[NOSQUASH] TouchScreenGUI refactor

[srifqi]

Goal of the PR
This PR refactors TouchScreenGUI code to make it more safe and maintainable.

How does the PR work?
There are two parts of this PR (separated by two commits):

1. Fixes memory leaks and code style issues.
2. Simplifies calculations by using better data type.

Does it resolve any reported issue?
There are no reported issue on this, but I noticed that some members/properties/variables were not removed correctly after its parent object is destroyed. There are also some repeating parts inside the code, mainly for the virtual joystick.

Does this relate to a goal in the roadmap?
Yes, this PR tries to refactor internal code.

If not a bug fix, why is this PR needed? What usecases does it solve?
From my tests, there are indications of memory leaks when using TouchScreenGUI code.

To do

This PR is Ready for Review.

How to test

To test the change statically,

1. Read the changes in the code base.
2. Verify that the old code and the new code are the same functionally.

To test the change practically,

1. Run Minetest on a touch-screen device.
2. Play in a world or on a server.
3. There should be no behaviour changes when using touch-screen GUI.
4. Back to Menu.
5. There should be no crashes.

To spesifically test memory leaks indication,

1. Run Minetest (without this PR) on a touch-screen device.
2. Play in a world or on a server.
3. Wait until its RAM usage has settled.
4. Note down its RAM usage.
5. Back to Menu.
6. Repeat step number 2–5 for a few times.
7. Exit to OS.
8. Run Minetest (with this PR) on a touch-screen device.
9. Repeat step number 2–7 on this new version.
10. Compare the results.

Android: APK builds

• armeabi-v7a
• arm64-v8a
• x86
• x86_64

Older APK builds

• armeabi-v7a
• arm64-v8a
• x86
• x86_64

[srifqi]

I tested this PR on Samsung Galaxy A72 (Android 13). I waited around 10 seconds in a world before sampling its RAM usage.

RAM usage in-game without exiting the app

Play Num.	master	(+)	This PR	(+)
1	224.0 MB	-	230.6 MB	-
2	229.3 MB	+5.3	235.5 MB	+4.9
3	234.2 MB	+4.9	237.5 MB	+2.0
4	238.0 MB	+3.8	239.3 MB	+1.8
5	240.5 MB	+2.5	241.4 MB	+2.1
6	244.4 MB	+3.9	243.8 MB	+2.4
7	248.3 MB	+3.9	246.1 MB	+2.3
Average		+4.1		+2.6

[SmallJoker]

The code looks good, the build passes and some tests were already done.
Minor comment, otherwise fine.

[SmallJoker]

I got the following error with your v7a binary on Android 6 even after deleting /storage/emulated/0/Android/data/net.minetest.minetest manually and re-installing the app. I tested a self-signed binary from master which worked flawlessly, thus I assume there were a few bugfixes that are yet not included in your binary. Whereas this issue is pretty sure not caused by the PR itself, I think it should be mentioned either way - just in case.

2023-07-10 20:05:09: ERROR[Main]: Irrlicht: PNG fatal error: IHDR: CRC error
2023-07-10 20:05:09: WARNING[Main]: Could not load icon file.
2023-07-10 20:05:09: ERROR[Main]: Irrlicht: PNG fatal error: IHDR: CRC error
2023-07-10 20:05:09: ERROR[Main]: Irrlicht: Could not load texture: /storage/emulated/0/Android/data/net.minetest.minetest/files/Minetest/textures/base/pack/checkbox_16.png
2023-07-10 20:05:11: ERROR[AsyncWorker-0]: HTTPFetch for servers.minetest.net/list?proto_version_min=37&proto_version_max=43 failed (Error)
2023-07-10 20:05:44: ERROR[Main]: Irrlicht: PNG fatal error: IHDR: CRC error
2023-07-10 20:05:44: ERROR[Main]: generateImage(): Could not load image "progress_bar.png" while building texture; Creating a dummy image
2023-07-10 20:05:44: ERROR[Main]: Irrlicht: PNG fatal error: IHDR: CRC error
2023-07-10 20:05:44: ERROR[Main]: generateImage(): Could not load image "progress_bar_bg.png" while building texture; Creating a dummy image
2023-07-10 20:05:44: ERROR[Main]: ModError: Failed to load and run script from /storage/emulated/0/Android/data/net.minetest.minetest/files/Minetest/games/minetest_game/mods/default/init.lua:
2023-07-10 20:05:44: ERROR[Main]: dcompressZlib: inflateInit failed
2023-07-10 20:05:44: ERROR[Main]: stack traceback:
2023-07-10 20:05:44: ERROR[Main]: 	[C]: in function 'orig_reg_fn'
2023-07-10 20:05:44: ERROR[Main]: 	...netest.minetest/files/Minetest/builtin/game/register.lua:445: in function 'register_decoration'
2023-07-10 20:05:44: ERROR[Main]: 	...les/Minetest/games/minetest_game/mods/default/mapgen.lua:1815: in function 'register_decorations'
2023-07-10 20:05:44: ERROR[Main]: 	...les/Minetest/games/minetest_game/mods/default/mapgen.lua:2491: in main chunk
2023-07-10 20:05:44: ERROR[Main]: 	[C]: in function 'dofile'
2023-07-10 20:05:44: ERROR[Main]: 	...files/Minetest/games/minetest_game/mods/default/init.lua:77: in main chunk
2023-07-10 20:05:44: ERROR[Main]: Für mehr Details siehe debug.txt.
2023-07-10 20:05:44: ACTION[Main]: Server: Shutting down

[SmallJoker]

Code LGTM

[srifqi]

I have updated the APK builds (see the main post) with this PR after rebase and current master of the Minetest Game.

[SmallJoker]

What I forgot to mention previously (and also applies to the newer build):

• MTG does not show up in the game menu and there is no prompt to install a new game (outdated builtin?)
• No icons can be loaded

I found that the directory /Android/data/net.minetest.minetest/files/Minetest/ is somehow cached by gvfs on Linux when mounting via MTP, thus I ended up looking at invalid directory contents without noticing. This is somewhat annoying, thus I have now opened an issue on libmtp in hopes to solve this.

Either way, the buildbot binary and yours both seem to install the same files. I double-checked the textures that failed to load but they show up in the file browser. Also a texture pack that I might have installed previously magically reappears even after deleting Minetest's data directory manually.

I cannot figure out what's wrong and since the build from master does work, I'd prefer to not dig into this any deeper.