Progress report for L3DT 2.3b

Release 2.3b is chugging along nicely. The dev plan, recently updated, lists the features and fixes that are ready-to-go, and those still unfinished. Some highlights include:

I made a few mods to the erosion routine, and it now gives better results for the same calculation time. Although I could use this extra efficiency to increase the speed of the calculation, I'm leaning towards increasing the default erosion strength and preserving the calc time. Which would you prefer? Faster or better, or a little of both? (let me guess - you want an interface so that you can choose, right?)

The current release of L3DT produces really lumpy-looking terrain when you use large horizontal-scale values for the heightmap (eg 500m/pxl, 10km/pxl, etc). This is because the strength of the peak-overlay algorithm ramps-up linearly with horizontal scale - which is fine for small values (eg 0.1m…10m), but becomes ridiculously for large values (>100m). I've modified the strength equation so that it asymptotes to a maximum value for large horizontal scales (S = MaxVal × (1 - e^{-λ×HorizScale}), for the math nuts). It seems to work OK across the tested range of 0.1m/pxl to 10km/pxl.

Also, the standard climates bundled with L3DT don't work very well when you use very small horizontal scale values (eg 25cm/pxl). I've re-balanced the temperate climate to be better-behaved, and I've also made a new climate file specifically for these small-scale maps (again based on 'temperate').

The was a nasty intermittent bug in the 'Export to Enviro' feature that sometimes garbled the file paths and caused odd errors. I think I've now fixed this, but some serious evaluation is required.

There was also a bug in the progress display dialog box that sometimes spewed thousands of 'CMapWrap::GetPixel' errors. It was a concurrent thread access problem, which only ever showed-up on slower CPU's (eg 450Mhz). I think I've nailed this one, but I'm waiting for confirmation of this.

Some good progress has been made on the function to export the texture-blending alpha-maps as RGBA images. The difficult part was building an interface to allow users to combine different texture layers into the same alpha-layer. This is to allow users to reduce the number of alpha layers from the 16-or-so typically used by L3DT when generating textures, to something more manageable for run-time blending (eg 4 layers in an RGBA image). Like most L3DT interfaces, it's probably not very intuitive. I'll smooth the rough-edges once I have some user feedback…once it's ready, that is.

Probably 1-2 months from now. I want to debug this one thoroughly.