deal with a index_buffer memory leak that Valgrind has always complained about
make HTL model buffers dynamic (get rid of MAX_BUFFERS_PER_SUBMODEL)
get rid of MAX_BUFFERS
make D3D vertex buffers dynamic, like OGL has already done
- don't alloc it for bmpman, and make sure to only call on textures that we are loading
- add text for model loading, gets rid of the long-wait issue when it appears to not be doing anything early on
- put glowmap activity back on a per-ship basis (via a SF2_* flag) rather than per-model
- same for glowpoints, back on a per-ship basis
- put specmaps and bumpmap back on a LOD0 and LOD1 affect (got changed to LOD0 only recently)
- fix glowmaps for shockwaves again
- add support for animated specmaps (mainly for TBP and Starfox mods)
some minor code cleanup and compiler warning fixes
minor cleanup of some modelinterp.cpp code, to make it more readable
fix tertiary thruster batcher which had the radius value swapped
comment out a couple places where we were doing a D3D or OGL check (it's always true so why bother)
add more support for NO_DIRECT3D so that it can be used on Windows to build without D3D
(there is more of this same checking on the way for the other malloc() calls during model loading, and vm_malloc() will be set to
Assert() on a 0 or less size as well)
more bmpman related fixes, mainly addresses the condition where it would try to render textures that have been released
have model_unload() run through Ship_info[] and clear out modelnum's when they are free'd (makes more sense here than doing it on level load to keep counts right)
fix missing subsystem warning messages
let gr_render_buffer() render untextured polys (OGL only until some D3D people fix it on their side)
add MR_SHOW_OUTLINE_HTL flag so we easily render using HTL mode for wireframe views
make Interp_verts/Interp_norms/etc. dynamic and get rid of the extra htl_* versions
add -noibx troubleshooting cmdline option to disable use of IBX files
don't try to set thuster object number is only one thruster bank is specified (default method should go into affect)
clean up some modelinterp code to not use memcpy() everywhere so it's more platform compatible and matches old code (Jens Granseuer)
NaN check to catch shards-of-death and prevent hitting an Assert() (Jens Granseuer)
fix jumpnode code to catch model errors and close a memory leak
make the call to bm_unload_all() after model_free_all() since we will get bmpman screwups otherwise
don't show hardware sound RAM when using OpenAL build, it will always be 0
print top-right memory figures in debug builds slighly further over when 1024+ res
taylor