[scilab-Users] Inexplicable result with scilab 5.0.2 and matrix multiplication

[Sylvestre Ledru]

It is a bug of gfortran 4.3 which didn't exist in the version 4.2.
I submitted a bug report:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37863
which has been very quickly confirmed as being a bug.

Sylvestre

Le mercredi 15 octobre 2008 à 13:46 +0200, Sylvestre Ledru a écrit :
> For those who are interested in this bug, the discussion continues in
> the bugtracker.
> (You can add your email address to the bug report to receive updates).
> 
> Sylvestre
> 
> Le lundi 13 octobre 2008 à 23:48 +0200, François Vogel a écrit : 
> > Looks like this one:
> > 
> > http://bugzilla.scilab.org/show_bug.cgi?id=2647
> > 
> > Francois
> > 
> > 
> > Sylvestre Ledru said on 13/10/2008 23:40:
> > > Hello,
> > > 
> > > It is most probably a bug in the printing functions... I can reproduce
> > > it with the binary (ie not compiled on my computer) but I can't on my
> > > dev version (compiled with gfortran 4.3.1)
> > > Once, we had a simular bug caused by a bug in gfortran. Maybe it is a
> > > regression.
> > > 
> > > Thanks for pointing this out.
> > > Sylvestre
> > > 
> > > Le lundi 13 octobre 2008 à 23:32 +0200, Christoph Dittmann a écrit :
> > >> Hi,
> > >>
> > >> I've got a problem with scilab 5.0.2. It seems to print certain values 
> > >> wrong. The following example shows the problem:
> > >> -->M = [1,2;3,2]; C = M^(-1)*M
> > >>   C  =
> > >>
> > >>      2.           1.110D-16
> > >>      1.110D-16    2.
> > >>
> > >> This is what I get with scilab 5.0.2 on Ubuntu 8.04 (x86) using the 
> > >> precompiled version from http://www.scilab.org/download/ . My problem 
> > >> are the 2's on the diagonal since I've no idea where they come from.
> > >>
> > >> With scilab 4.1.2, the one from the Ubuntu 8.04 repository, I get what I 
> > >> expect:
> > >> -->M = [1,2;3,2]; C = M^(-1)*M
> > >>   C  =
> > >>
> > >>      1.           1.110D-16
> > >>      5.551D-17    1.
> > >>
> > >>
> > >> Some further calculations with scilab 5.0.2 show that the problem seems 
> > >> to lie in the printing of the result and not its actual numerical value:
> > >> -->x = C(1,1)
> > >>   x  =
> > >>
> > >>      2.
> > >> -->x * 2
> > >>   ans =
> > >>
> > >>     2.
> > >> -->x + 0.01
> > >>   ans  =
> > >>
> > >>      1.01
> > >>
> > >>
> > >> Could that be a bug or did I just miss something crucial?
> > >>
> > >>
> > >> Thanks,
> > >> Christoph
> > > 
> > > 
> > > cd 
>