Ubuntu
octave3.0 package

Octave 2.9 left division gives incorrect results

Bug #132430 reported by Jacek C. Wojdel
2
Affects Status Importance Assigned to Milestone
octave3.0 (Ubuntu)
Fix Released
Medium
Unassigned

Bug Description

Binary package hint: octave2.9

In specific cases, the left division in octave 2.9 seems to ignore the below diagonal elements e,g:
octave2.9:1> y=[ 1; 2 ]
y =

   1
   2

octave2.9:2> A=[1 0; 0 1]
A =

   1 0
   0 1

octave2.9:3> A\y
ans =

   1
   2 <-- correct

octave2.9:4> A(2,1)=0.5;
octave2.9:5> A\y
ans =

   1
   2 <-- incorrect!

octave2.9:10> A
A =

   1.00000 0.00000
   0.50000 1.00000

Note that if A is initialised directly as A=[1 0;0.5 1], the bug does not manifest itself.
It seems as if some strange over-optimisation takes place and assumes A is upper triangular no matter what we do to it later (if the changed element is above diagonal, the results are correct).
Tested on up to date feisty

Revision history for this message
Stephan Rügamer (sruegamer) wrote : Please check if this bug is still valid for octave3.0

Dear Reporter,

please check if your bugreport is still valid for octave3.0.

If yes, please update your package to octave3.0.

Rational:

Octave3.0 is the latest stable release of Octave. All former versions
like octave2.1 or octave2.9 were development releases, which are
obsolete now (for Hardy that is).

As your bugs are not worth a StableReleaseUpdate for the still
maintained Ubuntu release, I would like to ask you to check out
Octave3.0 and eventually, when you bug is fixed, close your bugreport
or update your report to match source package "octave3.0".

Thank you for shaping the future of Ubuntu,

\sh

--
SysAdmin, OSS Developer
GPG-Key ID: 0xC098EFA8
Fingerprint: 3D8B 5138 0852 DA7A B83F DCCB C189 E733 C098 EFA8
http://www.sourcecode.de/

William Grant (wgrant)
Changed in octave2.9:
importance: Undecided → Medium
status: New → Incomplete
Revision history for this message
Jacek C. Wojdel (j-c-wojdel) wrote : Re: [Bug 132430] Please check if this bug is still valid for octave3.0

On 1/29/08, Stephan Hermann <email address hidden> wrote:
> Dear Reporter,
>
> please check if your bugreport is still valid for octave3.0.
>
> If yes, please update your package to octave3.0.

I'm sorry, but I cannot afford myself upgrading to Herdy on my
production machine at the moment, so I cannot check it against
Oct-3.0.
I tried to reproduce the bug on the current version of 2.9 (which was
published 2 days after my initial report), and it seems to be gone.
I'm pretty sure the bug can be closed.

--
Jacek C. Wojdel <email address hidden>
Evolution can go to hell as far as I am concerned.
What a mistake we are.

Revision history for this message
Peter McKenna (peter-mckenna) wrote :

I,ve tried this in Hardy Octave 3.0
octave-3.0.0:1> y=[1;2]
y =

   1
   2

octave-3.0.0:2> A=[1 0;0 1]
A =

   1 0
   0 1

octave-3.0.0:3> A\y
ans =

   1
   2

octave-3.0.0:4> A(2,1)=0.5
A =

   1.00000 0.00000
   0.50000 1.00000

octave-3.0.0:5> A\y
ans =

   1.0000
   1.5000

It appeard this bug has been fixed.

William Grant (wgrant)
Changed in octave3.0:
status: Incomplete → Fix Released
Revision history for this message
Jacek C. Wojdel (j-c-wojdel) wrote : Re: [Bug 132430] Re: Octave 2.9 left division gives incorrect results

I confirm that the bug has been fixed for sometime now.

As I wrote earlier to Stephan Hermann, the bug was gone in two days
after the report, only I did not notice (switched all my calculations
to 2.1 in the meantime).

On 1/29/08, Stephan Hermann <email address hidden> wrote:
> Dear Reporter,
>
> please check if your bugreport is still valid for octave3.0.
>
> If yes, please update your package to octave3.0.

I'm sorry, but I cannot afford myself upgrading to Herdy on my
production machine at the moment, so I cannot check it against
Oct-3.0.
I tried to reproduce the bug on the current version of 2.9 (which was
published 2 days after my initial report), and it seems to be gone.
I'm pretty sure the bug can be closed.

--
Jacek C. Wojdel <email address hidden>
Evolution can go to hell as far as I am concerned.
What a mistake we are.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.
Subscribing...

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.