Use glshader.h, and initialize shader stuff (once pre- and again post-window creation).

Not in use yet.

Add "vec/vectors" and "vecscale" control commands and "vecvar" data command.
When enabled, draws a vector at each point, as specified by some
triple of attributes given by "vecvar".  Length is [vecvar .. vecvar+2]
scaled by vecscale (default 1).

Add data for representing traj files (not yet used either).
Add data for drawing vectors at each point (this one is implemented now).

if we're using multiple source datasteps (in which case tfrac = realtime mod 1.0).
Check that the original realtime matches too.

of main window.  Apparently this used to be necessary to trigger initial redraw,
but (in more recent versions of FLTK?) just causes the main window to become
non-resizable.  Toss it.

Thanks to Jonathan Strawn <jonnyflash@gmail.com>,
UNM ARTS Lab, for figuring this out.

maybe related to the Fink version of autoconf? (/sw/bin/autoconf)
Try creating with
   env PATH=/bin:/usr/bin:$PATH autoconf -f

Don't check for FL/mac.r any more -- use makeapp.sh.  That's the only
way supported in modern Mac OS X anyway.

It's not a fatal configure error if we can't find fltk.  User might just want to build libparti.

(time - highest_integer_time).  Use step 0 for times < 0,
and highest-available timestep when out of range > 0.

Remove extra 'extern "C"' gibberish.  Just depend on #ifdef __cplusplus to know whether to put it in.

newly compiled partiview executable.  Don't use Mac resources any more,
they're obsolete.

Re-orthogonalize the matrix in FlyNav::navUpdate().
Otherwise it'll blow up within a few dozen iterations.

New "-fly" config option, and "fly <buttonpov>,<buttoncenter>,<tspeed>,<rspeed>"
control command.  I don't know which button is which, so it's configurable on the fly.

rather than the CAVE origin.  Use position only -- keep the orientation
of the CAVE.

This still might not be right.  Might be better to put the
target point just behind the head, so an origin crosshair doesn't bug us.

And don't confuse near vs. far.

to "parti.NNNN.log" for ProcessID NNNN.

as listed in galaxyorbit.h.  New "-galaxy <first-galaxyorbit-coef>" option
enables it.  It compiles, but probably doesn't work yet.

parti.NNNN.log, where NNNN is the SZGClient ProcessID.
Maybe this too should be changed -- should we just try to log
to one big file, to limit clutter?  Every line is already labelled with [PID] now.

Don't use exit(0) in command parser.  Set a shared quit flag instead,
and have everybody (?) exit later.

<string>.append( C_string, len ) never copies in the string's trailing \000,
so add it ourselves.

Only create a listenSocket if we're the master -- avoid races to grab the port.
I'm not sure when is best to determine master-hood, so out of superstition
we do it in the first preExchange callback.

Yes, we *do* need to initialize cmdbuf_ to be *empty*.

szg gets mad if we try to select size 0.