Test hello-world example with Android Emulator on CI

[Gordon-F]

No description provided.

[dvc94ch]

Neat!

[MarijnS95]

Haven't reviewed in detail, but looks cool and simple enough!

[MarijnS95]

Awesome, the test works now! We should probably come up with a more sophisticated test-case at some point, but validating app startup and "hello world" being printed is an important first milestone. Thanks!

[MarijnS95]

@Gordon-F Would you be able to help out? For quite some time now the CI seems to be unresponsive around adb install, timing out the job 3hrs after:

https://github.com/rust-windowing/android-ndk-rs/actions/runs/1452352305

[MarijnS95]

I vaguely remember writing down some more info in #169, too.

[Gordon-F]

@MarijnS95 It looks like an infrastructure problem on the GitHub runner side.

Is macos-latest MacOS BigSur powered runners? Maybe we should switch to macos-15 and run only on Catalina?

[MarijnS95]

@Gordon-F I bumped it from macos-10.15 to macos-latest in b665be0, and vaguely remember it started happening exactly since that patch. However:

https://docs.github.com/en/actions/learn-github-actions/workflow-syntax-for-github-actions#github-hosted-runners

The macos-latest label currently uses the macOS 10.15 runner image.

So it should not have made a difference?

[Gordon-F]

Also, the emulator starts 3 times for 3 different tasks, if I understood everything correctly. Is it necessary? It seems to me that it would be more optimal to start it only 1 time and run all tests on it.

[MarijnS95]

The broken run also shows that it was on 10.15:

Current runner version: '2.284.0'
Operating System
  Mac OS X
  10.15.7
  19H1419

Also, the emulator starts 3 times for 3 different tasks, if I understood everything correctly. Is it necessary? It seems to me that it would be more optimal to start it only 1 time and run all tests on it.

But we'd like to have a clean emulator for every artifact that we need to test. Could deinstall the application but that seems far from ideal. However, perhaps starting 3 of them instead of one is the problem "somehow" in scheduling?

[Gordon-F]

But we'd like to have a clean emulator for every artifact that we need to test.

Android-emulator-runner docs suggest using a clean snapshot to speed up CI in this case. What do you think about it? It should speed up CI time and maybe help with the runner issue.

[MarijnS95]

Yeah, we can try that :)

[MarijnS95]

It's a bit strange because this shouldn't happen regardless (I kept the set -x from aforementioned PR #169, and it doesn't show anything out of the ordinary).

[MarijnS95]

@Gordon-F It also seems to have been fixed by disabling animations, GPU etc from that page, see #198. We can add caching in a separate PR and see how well that speeds things up separate from fixing the regression lockup.