pin pyddx #2921
pin pyddx #2921
Conversation
There was a problem hiding this comment.
Thanks for this PR! DDX issues seem to have been the root for quite a few CI failures lately.
A quick thought on my end, since we're enforcing pyddx 0.3.0 here for now. In the ddx/CMakeLists.txt file in the external/upstream/ directory, there is a line for finding the pyddx Python module:
find_python_module(pyddx ATLEAST 0.3.0 QUIET)
Would it be necessary, or even desirable, to temporarily switch this to
find_python_module(pyddx EXACT 0.3.0 QUIET)
to simply not allow pyddx v0.4 to be used with Psi4 at the moment?
|
I'm a bit surprised this is needed as we did not introduce python-side changes in 0.4. Maybe there is a bug we introduced in 0.4? Can you point me to what is failing? On top I should add we follow semantic versioning, so all 0.3.x should be compatible. I don't know how conda / cmake deal with this, but in this case one should target the range 0.3.0 to less than 0.4.0 |
Sure, here's one: https://github.com/psi4/psi4/actions/runs/4735120869/jobs/8404940333#step:23:737
gtk, I can add |
|
So what's the current status of this PR? pyddx seems to be pinned to 0.3.0 from work in other PRs, although there seems to be discussion of pinning pyddx to 0.3.* instead. |
|
I haven't heard anything over at #2940, so I think a "hard pin" is in order for the release. That is, not just prevent CI from using the new 0.4 but update the external/ddx/CMakeLists.txt to avert users grabbing it. I'll fix this PR up. |
|
I've filed ddsolvation/ddX#142 over at the parent project. Further investigation makes me uncertain whether our test reference may just have a built-in bug, in which case I hate to pin to a worse version. |
Description
pyddx is in rapid development, so I think psi4 needs a PR to catch up to v0.4
Checklist
Status