flypig.co.uk

I thought I'd already had a go at applying patch 0038 "Fix mesa egl display and buffer initialisation" on Day 55. Looking back at my notes, I did try, but probably not hard enough. The patch failed to stick in many places and now I'm looking at the code it seems there are some really large gaps.

Trying to make a better job of it this time around I've been working my way carefully through the entire patch. I've manually applied all of the changes. As I mentioned yesterday, the patch contains a lot of material important for keeping Sailfish Browser working.

Now when I attempt to build the gecko library I start getting errors in far flung places in the code due to the way the API is accessed. The output below is typical of the errors I'm getting.

${PROJECT}/gecko-dev/gfx/gl/AndroidSurfaceTexture.cpp: In static member 
    function ‘static already_AddRefed<mozilla::gl::GLContextEGL> mozilla::gl::
    AndroidSharedBlitGL::CreateContextImpl(bool)’:
${PROJECT}/gecko-dev/gfx/gl/AndroidSurfaceTexture.cpp:78:52: error: too few 
    arguments to function  std::shared_ptr<mozilla::gl::EglDisplay> mozilla::gl:
    :DefaultEglDisplay(nsACString*, EGLDisplay)’
     const auto egl = gl::DefaultEglDisplay(&ignored);
                                                    ^

The reason for this error is clear: I changed some of the method signatures to include a display parameter. And while I've tried to update the code so that it's called correctly in all of the obvious places, it turns out there are some less obvious places too. In practice it looks to me like these are Android-specific, but if the compiler is going over them they'll need to be fixed.

An obvious fix would be to make the parameter optional. But these errors do serve a purpose in exposing when things have changed, so it's convenient not to make it optional just yet. Maybe once everything is working I can return to this and add it as an optional parameter for the purposes of simplifying the code and reducing the size of the patches. But for now the added protection of knowing exactly when and where the calls are made is convenient.

I've not yet got through all of the changes so I'll have to continue tomorrow. I feel like I've been making quite good progress though. I just need to get the last pieces compiling.

As an aside, I finally got to watch Peter Mack's (1peter10's) FOSDEM'24 talk on The Linux Phone App Ecosystem. I've been meaning to watch it from before the event and really enjoyed it, especially the way it straddled so many different Linux-based mobile operating systems. If you've not watched it yet yourself I highly recommend it.

I'm hoping to get that completed and compiling tomorrow. In the meantime if you'd like to read any of my other gecko diary entries, they're all available on my Gecko-dev Diary page.