On Thu, Jan 25, 2024 at 10:46:39AM +0100, Peter Krempa wrote:
On Wed, Jan 24, 2024 at 20:37:26 +0100, Andrea Bolognani wrote:
> We currently have a single test case called "minimal", which
> suffers from two big flaws:
>
> * it's limited to the x86_64/pc machine type;
> * it explicitly enables a number of devices.
>
> Add several test cases, one for each of the architectures and
> machine types that we have good support for.
>
> Unlike the existing one, they're *really* minimal: no devices
> or controllers at all are present in the input XML. So the new
> test cases demonstrate exactly what devices and controller
> libvirt will decide to add automatically.
>
> Note that we use the ABI_UPDATE variant of the test macros
> because, in some cases, the behavior for new guests is not the
> same as that for existing ones due to backward compatibility
> concerns, and we specifically care about the former.
IMO it would make sense to also add the non-ABI update cases.
Agreed.
even if you decide to do the above, you can use
Reviewed-by: Peter Krempa <pkrempa(a)redhat.com>
without reposting.
In order to flush things a bit, my plan would be to push patches
tests: Add usb-controller-automatic-unavailable-q35
tests: Add aarch64-panic-no-model
tests: Add title-and-description
tests: Drop existing <title> and <description> tags
tests: Rename and minimize no-memory
tests: Add minimal cases for many architectures
tests: Drop minimal
tests: Add default-models cases for many architectures
qemu: Fix a few comments
qemu: Default to no USB and no memballoon for new architectures
qemu: Clean up qemuDomainDefaultNetModel()
qemu: Drop qemuDomainFindSCSIControllerModel()
qemu: Drop qemuDomainSetSCSIControllerModel()
qemu: Add missing error handling
qemu: Move qemuDomainGetSCSIControllerModel()
that is, 01-14 and 16, with the following simple changes based on
your suggestions:
* add title-and-description to genericxml2xmltest instead of
qemuxmlconftest;
* have both ABI_UPDATE and regular variants of the minimal
qemuxmlconftest cases, as well as the default-models cases.
Can I go ahead without reposting?
--
Andrea Bolognani / Red Hat / Virtualization