On Wed, Nov 17, 2021 at 12:36:26 +0100, Boris Fiuczynski wrote: > On 11/17/21 12:12 PM, Daniel Henrique Barboza wrote: > > > > > > On 11/17/21 07:51, Peter Krempa wrote: > > > On Tue, Nov 16, 2021 at 21:11:50 -0300, Daniel Henrique Barboza wrote: [...] > > > This patch itself is acceptable as is, but doesn't really fix the real > > > issue. If you don't plan on addressing the root cause, please: > > > > > > I'll wait to see if that fixes the problem Boris was experiencing with > > s390x. Meanwhile I'll dig further to see if I can detect the root cause > > of the problem (e.g. can we default to an Intel CPU instead of AMD). > > Switching the default from AMD to Intel would also not address the root > cause. In addition the capability test data which is the base for the _LATEST caps are always based on a real qemu run which is unfortunately CPU specific. One potential fix would be to run it in a nested VM with an emulated CPU but that would defeat the purpose in certain parts. The X86 caps are generated mainly by me and I have an AMD box so for now there's no intention to change that in any way.
