InitRenderModel should be called before InitPosition (to have fixed
light positions)

better perspective projection handling (around cog+new icon in gui)

refactored Opengl_Window.handle()

should behave exactly like the old version, but now it is at least (somewhat)
readable and ready to be generalized for more complicated projection matrices

cleanup+add comments so that students can understand what's going on with
the projection stuff.

(still need to cleanup the zooming in handle() to use gluUnproject() instead
of the ad-hock stuff that only works for an orthographic proj)

little hack to draw a gradient in the background. Nice for flashy
presentations ;-)

tweak automatic polygon offset check

small cleanup + don't enter post stuff in select mode at the moment (we
might want to change this if we want to select post-pro views in the future)

- made meshed curves/surfaces selectable with the mouse
- better selection algorithm (use depth info to find the "closest" match)

removed unused Post.Scales option

a little more work on the axes stuff: enable global axes for the geometry and
the mesh, too. We can finally get rid of the ugly old "moving axes" bit.

(+ reverted changes in unproject: need to be more careful since unproject
can be called outside Draw())

- cleaned up (somewhat) + finished the euler angles/quaternion handling

- added a new little dialog in gui to specify the rotations/scales/translations by
  hand. This is similar to what I did in the Motif version of Gmsh a looooong time
  ago, and I was really missing this capability... (e.g. being able to specify
  exact rotation angles)

- the grid options now also apply to 3D views (this way we can
  finally get accurate size info for any data set; and the info
  is updated in real-time for gemeotry transformations & displacement
  maps)

- new option to specify the size of the moving axes

- generalized text align code (we now support the 9 modes that TeX
  does)

removed all unnecessary glPushMatrix/glPopMatrix calls

copyright update

- More intelligent way to detect if we need to apply a polygon offset
or not.

- better computation of translation coef in Draw2D

Added 2 options to control polygon offset. Instead of trying to guess
when a polygon offset is useful (it's no that simple when we combine
geo+mesh+some views), we just apply it conditionally to all surface
drawings.

simpler solution to 2D primitive sorting problem

translate all 2D objects before drawing them: this is the easiest way
to get GL2PS to put them "in front"

- plugin callbacks now draw their stuff until run/cancel is executed (so we
  manipulate the scene with the additional data displayed)
- made plugin code much shorter
- simpler defaults for StreamLines, CutGrid and Evaluate

New General.ClipFactor option to fine-tune the position of the near and far
clipping planes (default = 10; setting a smaller value leads to a better
z-buffer resolution, but zooms/rotations can then lead to undesired clipping).

Revert to GL_NORMALIZE instead of GL_RESCALE_NORMAL (since GL_RESCALE_NORMAL
is not supposed to work with anisotropic scalings)

trying to fix problem reported by T. Schumacher: use double[16] instead of
double[4][4] for opengl matrices

- make the rotation matrix a matrix of doubles (instead of floats)

- store the modelview and the projection matrices in CTX

- new unproject() routine to transform window coordinates to model
  coordinates

- cleaned up all the static variables in the Opengl_Window class

- small fix to reduce normal clamping when zooming

added W coordinate for light sources (so that we can have "spot sources")

fix support for multiple light sources (+ add a couple of options to set
the light color and specular reflection exponent)

draw intersection of cutting planes with the bounding box

quick hack to implement "fake transparency" as in xpost

Test if we have OpenGL 1.2 to use GL_RESCALE_NORMAL. If not, use the more
expensive GL_NORMALIZE. Thanks Dave!

slightly better computation of the z-scaling in lasso zoom to avoid
normal clamping

- don't call glNormal unnecessarily

- do The Right Thing (TM) with the light position, i.e., set it
  *before* we multiply by the modelview matrix. It's much easier
  to achieve predictable lighting results like this: (-1,0,0)
  simply means "light from the left", (0,1,0) means "light from
  the top", (0,0,1) means "light from the front", etc.

- fixed orientation of boundary of post tet

- added new LightTwoSide option: with this turned off, we achieve
  roughly the same performance as SceneViewer on large data sets, i.e.,
  around 1 million triangles (SceneViwer is the "industry standard"
  OpenGL viewer from SGI). Nice... ;-)

use GL_NORMAL_RESIZE instead of GL_NORMALIZE: it's faster

display the interactive messages in the graphic window (instead of the
status bar)

removed unused geom.highlight option + small cleanup of the selection code

removed glEnable(GL_ALPHA): this is not even in the GL doc :-)

- only save the options that differ from the default values (unless we
  explicitly ask to save everything)

- tweak the auto doc routines

Fixed orientation of polygons produced by CutTriangle2D so that we can
use a TWO_SIDE light model everywhere. This also fixes the ugly "checkboard
patterns" when using custom scales on continuous maps.

- Simplified the bounding box computation. It is now computed as
follows (after each include/merge/open):

1. If there is a geometry (i.e., at least one geometrical point), the
bounding box is taken as the box enclosing all the geometrical points;

2. If there is no geometry but there is a mesh (i.e., at least one
mesh vertex), the bounding box is taken as the box enclosing all the
mesh vertices;

3. If there is no geometry and no mesh, but there are some
post-processing views, the bounding box is taken as the box enclosing
all the primitives of the last post-processing view.

This should fix the weird things that sometimes happen when we animate
scenes that do not contain any geometrical entities.

- Deprecated MergeWithBoundingBox()

- Removed Replot()

- Removed the "Last_NumberOfPoints" hack in the parser

- Removed CTX.expose

More "airplane" work...

- rewrote the way we handle lights: glEnable(GL_LIGHTING) is now used
  at the lowest level, where it belongs. This fixes many lighting
  glitches, and should make the whole stuff much more predictable;

- simplified the display modes for geometry and mesh: the choice
  between wireframe/solid is now independent of the choice of
  lighting/no lighting (and there is a new "Enable lighting" option
  for the geometry, too--even if it does only affect things like
  normals/tangents at the moment);

- brand new code to draw really nice (shaded) 3D arrows + 3 options
  that fully parameterize them (they can degenerate into pyramids,
  cones, w/ or w/o stems, etc.);

- new options so that we can also use the new arrows outside the
  post-processing module (e.g. for tangents/normals).

- "alt+d" now simply switches between solid and wireframe mode;

- new shortcut "alt+w" to switch the lighting mode for all the modules
  (geo/mesh + all post views) at once.