On Tue, 2018-08-07 at 13:19 +0100, Daniel P. Berrangé wrote:
On Mon, Aug 06, 2018 at 11:38:06AM +0200, Andrea Bolognani wrote:
I wonder how to handle architectures where QEMU never declared a default machine type, such as aarch64 and riscv64, though: I think it would make sense to prefer the virt machine type there, but I'm not entirely sure that wouldn't cause any breakages either...
Existing libvirt behaviour is that we'll pick the first reported machine type, so we have to preserve that.
Right, makes sense. I guess with aarch64/virt first and x86_64/q35 now it won't take too long for applications to figure out they should specify the machine type explicitly rather than leaving that up to libvirt :) -- Andrea Bolognani / Red Hat / Virtualization