[daip] VTESS problems under g77 3.2.2

[Eric Greisen]

rlaing at eso.org writes:

 > I recently attempted to use VTESS under a version of 31DEC04 downloaded and
 > compiled on 2004 Jan 16 using g77/gcc 3.2.2 under Linux RH9. This failed
 > consistently on a number of datasets, giving 0 and NaN for step and gradient,
 > and always generating a flat image. Reverting to a VTESS.EXE compiled using
 > an older g77 under RH6.3 gave sensible results [I'll send fuller details if they
 > are of any use to you].
 > 
 > Would you recommend returning to gcc 2.95.3 or is there a known work-around for
 > 3.2.2 (e.g. in a more recent incarnation of 31DEC04)?

I just rebuild my system with the RH9 build 3.2.2 compiler and ran a
VTESS with which I am familiar and it produced answers identical to
those of 3.2.3 and certainly not NaN's and such.

2.95.3 is always a solution so far but perhaps this needs more
debugging.

VTESS is a very touchy program and can go off into the weeds pretty
easily but usually not with NaN's.  Are you using a positive value of
FLUX - try a negative value.  Does it produce the Nan's immediately or
only after some iterations?  perhaps I need to see the inputs and
maybe even use one of your images...

Thanks,

Eric Greisen