depth testing and 2 pass render in MCv3 - why?

[zb3]

Hey!

I'll probably ask more questions as I'm preparing a new commit to my freej2me work (it already includes your engine but the new commit will switch gles2 wrappers), but now - I compared some rendering cases on the SEMC sdk emulator and yours and I see that SEMC doesn't seem to have any kind of depth testing.. it just relies on polygon sorting.. or at least, it appears to.

(note: I don't have a real SE phone, if the SEMC emulator diverges here - apologies, I didn't know about that, that's why I ask)

But here I see that you render in two passes, with flushStep == 1 being for opaque polygons while flushStep == 2 being for those transparent. Why is it done this way in your code?

Like I said, there are many more differences I noticed, but I'll ask about these when the code is ready..

[woesss]

I have never worked with 3D, so it turned out the way it did - this is the only explanation why I did it this way.
The original render is software and as lightweight as possible, so it is unrealistic to achieve an exact match of the output image using GLES2 (at least for me)