How confident can I be that the code to get & set
pixels in the example code (under “Drawing directly
to the display”) will be correct in the future?
As long as the concept of a framebuffer doesn’t change, you should be fine.
Is there a reason SDL doesn’t provide “SDL_GetPixel()/
SDL_SetPixel()” calls?
Getting and setting a pixel is a slow operation.
My thought that was if you need to do this, you can probably optimize the
surface access better than I can, by knowing more about the type of access
and format of the surface than the library will.