Bump to SDL2 2.0.10 & extract .java from SDL2 tarball: merge conflicts fixed

[inclement]

This is the code from #1779 with (minor) merge conflicts fixed.

[inclement]

If the tests pass, I intend to merge this right away.

[opacam]

The flake 8 errors now are fixed by #2114, a new merge/rebase of develop branch should fix that

[AndreMiras]

Yes 5K lines of code less 👏
Can we rebase it to see if tests are passing?
Can't wait to have it merged!

[AndreMiras]

We could save one line by simulating the --parents flag:

Suggested change

	if not os.path.exists(target_dir):
	os.makedirs(target_dir)
	os.makedirs(target_dir, exist_ok=True)

[ghost]

I think I wrote it like this to maintain Python 2 compatibility. So yes, for 3.x-only shortening it like this makes sense 👍

[inclement]

Unsurprisingly there are some minor test failures, but the app builds seem to work fine. I'll see if I can fix it quickly.

I'm finding that test_pythonpackage.py runs out my ram within a few seconds, which didn't happen in the past. Not sure what's up there.

[AndreMiras]

I actually often skip the test_pythonpackage.py in local because for me it always ran for ages

[opacam]

If I remember well...the problem of these tests are that we don't mock at all the file/dir reading/creation and this has a serious impact to the tests performance. So, as @AndreMiras, I almost always skip these tests as well.

Anyway, I also tested one of the generated apks for on_device_unit_test_app, the one for arm64-v8a, and it works ok, actually I think that it works better, since the kivy's keyboard features now I think that are working as expected.

[ghost]

For what it's worth, test_pythonpackage.py tests with pip because pip has no stable interface and likes to break things, and there is no libpip or other library-level alternative. So this is kind of an external problem that cannot be easily solved unless pip decides to do so.

Edit: the out of memory thing sounds strange though. How much did it have available to run?

[inclement]

Got sidetracked on this but still trying to prioritise it over anything else.

I fixed a bug with the sdl2 patching, on runs after the first one it was partially-applying the patch before deciding the patch wasn't needed, breaking the java code. I've made it do a dry run instead, which seems to resolve that issue.

I want to do a bit more testing, but as far as I know it's working fine now.

[inclement]

Fixed tox, I think, by fixing recipe cache clearing.

[inclement]

Fixed another bug, in test_recipe.py. This one only occurs if you run the tests from an unexpected directory, as the test tries to listdir a directory that doesn't exist because the exists call has been overridden. To keep things simple, I changed the test state so that this directory is not in the search path, but I might revisit this in another PR.

[opacam]

💯

Awesome, ¡¡so many thanks!!

[AndreMiras]

Looking good! Thanks for spending time on resuming this work. Awesome to see 5K lines ditched 👏
Thank you @etc0de for the initial initiative ❤️