Mon 08 Aug 2016 01:36:03 AM UTC, comment #12:
Verified fix, closing report.
|
Fri 05 Aug 2016 06:35:31 PM UTC, comment #11:
I checked in the changeset from file #36374 here:
http://hg.savannah.gnu.org/hgweb/octave/rev/21684fa513ce
Can this report be closed?
|
Wed 06 Jul 2016 07:48:21 AM UTC, comment #10:
Hi, I would like to help here, but which patch(es) should I apply to development? file #36374 applies smoothly, but afterwards I cannot apply file #36373 from bug #46912.
|
Mon 27 Jun 2016 06:22:31 AM UTC, comment #9:
I've increased the importance of this bug to reflect that it silently produces completely bogus results -- a class of bugs that I consider fatal to a language.
I hope massimiliano's patch is one of those to make it into 4.2.
|
Thu 18 Feb 2016 05:05:43 AM UTC, comment #8:
Here's a new patch to reflect jwe's recent changes.
(file #36374)
|
Sat 23 Jan 2016 05:40:52 AM UTC, comment #7:
On closer inspection, there are a couple of minor issues with Massimiliano's patch. (It doesn't support qr(A,B,0) for full A, and the variable b_mat is redundant given have_b.) Both issues are fixed by the changeset at bug #46912.
|
Sat 23 Jan 2016 03:07:19 AM UTC, comment #6:
Massimiliano's patch looks good to me. I've updated it to reflect recent changes to the code base.
There are still some issues the the qr code; I'll submit another patch at bug #46912.
(file #36135)
|
Mon 18 Jan 2016 06:30:33 PM UTC, comment #5:
Your patch has been noticed, we just have few patch review resources in Octave.
|
Fri 12 Sep 2014 02:00:44 PM UTC, comment #4:
I'm starting to think that my patch went unnoticed. That is why I'm pinging, as I was suggested to do on IRC.
Any feedback would be much appreciated.
|
Sun 09 Mar 2014 04:11:13 PM UTC, comment #3:
I have submitted a patch that should fix the bug, with relative tests.
(file #30836)
|
Thu 13 Feb 2014 05:50:42 PM UTC, comment #2:
The expected behavior, as per both the Octave and Matlab documentation, is to return Q'*B if B is a matrix, regardless of whether A is sparse.
|
Thu 13 Feb 2014 05:42:40 PM UTC, comment #1:
Confirmed in version 3.8.0, although I am not sure what the expected behavior is here. The code for the qr function only uses the B argument if A is sparse. If A is not sparse, the second argument serves the same purpose as the documented '0' argument regardless of its value. Does the behavior or the documentation need fixing?
|
Wed 12 Feb 2014 01:45:41 PM UTC, original submission:
if, for example,
A = randn(3, 3); B = randn(3, 2);
[Q, R] = qr (A, 0);
works as expected, returning a QR decomposition with QR = A.
According to the documentation,
[C, R] = qr (A, B);
should return C = Q'*B -- but it actually just returns C = Q, which is quite different.
|