On Thu, May 08, 2025 at 02:00:07AM -0700, Andrea Bolognani wrote:
On Thu, May 08, 2025 at 07:51:29AM +0200, Peter Krempa wrote: [...] seems to indicate that they have just applied a naive s/borzoi/collie/ without considering the semantics. With this patch applied:
$ grep collie tests/qemucapabilitiesdata/caps_8.2.0_aarch64.xml <machine type='kvm' name='collie' maxCpus='1' defaultCPU='pxa270-c0-arm-cpu' acpi='no'/> <machine type='kvm' name='collie' maxCpus='1' defaultCPU='sa1110-arm-cpu' defaultRAMid='strongarm.sdram' acpi='no'/> <machine type='tcg' name='collie' maxCpus='1' defaultCPU='pxa270-c0-arm-cpu' acpi='no'/> <machine type='tcg' name='collie' maxCpus='1' defaultCPU='sa1110-arm-cpu' defaultRAMid='strongarm.sdram' acpi='no'/>
i.e. the collie machine shows up twice, with completely different attributes associated with it each time, in the resulting capabilities data. This obviously will not make any sense to a user or application parsing the information.
Disregard this part. I was looking at an earlier version of the patch set which displayed this behavior, but upon further inspection I see that's not present in the version that was posted. The NACK stands for the other reasons. -- Andrea Bolognani / Red Hat / Virtualization