[daip] FRING Problem

[Amy Mioduszewski]

Hi Thomas,

So the problem seems to be that the HP compiler is completely
out to lunch.  O.K., maybe I am overstating things.  This has
just been frustrating.  The reason FRING fails on your HP and
not on Linux or Sun boxes is that the HP compiler cannot add
2.315E-5 to another number.  FRING pads the end scan time
by 2 seconds (2.315E-5 days) and on the HP 2.315E-5 is taken
for 0 and no padding occurs.  Eventually, after much trial
and error, I did get it to do the padding, and FRING ran to
completion.  So this really is the problem.  Unfortunately,
the thing that got it to work makes no sense, i.e., I just added
a couple lines of code to test various methods and the original
line suddenly started to do the right thing.  It is not like
2.315E-5 is at the edge for a real number and it is being added
to numbers like 3.03239e-3, so this should be a stretch for a
FORTRAN compiler.

Eric is gone and he has more expertise with compilers than I,
maybe there is an option we can use on the HP FORTRAN compiler
that will make this work correctly.  E.g., AIPS uses option
+Oaggressive when it compiles, maybe if we tone that down it
will work.  When I compile the program with no optimization it
works.

Because of this I would not really trust any program in AIPS
compiled with this compiler, so I would recommend not using the
AIPS you have installed on your HP boxes.

Cheers,

Amy