Disable macOS M1 testing

[gsmet]

It has been completely unstable for months now.

@holly-cummins sorry it has to come to that but it's not sustainable to have this thing failing 95% of the time (if not more).

[holly-cummins]

I've noticed that when we have test failures, we tend to disable the test (ie, we accept the break). To me, this also reduces the value of this infra for catching M1 problems.

So we should probably have a conversation about what level of M1 support we want to aim for.

Before we disable the platform entirely, should we consider paying for a Docker license? The stability issues we see are almost entirely related to Podman, so installing Docker on the system would allow us to keep the M1 coverage (although we'd be losing Podman coverage, of course).

[holly-cummins]

Just to expand on my comment about the issues being caused by podman, we see three classes of problem:

• Podman machine fails to start, killing the build in in its first 20s or so - Podman machine start fails with 'no such file or directory', reboot resolves issue on Mac M1 containers/podman#16163 covers some of these cases, and we sometimes see other error messages.
• Disk space leaks. I don't know the root cause of these. Something is using much too much purgeable space, and MacOS is not properly clearing the purgeable space. It's clearly a MacOS bug, but I don't know what's exposing it. It may or may not be podman.
• A particular service fails to work with podman. In most cases these are genuine product issues, but not necessarily ones we can fix, since they involve third parties.

[gsmet]

Well, we can start discussions on how to fix it in the future but it has been months now that this thing is not working correctly and that it fails most of the time.
And from what I can see, the issues don't look trivial or you would have already fixed them, given the time you spent on this thing.
So let's disable it for now and enable it again if we can get it working more reliably (for instance by testing in a PR several times).

I know @geoand had the same opinion because he created a similar PR a while ago.

[geoand]

I know @geoand had the same opinion because he created a similar PR a while ago.

+1