mutter arm64 tests passed in PS4.5 but not PS6

[Brian Murray]

I was looking at mutter tests in Jammy (because the failing test is blocking an SRU) and noticed that a couple of failures were in bos01 or bos02 and the passing one was in bos03. I did not check all the log files so it might have also failed in bos03 but I think this warrants some further investigation. Let's check and see if there is a difference the availability of gpus.

From the failing log file:

3079s Bail out! libmutter-FATAL-WARNING: Failed to open gpu '/dev/dri/card0': Failed to open device '/dev/dri/renderD128': Permission denied

Here are the lists of tests which I was looking at:

https://autopkgtest.ubuntu.com/packages/m/mutter/jammy/arm64

If we do find out that there is a difference in device availability we should check for other packages which might have failed tests.

[Brian Murray (brian-murray)]

We should also manually run the test in bos01 / bos02 (if possible) and provide the results to bug 2059847 so the SRU team can proceed with releasing mutter.

[Brian Murray (brian-murray)]

I had a look at regressions and didn't find any other test failing to open the GPU but we should look at all recent failures i.e. even ones that don't appear in update-excuses.

 $ for series in bionic focal jammy mantic noble oracular; do ./retry-autopkgtest-regressions --series $series --log-regex "Failed to open gpu" | grep arm64; done
https://autopkgtest.ubuntu.com/request.cgi?release=jammy&arch=arm64&package=mutter&trigger=mutter%2F42.9-0ubuntu8&trigger=xorg-server%2F2%3A21.1.4-2ubuntu1.7~22.04.11

[Brian Murray (brian-murray)]

I've set this to Medium until we determine whether or not there really is a difference between PS4.5 and PS6. If we do discover there is a difference than we should adjust increase the importance of the bug.

[Daniel van Vugt (vanvugt)]

I don't think this bug is needed. It looks like mutter 42 (jammy) is the only release that still has the problem, so I will patch it (bug 2068119).

[Brian Murray (brian-murray)]

I'm setting this to Invalid given Daniel's comments and the need for mutter's test to better deal with hardware availability.