My last patch to eigens was to make it barf whenever it received a non-symmetric matrix. That's because the ssl thing fails so frequently it isn't worth even attempting -- even though I put in code to post verify the eigenvalues and eigenvectors, and turn them to NaN's on failure.
I also updated docs to say that asymmetric matrixes are not supported for this release, in anticipation of that decision -- which Chris Marshal also just recommend.
One can argue that this is a minor improvement over
2.4.2, because at least we know that we can't handle non-symmetric matrices. I think that 2.4.2 didn't know it couldn't handle them, and produced wrong results? Not sure.
To answer your question, I am fine with leaving it as is, for
2.4.3 release (or whatever you'd like to call it). Personally, I only need eigens for symmetric matrices. Since Chris Marshal has brought up the topic several times, I am guessing he'd like to see asymmetric matrices solved, which would be nice. I am further guessing that you haven't seen a lot of requests for eigens of asymmetric matrices. I also think the gsl will be a major change, and there are sufficient changes for a
2.4.3 release. gsl can wait, imho.
Bill
On 7/18/06, Craig DeForest <deforest@xxxxxxxxxxxxxxxx
> wrote:Okay, I've marked 1497870, 1497875, and 1221742 as fixed.
I just fixed the constructor bug as well.
That leaves:
1465414 (lib64 version and lib paths)
1465864 (flexraw fails on Intel OS X with gfortran)
1505171 (failure in MatrixOps::eigens in CVS)
Bill -- for the point release do you want to do a full eigens fix, or
just patch it back to the symmetrizing code?
Cheers,
Craig
On Jul 18, 2006, at 1:07 AM, Sisyphus wrote:
>
> ----- Original Message -----
> From: "zowie" <
deforest@xxxxxxxxxxxxxxxx>
> To: "Bill Coffman" <bill.coffman@xxxxxxxxx>
> Cc: "perldl list" <
perldl@xxxxxxxxxxxxxxx>
> Sent: Tuesday, July 18, 2006 3:36 PM
> Subject: Re: [Perldl] next version
>
>
>> I'd like to see the "major" bugs on Sourceforge closed.
>>
>> The bugs that I'd like to see fixed before the next release are:
>> - 1221742 (failure of some PDL tests with perl-5.8.7)
>> - 1465414 (lib64 version and lib paths)
>> - 1465864 (flexraw fails on Intel OS X with gfortran)
>> - 1497870 (limits_keyspecs.t fails Test::More 0.62 [related to
>> 1221742?])
>> - 1497875 (limits_normalize_dsets.t faile Test::More 0.62 [related
>> to 1221742?])
>> - 1505171 (failure in MatrixOps::eigens in CVS)
>> - 1524150 (pdl constructor undefined behavior)
>>
>> Three of those (the test failures) are probably related.
>>
>
> Those three are fixed. Please close them.
> 1497870 and 1497875 are probably not related, but are certainly fixed.
> 1221742 is related to both in that it was the first reporting of the
> '1497870' and '1497875' bugs.
>
> Cheers,
> Rob
>
_______________________________________________
PDL-porters mailing list
PDL-porters@xxxxxxxxxxxxxxx
http://mailman.jach.hawaii.edu/mailman/listinfo/pdl-porters
--
Bill Coffman
~~~_\\\|/
~~~-(@ @)
=oOO=(_)==OOo===
_______________________________________________
Perldl mailing list
Perldl@xxxxxxxxxxxxxxx
http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
|
