Evgeni Burovski

Wow. So the trick is to add "bare" prototypes, in the form of `void dpotrf_(char* uplo, int* n, double* a, int* lda, int* info);` etc. I honestly don't understand how...

So the error here is that setting `parametrization` works on a regular instance and fails on a frozen instance, correct? FWIW, I'm not sure about the fix here. While it...

Confirmed on linux, with an older scipy. So the issue is there "forever". Hmm. Any chance you are set up to step it in gdb to see where the nans...

Took a quick look to see if it's not fixed by gh-17212. Of course it is not. In the OP, nans appear in the spline coefficients which are computed here:...

Thanks Ralf ! I think this would be nice to add to some building from source docs. Not necessarily the Fortran info, may be similar info for more popular languages...

A quick note so far: nans start appearing around this line https://github.com/scipy/scipy/blob/v1.3.0/scipy/interpolate/fitpack/fpgrre.f#L262 when `nx == nxe`, the maximum allowed number of knots in the x direction. https://github.com/scipy/scipy/compare/main...ev-br:scipy:rectbivspline_nans?expand=1 here's a branch...

There was no recent changes in or around RectBivariateSpline. git blame concurs: https://github.com/scipy/scipy/blame/v1.14.1/scipy/interpolate/_fitpack2.py#L1526-L1611

Well, the next question is of course why there's a zero. `fpback` is a backsubstitution solve of a linear system with the triangular l.h.s. The triangular matrix itself comes from...

Checking locally with the patch from https://github.com/scipy/scipy/issues/17787#issuecomment-2565026701, the OP example fails with a legible ValueError, so it's an improvement: ``` ValueError: The required storage space exceeds the available storage space:...

Yes, s=0 is interpolation (spline(x) == y). Commenting out the goto is not the right fix, unfortunately. For zero s, the number of knots is basically equal to the number...