[FernetMenta]

- rework interface to video renderers: https://github.com/FernetMenta/xbmc/comm...511aa089ad
- refactoring of renderManager and LinuxRendererGL: https://github.com/FernetMenta/xbmc/comm...9dc82b46ff

This allows use of appropriate render method on demand, i.e. hw decoders like VAAPI have their own renderer

- kill global variable g_renderManager and move all player related methods to a single interface: IPlayer.h https://github.com/FernetMenta/xbmc/comm...1d70056a0d

- rename dvdplayer to VideoPlayer: https://github.com/FernetMenta/xbmc/comm...1692e2ca96

I have already adjusted the OpenGL based platforms: Linux and OSX.

[Memphiz]

So for ios i would start by implementing:

xbmc/cores/VideoRenderers/HwDecRender/RendererVTB.cpp for example and use the needed bits from the old VIdeoRendererGLES.cpp - right?

Also i would need to mimic your refactor of LinuxRendererGLES - though i am not quiet sure what the constraints for this refactor are.

[FernetMenta]

Yes, you would need to adapt LinuxRendererGLES and RenderVTB similar to LinuxRendererGL and RenderVDA. All platform dependent code should leave LinuxRendererGLES. No need to care about the other platforms in this stage when cleaning LinuxRendererGLES.

[popcornmix]

I have a branch where this code builds for Pi, and on a quick test, software decode, MMAL accelerated and omxplayer were working:
https://github.com/popcornmix/xbmc/tree/fernet

[Martijn]

@FernetMenta would you prefer devs create PRs against your branch and you squash them in or some other work method?

[FernetMenta]

for the time being I think it's best to PR against my master. I don't know yet how to structure the changes for submission to mainline. I always try not to break bisect but not sure if this will be possible in this case or it will result in a couple of huge commits.