2010/1/28 Stefan Dösinger stefandoesinger@gmx.at:
On Wednesday 27 January 2010 23:41:27 Henri Verbeet wrote:
The problem with the effect interface is that there are several fairly large parts/dependencies to implement. For example, it has important dependencies on both the (non-existent) HLSL compiler and the only partially merged shader assembler
There's another problem with any d3dx9 project: You cannot test a small subset of D3DX9 with any real-word game if the rest of the library is missing because no games use only the event interface or use only the assembler etc.
Matteo faced this issue - there was no game that used the assembler, but didn't use some of the Font or Texture loading code Tony was working on. This meant he could only test with test cases written himself and had no success moment when a real game started working without native d3dx9 thanks to his work.
Yes, that's a relevant issue. I didn't find a single game, or even DirectX SDK sample, which needed only the shader assembler functions and not also some other unimplemented part of d3dx9 (expecially the texture functions are often used). This essentially means that you'd have to rely only on conformance tests to have some measure of correctness of your code (and fidelity to the behavior of native implementation), at least in the short to medium period.
So before any further d3dx9 projects have a realistic chance at succeeding we need the existing code merged. If necessary I can pick up Matteo's patches, split them up and send them in, although it would be preferable if Matteo could do this himself.
I have pretty much splitted the patches up in reasonable enough (I believe) pieces, now I just need to address the things Henri noticed last time I sent the main patches. That's not much to fix, but these days I'm really busy. Hopefully in some days I could dedicate some time to this.