Re: dMatrix.cc test failures

[Tatsuro MATSUOKA]

Hello

Sorry I have overlooked
*******************
I pushed a changeset.

        http://hg.savannah.gnu.org/hgweb/octave/rev/0413227e407b

Ben
******************
However the problem seem to be setting of tolerances.


--- On Wed, 2011/8/24, Tatsuro MATSUOKA wrote:

> Hello
> 
> Perhaps　the changes two new FAILS in MinGW build in the dMatrix.cc test.
> However the problem seem to be setting of tolerances.
> 
> Regards
> 
> Tatsuro
> 
> ***************************
> >>>>> processing 
> >>>>> e:\usr\Tatsu\mingwhome\octaves\hg\octave-work\liboctave\dMatrix.cc
>   ***** assert([M*cv,M*cv],M*[cv,cv],1e-14)
> !!!!! test failed
> assert ([M * cv, M * cv],M * [cv, cv],1e-14) expected
>    127.2762   127.2762
>    238.3372   238.3372
>     71.3559    71.3559
>   -107.4919  -107.4919
>    108.3659   108.3659
>     28.0889    28.0889
>    -42.4246   -42.4246
>      9.8570     9.8570
>    134.3780   134.3780
>    153.5574   153.5574
> but got
>    127.2762   127.2762
>    238.3372   238.3372
>     71.3559    71.3559
>   -107.4919  -107.4919
>    108.3659   108.3659
>     28.0889    28.0889
>    -42.4246   -42.4246
>      9.8570     9.8570
>    134.3780   134.3780
>    153.5574   153.5574
> maximum absolute error 2.84217e-014 exceeds tolerance 1e-014
> shared variables 
>   scalar structure containing the fields:
> <snip>
> 
>  ***** assert([rv*M;rv*M],[rv;rv]*M,3e-14)
> !!!!! test failed
> assert ([rv * M; rv * M],[rv; rv] * M,3e-14) expected
>  Columns 1 through 8:
> 
>    -35.212  -117.988   138.150   -66.489  -263.431   -66.692   -33.386  
> -144.804
>    -35.212  -117.988   138.150   -66.489  -263.431   -66.692   -33.386  
> -144.804
> 
>  Columns 9 and 10:
> 
>   -134.980   -86.433
>   -134.980   -86.433
> but got
>  Columns 1 through 8:
> 
>    -35.212  -117.988   138.150   -66.489  -263.431   -66.692   -33.386  
> -144.804
>    -35.212  -117.988   138.150   -66.489  -263.431   -66.692   -33.386  
> -144.804
> 
>  Columns 9 and 10:
> 
>   -134.980   -86.433
>   -134.980   -86.433
> maximum absolute error 5.68434e-014 exceeds tolerance 3e-014
> shared variables 
>   scalar structure containing the fields:
> <snip>
>  
> --- On Wed, 2011/8/24, Marco atzeri  wrote:
> 
> > changeset f217edac2c71 "fix dimension check for A'\B (bug #33997)"
> > introduces the "Mt, rvt2 but forgot to declared as shared
> > 
> > some other tests are failing (on cygwin) for tolerance
> > 
> > proposed change:
> > 
> > diff -r 7aa9cb7aaa32 liboctave/dMatrix.cc
> > --- a/liboctave/dMatrix.cc      Tue Aug 23 14:00:02 2011 -0500
> > +++ b/liboctave/dMatrix.cc      Wed Aug 24 00:02:58 2011 +0200
> > @@ -3111,7 +3111,7 @@
> >  */
> > 
> >  /* Test some simple identities
> > -%!shared M, cv, rv
> > +%!shared M, cv, rv, Mt, rvt
> >  %! M = randn(10,10)+100*eye(10,10);
> >  %! Mt = M';
> >  %! cv = randn(10,1);
> > @@ -3119,10 +3119,10 @@
> >  %! rv = randn(1,10);
> >  %! rvt = rv';
> >  %!assert([M*cv,M*cv],M*[cv,cv],1e-14)
> > -%!assert([M'*cv,M'*cv],M'*[cv,cv],1e-14)
> > -%!assert([rv*M;rv*M],[rv;rv]*M,1e-14)
> > -%!assert([rv*M';rv*M'],[rv;rv]*M',1e-14)
> > -%!assert(2*rv*cv,[rv,rv]*[cv;cv],1e-14)
> > +%!assert([M'*cv,M'*cv],M'*[cv,cv],3e-14)
> > +%!assert([rv*M;rv*M],[rv;rv]*M,3e-14)
> > +%!assert([rv*M';rv*M'],[rv;rv]*M',3e-14)
> > +%!assert(2*rv*cv,[rv,rv]*[cv;cv],3e-14)
> >  %!assert(M'\cv,Mt\cv,1e-14)
> >  %!assert(M'\rv',Mt\rvt,1e-14)
> >  */
> >
>