Re: [eigen] Error when using tridiagonalization_inplace_selector

[Christoph Hertzberg <chtz@xxxxxxxxxxxxxxxxxxxxxxxx>]

You should probably change the type of the solver to either

    Eigen::SelfAdjointEigenSolver<Eigen::Matrix<double,9,9> >
or
    Eigen::SelfAdjointEigenSolver<Eigen::MatrixXd>

Not sure, if we really want to support `Eigen::Matrix<double, -1, 9>` 
here. It would of course be possible, as this implicitly implies that 
the number of rows is 9 as well (otherwise, there is no Eigenvalue 
decomposition).

Cheers,
Christoph

(sorry for the empty response before ...)

On 08/09/2021 16.17, Alberto Luaces wrote:
> Yes, sorry, that's exactly the error I mean.
>
> It seems that previously it was ok to build the solver from AᵀA, but not anymore.
>
> Thanks for deploying it in godbolt!
>
> On Wed, Sep 8, 2021, at 15:11, Jens Wehner wrote:
> > Hi,
> >
> > what is the build error and which system/compiler/options?
> >
> > Is the build error like this?
> > https://godbolt.org/z/KGTejxMdq
>   <https://godbolt.org/z/KGTejxMdq>
> > Compiler Explorer - C++ (x86-64 gcc (trunk)) <https://godbolt.org/z/KGTejxMdq>
> > int main(){ using MatX9 = Eigen::Matrix<double, -1, 9>; MatX9 epipolar_constraint(7, 9); epipolar_constraint.fill(0.0); // EncodeEpipolarEquation(x1, x2, &epipolar_constraint); Eigen::SelfAdjointEigenSolver<MatX9> solver (epipolar_constraint.transpose() * epipolar_constraint); return 0; }
> > godbolt.org
> >
> >
> > Escience Center Logo
> >
> > *Jens Wehner*
> >
> > Research Software Engineer
> >
> > +31 6 438 666 87
> >
> > Netherlands eScience Center
> >
> > *www.esciencecenter.nl*
> >
> >
> >
> > *From:* Alberto Luaces <eigen@xxxxxxxxxxxxxxxxxxxx>
> > *Sent:* Wednesday, September 8, 2021 14:12
> > *To:* eigen@xxxxxxxxxxxxxxxxxxx <eigen@xxxxxxxxxxxxxxxxxxx>
> > *Subject:* [eigen] Error when using tridiagonalization_inplace_selector
> >   
> > Hi, I have found a build error in openmvg when using 3.4.0, shown with this snippet.  Do you think it is a bug in Eigen?
> >
> > Thanks!
> >
> > #include <Eigen/Dense>
> >
> > int main(){
> >      using MatX9 = Eigen::Matrix<double, -1, 9>;
> >      MatX9 epipolar_constraint(7, 9);
> >      epipolar_constraint.fill(0.0);
> >      // EncodeEpipolarEquation(x1, x2, &epipolar_constraint);
> >      Eigen::SelfAdjointEigenSolver<MatX9> solver
> >          (epipolar_constraint.transpose() * epipolar_constraint);
> >      return 0;
> > }
> >
> >
> > --
> >    Alberto
>
> --
>    Alberto