On Mon, 27 Jan 2020 at 15:07, Alex Bennée <alex.ben...@linaro.org> wrote: > Robert Foley <robert.fo...@linaro.org> writes: > > > Hi Drew, > > > > On Mon, 27 Jan 2020 at 12:27, Andrew Jones <drjo...@redhat.com> wrote: > > > >> > > >> > I suppose we could check the version of QEMU and use the above > >> > defaults only for earlier versions of QEMU. > >> > This is something we will probably move to aarch64vm.py since it is > >> > common. > >> > >> What versions of QEMU do these tests *have* to support? Because we could > >> just skip the tests for QEMU that doesn't support cpu=max,gic-version=max. > >> 'max' is indeed the nicest selection for using the same command line on > >> KVM (gicv2 and gicv3 hosts) and TCG. > > > > I believe these test scripts which build/launch the VM have to support > > the older version of QEMU since > > this is the version of QEMU currently used when these VMs are > > launched. I don't know the history on > > this, but it seems intentional that we use one older/different version > > of QEMU to launch the VM, > > Well we defer to the system QEMU as it should be stable. It can be > overridden with the QEMU environment variable which worked well enough > when we only had VMs of one architecture. Perhaps we needs a new way to > say "use the appropriate QEMU from this build"?
Good idea. This is a pretty common use case and it makes sense to make it easy to use. I will add some support for this in my patch series. Thanks & Regards, -Rob > > > while we test the 'current' build of QEMU inside the VM. > > It also seems like a 'nice to have' to automatically support the > > latest version where we could > > use max as you pointed out. > > > > Thanks & Regards, > > -Rob > > > -- > Alex Bennée