Re: wierd error

[John W. Eaton]

On 12-Jun-1998, Heber Farnsworth <address@hidden> wrote:

| I just upgraded to 2.0.13 and I get frequent 
| floating point exceptions.  I can get one 
| anytime I issue the following three commands 
| just after startup
| 
| Octave, version 2.0.13 (alphaev56-dec-osf4.0d). 
| Copyright (C) 1996, 1997, 1998 John W. Eaton.
| This is free software with ABSOLUTELY NO WARRANTY. 
| For details, type `warranty'.
|      
| octave:1> multiplot(2,2)
| octave:2> x = randn(100,1);
| octave:3> y = randn(100,1);
| error: floating point exception -- trying to return to prompt
|      
| octave:3> 
|      
| Changing arguments to randn doesn't change anything.  This is not
| the only time I get such exceptions but this is the only way I can
| get them predictably.
|      
| I suspect this is only a problem with DEC machines but can anyone
| verify this?

I don't see this problem on the Alpha systems I have.  What
compiler(s) and options did you use to compile Octave?

On 15-Jun-1998, PETER HOPFGARTNER <address@hidden> wrote:

| If you use Digital's Fortran, add -fpe3 to the Fortan compilation flags 
| (FFLAGS), for GCC, G++ and  G77 
| -mieee-with-inexact should be already set.

If you use Digital Fortran, Octave's configure script should already
set FFLAGS to include -fpe1, which I think is OK (the only difference
in -fpe1 and -fpe3 that the man page mentions is that -fpe1 sets
underflow values to zero, and -fpe3 allows `gradual underflow').

If you still have trouble, please submit a complete bug report to
address@hidden

Thanks,

jwe