Other than creating the test scene it's pretty much the same as before. Variables for the light position/intensity have moved into uniforms so they're changable from code, but I'm not worrying too much about the few constants and arbitrary scale factors yet as without a proper game behind it I don't know which needs to be tweekable and which can remain hardcoded.
Other than being fun and quite interesting, these graphics tests are also helping me design my new graphics renderer. My requirements for a 2d renderer are quite different from a regular 2d 'engine', or indeed a fancy 3d renderer. Most 2d OpenGL renderers attempt to emulate old-school blitting APIs, and so while they're fast and easy to use they don't really take full advantage of the graphics card. 3d renderers on the other hand tend to favor an explicit feature set which is very tweekable.
I'm after something different - I've got a whole bunch of games all running off the same codebase, and I like to do something different with each ones. Doing this like a traditional 3d engine would result in a massive, complicated renderer with all sorts of effects built in - but each game is only using a one or two effects each, so theres a lot of unwanted complexity and baggage.
What I'd like is a collection of parts which I can quickly snap together to create a custom renderer really quickly. They've got to work well with each other, but they've also got to make adding small custom chunks of code easy so that new effects can be added. In return I'm willing to sacrifice a certain amount of performance - more specifically, I'm not going to worry too much about CPU usage or raw geometry throughput. CPU doesn't matter too much because 2d games always tend to have plenty of cycles to spare. Geometry throughput doesn't matter too much because I'm always fill-limited (practically everything in a 2d game has to be blended in some way and it all adds up).
I'm still juggling ideas in my head, and I suspect I will be for at least another couple of weeks before things start to solidify. In particular I've got to figure out where to draw the line between geometry state and effect state, and whether I'm going to make multi-pass drawing built-in and handled for you or whether you'll have to manage it explicityly yourself (which is going to be a classic ease-of-use vs. customability trade off I think).