Korlib will have two components: a python implementation in Korman and a
C++ "_korlib" module that will be insanely fast. Users will apreciate it,
and devs will be glad they don't have to compile the C version.
To ensure that there are really, really, REALLY no race conditions related
to coordinate interfaces, we now run through all modifiers before we
export and ask them if they need to make Coordinate Interfaces. I was
hearing some comments about clickables warping around. This sounds like
physical coordinate issues to me...
- Add an opacity value for exported layer.
- Add a conversion button in the toolbar for old PyPRP Ages
which used the diffuse color factor as the opacity value.
After discussion in #writers IRC and observing how Blender's image
environment maps work, it makes more sense to rely on our superior
implementation of dynamic maps.
This means we now export BOTH plDyanmicEnvMap and plDynamicCamMap. The
latter will only be exported on MOUL+, sadly. The former is "cube" mapping
and the latter is "plane" mapping. Instead of manually specifying the
refresh rate, I decided that Blender's "static" option means that it only
refreshes once (and never again) and animated means we refresh every 0.01
second, which seems to be the Cyan standard for "update every frame, blast
it!" NOTE: when cameras are exporting, we need to support Camera based
DCMs.
Static image based CEMs coming soon...
Cyan calculates the level sizes wrongly such that a 1x1 DXT5 mip level
only has one block for alpha and therefore lacks the color block. This can
cause Blender to crash with an access violation.
Still lots of considerations:
- Need to make faux lightgroups to avoid baking Plasma RT lights
- Still need to bake vertex colors
- Still need to export Plasma RT lights (D'oh!)
But, this is definitely a step in the right direction!
We now harvest information about texture images throughout the export
process and don't pull the trigger until we're totally done. There are
still some issues with regard to confusing UI, but we'll get to it.
Yeah, korlib was a bit faster than this implementation, but, honestly,
having a second python module made life more difficult than it needed to
be. This fixes that.
It appears that we were storing PyHSPlasma objects in class attributes, so
they were being held after the plResManager went away--therefore their
keys were nuked. We should probably be holding keys and not objects, but
this fix works regardless.
