Re: Interpreting Results of Make Check

[Daniel Sebald]

Hello,

I don't receive the octave-help list, so sorry I didn't see this until now.


> This looks like one for Dan - aren't these supposed to be
> coming out integer?  That's why you took off the eps arg,
> right?

Correct.  I'm testing the example here and get integer.  All tests pass.

Let's first confirm that Alan has patched the file correctly.  Alan, near
the end of the routine after fftfilt.m is patch there should be:

  ## Transpose after cleanup, otherwise rounding fails.
  if (transpose)
    y = y.';
  endif

Do you see that?  If not, the first hunk of code didn't patch while the
second hunk might have.  (You should find a reject file somewhere in that
case.)

The rounding routine should not be influenced by the FFT library.  The tests
for rounding are based only on properties of b and x.

There is an error in the code I see, but it doesn't influence any of the
results.  In the patch is a small hunk of unaltered code: 

   if (! any (b - fix (b)))
     idx = !any (x - fix (x));
     y(:, idx) = round (y(:, idx));
   endif

I believe the line

    idx = !any (x - fix (x));

should read

    idx = find (!any (x - fix (x)));

because "any" returns a logical scalar, which doesn't make such a good
index.  I will fix that and attach a new patch to the bug report.

Dan




--
View this message in context: 
http://octave.1599824.n4.nabble.com/Interpreting-Results-of-Make-Check-tp4645175p4645349.html
Sent from the Octave - General mailing list archive at Nabble.com.