Re: Different numerical results between sparse and full matrices

[John W. Eaton]

On 17-Feb-2012, Michael Goffioul wrote:

| On Fri, Feb 17, 2012 at 1:02 PM, Jordi Gutiérrez Hermoso
| <address@hidden> wrote:
| > On 17 February 2012 07:41, Michael Goffioul <address@hidden> wrote:
| >> octave.exe:28> sparse(x)./sparse(0)
| >> ans = Compressed Column Sparse (rows = 1, cols = 1, nnz = 1 [100%])
| > [snip]
| >> In other words, complex division by 0 in sparse matrices gives NaN
| >> instead of Inf. Does anybody know where I should look at to find where
| >> this could come from?
| >
| > They're in liboctave/Sparse-op-defs.h... which seems to be one giant,
| > endless macro. I'm now itching to rewrite it without macros, if you
| > don't beat me to it.
| 
| No, please, be my guest :)
| 
| In the meantime, I found out that the test is marked with "xtest",
| which makes is a "known failure". I traced the change back to
| http://hg.savannah.gnu.org/hgweb/octave/rev/3e4dab6a136c, but that
| does not give more information about why the change was made.

This change was discussed in the thread that starts here:

  https://mailman.cae.wisc.edu/pipermail/bug-octave/2007-October/010077.html

jwe