Make it work under WIN32 (ugh); and allow CAVETime.
Allow null time ranges so partiview can suppress time bar
if nobody has nonempty range.

Add parti_allobjs() to implement "gall" command.

Don't show time-control bar unless some object has a nonzero
time range.

"depthsort" adds depth-sorted polygons.
New texture qualifiers -i(ntensity) -a(lpha) -A(dd).
New "gall" (a.k.a. "allobj") cmd sends following command to all objects.
"everycomp" says whether to compensate for "every" subsampling.

Use config.h's WORDS_BIGENDIAN.

Add new "trange" command to specify a limited range of times we're
allowed to view.

Add "kira track" command for particle tracking.

Re-extract whenever kira_set changes something.
Add particle tracking.
Compute ring sizes correctly, and draw rounder rings.

Keyboard shortcuts "<" and ">" stop animation.
Shortcuts "{" and "}" start animation (backwards & forwards).

Add parti_nudge_camera() to aid in tracking objects.

Get slider position right if min/max values change.

Improve color choices, jog-roller resolution.

Fix some clock bugs.
Middle- or right-click on "step" buttons advance by .1 or 10 times normal step.
Tie step to speed: step = 0.1sec * speed.

Tidy up messages -- don't append \n in msg() calls.

Add ring markers for non-isolated stars.  Not finished yet, but seems usable.
No GUI yet, but new commands:
   kiractl nodes  {on|off|roots}
   kiractl rings  {on|off|roots}
   kiractl ringsize {sep|semi}   (ring size from curr. separation or semimajor axis)
   kiractl scale  <scalefactor>
   kiractl span   <minpixels> <maxpixels>   (allowed range of ring radii)

Change encoding of nclump: n for leaves, -n for center-of-mass nodes,
   so colormap can make them distinguishable.

Bug: after kiractl command, must change timestep before it takes effect.
Bug: ring markers aren't colored yet for some reason.

Add -lm to -lGL/-lMesaGL tests, to ensure it doesn't fail to find GL
just because libGL/libGLU use sqrt etc.!
Discard tests which search for IrisGL and PEX, since we can't use them anyway.
Re-install "depended"/"Makedepend" fix.  We really do want to force
Makedepend to be rebuilt, I think.

Reduce SPECKCHUNK to 2001, not 10001.  Too many machines have tiny stacks.

Lighten the text in timebase panel.

Rather than archiving an (incorrect) copy of Makedepend,
just have "configure" create an empty Makedepend and force
"make depend" to be invoked if the user then says "make".

We don't want to archive this file -- it's system-dependent!
Instead, "configure" does a "touch Makedepend",
and "Makefile" uses a "depended" tag-file to ensure that
"make depend" actually gets run once.

Handle wrap band correctly (no extra ";"!).

Add flag for "user limited time range", and wrap-band interval.

User-adjustable wraparound time (but no way to adjust it yet).
Recompute colors and sizes whenever we update dynamic data.
Don't update dynamic data if time hasn't changed.
Discard old specks when updating dynamic data -- but allow for
keeping "ghost" copies (up to st->nghosts) for making trails.
We don't do trails yet though.

Recode to be speedier (use recursion) and to provide more information to partiview:
	[0] worldline index (= kira index for singles, - and unique for others)
	[1] mass
	[2] number of stars (leaf nodes) in clump
		+ 100 if this is a member of an unperturbed binary
	[3] top-level name (small integer, = worldline index for single stars)
	[4] tree address in clump (0=single, 1=root, children of <i> are <2i>,<2i+1>)

Make ...::format() return count.

Add new "nghosts" field for keeping animation history (trails).

Tidy.

Add default case to ...::printvalue().

Lots of changes to support new time flow and additional
time-related GUI.  Probably not all there but at least somewhat usable.

Add new files kira_parti (Steve McMillan's glue to starlab),
sclock (clock), notify (notification service -- not yet used),
Fl_Log_Slider (log-scaled slider).

Glue code to read starlab (kira) files and interpolate particle positions
into partiview internal form.

Now use worldline.h from ${KIRA}/inc directory rather than
the one hidden deep in the source.

Don't report error if WORDS_BIGENDIAN is undefined -- config.h might
just /* #undef */ it.  Just assume undefined -> false.

Add default rules for .cc (and .C and .cpp) files -- not all versions of make
have them built-in.

We're little-endian if WORDS_BIGENDIAN is defined (in config.h) and set to 0,
so use #if rather than #ifdef to test endian-ness.