One test in data.cc crashes Octave 3.1.51

[Jaroslav Hajek]

On Mon, Aug 11, 2008 at 10:54 PM, Thomas Treichl <Thomas.Treichl at gmx.net> wrote:
> Hi,
>
> I have still some more problems after compiling 3.1.51 for PPC (but overall just
> 7 test failures). One special test in data.cc crashes Octave, I have found out
> the source where it appears (data.cc, line 1064):
>
>   %!assert(log2(complex(0,Inf)), Inf + log2(i));
>
> I split the commands and found out more:
>
>   octave:1> A = Inf
>   A = Inf
>   octave:2> B = log2(i)
>   B =  0.00000 + 2.26618i
>   octave:3> A+B
>   panic: Segmentation fault -- stopping myself...
>   ans = Inf + attempting to save variables to `octave-core'...
>   save to `octave-core' complete
>   Segmentation fault
>

I can't reproduce this crash. Can you share your config.log, or just
tell me whether you have log2 in libm?


> And I don't have any idea where to go from here looking for this bug. My second
> test was:
>
>   octave:1> complex(0,Inf)
>   ans = 0 + Infi
>   octave:2> log2(complex(0,Inf))

again, works fine here: Inf + 2.266i. It almost looks as if you have
still the old log2 around!
What does "which log2" say?


>   ans = NaN - NaNi

Octave can use -NaN instead of NaN. I sent a simple fix some time ago,
but apparently it was never applied, even though John stamped it "a
reasonable thing to do" later in the same thread.
See <http://www.nabble.com/Re%3A-automatic-Inf%2BNaN*i-conversion-p16895792.html>


-- 
RNDr. Jaroslav Hajek
computing expert
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz