moving a window under
Windows for example).

-- 
Anthony PERARD