The main issue is that we changed the defaults of atol from None to 0., hence when that code hits most of these will still complain about a different warning. That's why all that if solver this do that frenzy

Looks much better but still won't fix the actual problem there.

Looks much better but still won't fix the actual problem there.

I'm not claiming to solve any problem here, it's explicitly meant to be NFC (non-functional change). It's meant to reduce the necessity for test changes in #18391

In the spirit of "if you care strongly, I'll concede", I'm happy to back out 31fd623 if anyone objects. It's based on a single change that I had picked up from #18391 and applied to the whole file.

This is much better than what I did in the PR anyways because I basically got tired by looking at failing tests for long time. So +1 to merge from me.

Glad to hear it @ilayn.

@tupui, do you want to review (not required, just an offer)? I'm still still planning to remove self.Poisson1D & 2D from the class, which I overlooked previously, but that's only two lines. I'll do it once I'm back at a computer in about an hour

@tupui, not sure if your 👍 means you want to review or you're happy as is.

It means I m happy as is 😃