1. When running on rs/6000 or DecStation, problem with asserts.  I think
   it is the load.c file, maketree is the routine.  Lorenz thinks it is
    an assert in expandbox.  Anyway, you have to have the asserts in
    the code (comment out #define of NDEBUG) to make code run.

2. Here's another from Lorenz:
I ran across the following: xrand(0,0.999) may return 0.0.  This might 
cause the line 'r = 1 / sqrt (pow(xrand(0.0, MFRAC), -2.0/3.0) - 1);'
in testdata to fail (sqrt of a negative number).  

I imagine this is straight out of the original.  Perhaps it can be 
safely changed to xrand(0.00000001,MFRAC) ?
