Ways to deal with broken machine types

[Michael S. Tsirkin]

On Tue, Mar 30, 2021 at 12:21:37PM +0100, David Edmondson wrote:
> > Unfortunately I do not think this is practical :(.
> >
> > All examples of breakage I am aware of, we did not
> > realise some part of interface was part of guest ABI
> > and unsafe to change. We simply would not know to write a
> > test for it.
> 
> While agreeing that it would not be possible to cover all aspects of the
> ABI immediately, does that mean that some level of coverage would not be
> useful?

Our testing already warns about ACPI table changes (which is what
happened here). We just verified them manually and thought they are
fine.

-- 
MST