So, I opened two bugs :
“Android, is eglGetProcAddress still busted ?” (bug #4040)
“Android, SDL_Renderer OpenGLES 1 is loading GLESv2 library” (bug #4041)
Well, I opened a bug too, since I was the one who discovered the problem. Indeed you suggested that I should. Since neither of yours even mentions SDL_GL_GetProcAddress, which is the failing function, Bugzilla didn’t suggest them as similar.
I’ve been doing a little more research on this issue. As noted previously, on platforms other than Android SDL_GL_GetProcAddress simply calls eglGetProcAddress, and to fix my crash I’ve patched sdl_egl.c so that it does so on Android too. However the docs of eglGetProcAddress state that it should be used only to return the address of an OpenGL(ES) extension function, not a core function.
It seems to me that this is potentially a problem. The docs of SDL_GL_GetProcAddress state that although it is usually used “to retrieve function pointers to OpenGL extensions” it may also be used to retrieve the address of a core function so long as you verify “that the version of OpenGL you’re using offers the function as core functionality”.
This is the case in my app: I’m using SDL_GL_GetProcAddress to retrieve the addresses of glEnable and glDisable. But if the docs of eglGetProcAddress are to be believed surely this is not guaranteed to work, because these are core functions. So is SDL relying on undocumented behavior here, and can I be confident it will always work?