On Mon, 5 Jun 2023 at 11:51, Paolo Bonzini <pbonz...@redhat.com> wrote: > > On 6/5/23 11:46, Peter Maydell wrote: > >> To avoid this issue, tests/requirements.txt should use a ">=" constraint > >> and the version of Avocado should be limited to what distros provide > >> in the system packages. Only Fedora has Avocado, and more specifically > >> version 92.0. For now, this series reverts to the older requirement > >> (version >=88.1) while leaving further version bumps to future changes. > > If the new Avocado version is broken, don't we also need a < constraint > > so we don't get it by mistake ? > > I expected those to be bugs that get fixed in 102 or 101.1, so not a > reason to impose a strict constraint. But you're right, the version > that would be installed from PyPI is the latest; I didn't notice because > I do have avocado installed outside pyvenv/. > > Is the logging issue limited to the one fixed by > https://www.mail-archive.com/qemu-devel@nongnu.org/msg962758.html? Or > is there something more?
I don't know, as I haven't tested with that patch. > > In particular, for a local build tree that currently has 101 installed, > > if the tree is updated to include these two patches together, will that > > correctly downgrade it to 88.1? > > No, it won't. What you can do is "pyvenv/bin/pip uninstall > avocado-framework". I think if we're going to revert back from Avocado 101 we should actively do that, not merely allow older versions. -- PMM
