fix(ci): required ubuntu 24 runner forward ports as 24 is now latest

[mikehardy]

• with much less default software in the runner image, don't manually liberate space
• lzma no longer included, install a package with stable name that will always depend on it (xz-utils) the liblzma-dev package which should provide it

The need for this was noticed when a simple dependabot PR failed:

= note: /usr/bin/ld: cannot find -llzma: No such file or directory

https://github.com/ankidroid/Anki-Android-Backend/actions/runs/11278308705/job/31366466007?pr=421#step:13:40

...which led to a quick look at the runner image in use, noticed it was ubuntu-24 now and saw it just rolled out https://github.com/actions/runner-images/blob/ubuntu24/20241006.1/images/ubuntu/Ubuntu2404-Readme.md

Success criteria for merge is that the Ubuntu runner goes green in CI with runner image 24

[mikehardy]

Confirmed expected ubuntu-24 runner image:

Runner Image
  Image: ubuntu-24.04
  Version: 20241006.1.0

https://github.com/ankidroid/Anki-Android-Backend/actions/runs/11279185141/job/31369310967?pr=422#step:1:8

But xz-utils was apparently already installed, at the correct version 🤔

https://github.com/ankidroid/Anki-Android-Backend/actions/runs/11279185141/job/31369310967?pr=422#step:3:48

xz-utils is already the newest version (5.6.1+really5.4.5-1build0.1).
0 upgraded, 0 newly installed, 0 to remove and 24 not upgraded.

So I expect this to fail again with -llzma: No such file or directory and I'll have to do something different

[mikehardy]

Definitely installed something new at least, when using liblzma-dev as package name to install, and that did the trick, we got past that point in the build

Reading package lists...
Building dependency tree...
Reading state information...
Suggested packages:
  liblzma-doc
The following NEW packages will be installed:
  liblzma-dev
0 upgraded, 1 newly installed, 0 to remove and 24 not upgraded.

[david-allison]

We've possibly hardcoded a path?

/usr/bin/sh -c \printf 'hw.cpu.ncore=2
' >> /home/runner/.android/avd/test.avd/config.ini
/usr/bin/sh: 1: cannot create /home/runner/.android/avd/test.avd/config.ini: Directory nonexistent

[mikehardy]

I'm still working through this. The emulator runner isn't working with ubuntu-latest == ubuntu24 yet, there is an issue upstream

There was also this nugget in the log which I'm working through:

Warning: This version only understands SDK XML versions up to 3 but an SDK XML file of version 4 was encountered. This can happen if you use versions of Android Studio and the command-line tools that were released at different times.

Got a few things local but I'm working on this in parallel with a few other things for other repos so it'll be stop starty

but I'm on it

[mikehardy]

I may have a local fix for the upstream issue but I need to close/reopen so that my fork of the action is on the allowed list

[mikehardy]

upstream issue on emulator runner will likely resolve soon, but now I've got rsdroid.so not found when running the emulator test

@david-allison does this look familiar to you at all?

net.ankiweb.rsdroid.BackendIntegrationTests > schedTimingTodayCall[emulator-5554 - 6.0] FAILED 
	java.lang.UnsatisfiedLinkError: dalvik.system.PathClassLoader[DexPathList[[zip file "/system/framework/android.test.runner.jar", zip file "/data/app/net.ankiweb.rsdroid.instrumented.test-1/base.apk", zip file "/data/app/net.ankiweb.rsdroid.instrumented-1/base.apk"],nativeLibraryDirectories=[/data/app/net.ankiweb.rsdroid.instrumented.test-1/lib/x86_64, /data/app/net.ankiweb.rsdroid.instrumented-1/lib/x86_64, /vendor/lib64, /system/lib64]]] couldn't find "librsdroid.so"
	at java.lang.Runtime.loadLibrary(Runtime.java:367)

I don't have an Ubuntu 24 env laying around at the moment to check myself, but I worry that similar to liblzma not being on the system, this could be a subtle linkage issue.

Or it could just be some obvious failure that the build system silently glosses over - there doesn't seem to be a lot of "did the expected products from this stage actually get produced correctly?" type verification as the build works

[david-allison]

Do we have the .so so we can test it manually?

Not enough information from that error to know where the issue came from

[mikehardy]

No, the workflow would need some hackery to post it as an artifact or similar, which would require finding out exactly where it is, at which point it's probably faster to spin up an ubuntu 24 vm locally but I won't have time for that for a while unfortunately

[mikehardy]

Oh! I see that you actually do upload those artifacts, you must have set that up long ago, I see it on another run https://github.com/ankidroid/Anki-Android-Backend/actions/runs/11297029758?pr=423 so perhaps we have the .so

This isn't the highest priority (I pinned the runner to ubuntu-22 today on main) so will catch this as I can, or feel free anyone else reading this...