Uhm, nothing important has changed since then and every existing backend is, like, a properly working and up to date example of a backend I suppose. But something like Magic Leap support doesn't even require its own backend, it's just another VR/AR device after all if I'm not mistaken. Look at the Rift/Vive/Hololens support in Kore.

The html5 backend is less aggressive in throwing out unnecessary data but that might change so be careful. It won't work at all in any of the native backends without readable set to true.

For non-rendertargets use the at function instead (we should probably unify that). But for that to work, the images have be be loaded with the readable option. When your assets are defined in your khafile you can just add that option in there ala project.addAssets("Whatever/**", { readable: true }); and if you load your images manually readable is a parameter of Assets.loadImageFromPath. Ideally you only use readable for images which really need it because it doubles RAM usage (images have to be kept in GPU mem for rendering and in CPU mem for direct access).

It's Kha/make unity. But I haven't updated the Unity backend in a long time, it's currently not working. Unsure whether it's worth it now that we have direct console support.

Nope, not supported but feel free to send a pull request.

The original OpenGLWindow.h issue is now fixed.

Oh, please don't double-post. For reference, here is the answer: https://github.com/Kode/Kha/issues/894

Shouldn't be necessary to add that include path manually, how exactly did you create the Xcode project?
Clearing via Metal might indeed be broken, will have a look.

It's no longer a thing, everything moved into Main.hx.

The Switch backend is a plugin which we are only allowed to send to you if you can prove that you have a valid Switch developer account with Nintendo. Same for PlayStation and Xbox. If you do have an account please send me a private message right here in the forum.

Yes (just for the record, already talked about it in irc).

Oh, that's mysterious.

Ah, that's what the chat post was hinting at. Hm. So the old code works... sometimes?

Yes, you're at the right spot.

No, that's just the user-callback. The backend is supposed to call the __leave function to trigger the user-callback.

They should work but it in OpenGL only (cross-compilation to HLSL isn't finished and there just are no geometry shaders in Metal). Upload a test project please.

The hide the mouse for the whole screen is problematic and it's not how it works on other targets. People won't know how to close the window. Guess we need to build in some mechanism that automatically reverts the mouse-hide call when the mouse moves out of the window.

Oh and one more hint: You can create a Kore application to test/implement things in native backends so you don't have to compile all of the Haxe things.

There's a lib called xfixes? That's quite hilarious. Just read up on it and I think it's ok to include it. Would you like to create a pull request? That'll give you proper credit in the git commit list.

Oh Maybe it'll fix itself at some point then.