On Tue, 2019-10-22 at 14:32 +0200, Boris Fiuczynski wrote:
On 10/22/19 1:39 PM, Andrea Bolognani wrote:
> On Mon, 2019-10-21 at 10:55 +0200, Shalini Chellathurai Saroja wrote:
> > .../domaincapsschemadata/qemu_4.1.0.s390x.xml | 221 +
> > tests/domaincapstest.c | 4 +
> > .../caps_4.1.0.s390x.replies | 22171 ++++++++++++++++
> > .../qemucapabilitiesdata/caps_4.1.0.s390x.xml | 3272 +++
> > 4 files changed, 25668 insertions(+)
>
> Can you please do 4.2.0 instead? Jirka is working on a patch series
> that will use a feature introduced in the upcoming QEMU release, and
> it would be useful to test it on s390x along with all the other
> architectures.
>
> Note that it's fine to just use the latest commit on master for now
> and only later, once QEMU 4.2.0 is our, post a much smaller patch
> that contains just the differences.
>
We can do the (pre) 4.2.0 on s390x in addition to the 4.1.0 version. In
that way when backporting a matching qemu version is available.
I'm not against having 4.1.0 replies per se, but I just don't see
much value in adding them at this point.
As mentioned in a previous message, we generally only add replies
when doing so improves test coverage: in the case of QEMU 4.1.0, it
looks like they were originally introduced to test CPU feature
translation, which AFAICT is x86-only; accordingly, we only have
x86_64 replies for that QEMU release.
I can see the value in having replies for 4.2.0 on s390x; replies
for 4.1.0 look to me like they would only add a few hundred KiBs to
the repository and release archives for no real benefit.
--
Andrea Bolognani / Red Hat / Virtualization