Why does refscf disagree between mints2-bse and mints2?

I think Ben mentioned to me that it's because of revision of the basis set, and if a :0 (equals old PNNL BSE) version is used they match exactly.

The output file included in the PR uses :0 but is still showing disagreement with mints2. (The input and output files have different reference energies, which is more confusing.)

The "base" basis changed, too, from DZ to 6-31g. Only one atom left with it, if I'm reading it right.

mints2 (usual)

   # First, assign DZ to all atoms
   assign DZ
   # Now, assign 3-21G to all carbon atoms
   assign C my3-21G
   # The two atoms labelled H1 get a STO-3G basis two
   assign H1 sto-3g
   # Carbons 3 and 5 get a STO-3G basis, too
   assign C1 sto-3g

mints2-bse

   # First, assign DZ to all atoms
   assign bse:6-31g
   # Now, assign 3-21G from bse library all carbon atoms (latest version)
   assign C bse:3-21g
   # The two atoms labelled H1 get a STO-3G basis from psi4 library
   assign H1 sto-3g
   # Carbons 3 and 5 get a STO-3G basis from BSE, but specify v0
   assign C1 bse:sto-3g:0

Thanks. In that case, can we change the comment at the start of the file to mention that? As written, I thought the two energies were supposed to be identical.

Thanks. In that case, can we change the comment at the start of the file to mention that? As written, I thought the two energies were supposed to be identical.

Sounds good. This is why I like parameterized pytest for seeing the input differences.

@bennybp, did I get the why-different reasons right above? And if you're changing that comment, maybe regenerate the ref file w/o test fail:

https://github.com/psi4/psi4/pull/2905/files#diff-667f7008a6b3ab105b3130447305716990d91451c161ed7eaf1f04886bd32a00R361

Well that was sloppy of me. Thanks for catching that

I'm letting @loriab merge, since she's more familiar with MolSSI interfaces.