Jonathon Misiewicz

That agrees with my knowledge of `libdiis` history: I think it was made in the late 2000's for `dct` purposes. Noted about the poorly conditioned equations. That possibility didn't even...

You can bypass this by changing the bra and ket of the `mints.ao_eri` call so that the two `orbital_basis` arguments are last. I'm not sure _why_ this makes any difference.

@andysim would probably have an idea as to why this is happening. I suspect this is related to us switching to Libint 2.

Thanks for the report. I've been able to reproduce the crash. (I observe a fifth decimal place disagreement in DF-RKS SAD energy and a third decimal place disagreement in all...

Good call. The geometry is about a factor of 2 too compressed. (Treating the geometry as bohr rather than angstroms makes things _worse_.) @averyparr Where did this geometry come from?

To sum up the discussion so far: * The original geometry was malformed. Not a Psi4 problem. * We've noticed large changes in energy in the early iterations, with the...

> @hokru For ωB97M-V, it doesn't seem to be possible to change the ω parameter directly from the psithon interface—there's a check at > > https://github.com/psi4/psi4/blob/d9093c75c71c2b33fbe86f32b25d138675ac22eb/psi4/src/psi4/libfunctional/LibXCfunctional.cc#L218 > that disallows it....

Good catch, and thanks for the response. Requiring `solver_n_guess` to be at least the computed `nirrep` sounds good to me. Should I edit the original issue, or leave it as-is?

i.e., confirm https://github.com/psi4/psi4/issues/2476 and https://github.com/psi4/psi4/issues/2594 are actually closed and then add as test cases

2594 is not an option for adding to the test suite, as the test takes too long tor un. 2476 _may_ be. Per @loriab, that depends on increasing the AM...