Re: [Qemu-devel] [PATCH v3 1/4] serial: reset thri_pending on IER writes with THRI=0

[Paolo Bonzini]

On 15/12/2014 14:37, Dr. David Alan Gilbert wrote:
>> > Right, the question is whether you really want to do this. :)
> Well, it would solve this type of problem less painfully, your migration
> would always succeed and destinations that knew what to do with it would
> always do the right thing.

Yes, that's a determination to make.  I agree that this particular case
is one where losing the subsection is not the end of the world, and
keying off the machine type could be a good call for downstreams.

But that's also because downstreams can also patch older source
versions, while QEMU will not release another 2.1 version.  In fact...

> Of course you could also solve the problem
> if the source knew the version of the destination, but you didn't like
> that idea either.

... if the source could be patched by someone to know the version of the
destination, they could also add the subsection with a "needed" function
that always returns 0.  This way it is never migrated, but it is
interpreted.  You do not need the version of the destination.

Paolo