Ilhan Polat

Are there basic but nontrivial problems for the testing that won't take too much time? Alternatively, we can create our own simple test suite by just taking a subset of...

Yes of course, I'll circle back to logging in the original source. We probably leave this merge to 1.14 though just to go through things a bit more thoroughly.

If there is such a bound then it should be in the wrappers as fortran code does not check it This is the old wrapper https://github.com/scipy/scipy/blob/1d3a067c2ccd0a6efddeb3194163aa9a3879d26e/scipy/special/cdf_wrappers.c#L243-L252 This is the fortran...

Yeah, this bound should only appear on `ncfdtrinc` as an algorithmic shortcoming but it doesn't https://docs.scipy.org/doc/scipy/reference/generated/scipy.special.ncfdtrinc.html#scipy.special.ncfdtrinc

Yes this is a very funky way of wrapping things for me too. So typically what I do is first get the function name to be tested, from the test...

OK in #20069 we tried ```python def _normal_cdf(sigma, x, mu): x = mpmath.mp.mpf(x) mu = mpmath.mp.mpf(mu) sigma = mpmath.mp.mpf(sigma) q = mpmath.mp.ncdf(x, mu, sigma) return q [snipped] class def test_nrdtrimn(self):...

I managed to make this work but the main issue is the closed form mpmath or some other credible source to compare for the other ones. For now, I'm just...

Sigh, if the fortran torture finishes, I'll rewrite this. It is always causing some minor issues. For now we can use a bit more stringent condition for the real squareroot...

So what is happening here in these lines https://github.com/scipy/scipy/blob/715913255756af431786ad587392650bdb6f11ce/scipy/linalg/_matfuncs_sqrtm.py#L173-L178 is that, if the Schur decomposition returns 2x2 blocks on the diagonal, that means the matrix is going to have a...

Not to mention, we lost Nick Higham last month, I was hoping to visit him this year during a business trip. Just a tiny remembrance for all of us and...