MT JW Bespokefit deployment CI failed because the fetch-tags keyword in the actions/checkout action is a liar and didn’t actually fetch the most recent tag. If other deployment CI starts failing this is almost certainly the cause. MT – We don’t actually use fetch-tags and this rabbit hole likely goes deeper. This does coincide with a new release of git itself, but the git mailing lost doesn’t show any smoking guns or similar bug reports. (general) – We don’t know why this broke or why JW’s changes fixed it. We’ll leave this be unless it pops up again.
Thanks for the masterful issue discussion and technical leadership while I was out (I mean, not just then, also while I’m around, but in particular while I was out) Charger arrived? To benchmark our protein FF using OpenFE, we’ll need to get SMIRNOFF proteins running in their infrastructure. This will likely have two stages An initial protocol in pontibus (It’s not just for SFEs, it’s for everything experimental) A full systemgenerator replacement I think this will not be much coding work, but will require a lot of speccing since OpenFE wants to put it in their core code once and have the interface/behavior be basically final. I’m currently on the hook to find a time to start scoping this with Irfan, I may loop you in if timing permits.
MT – For the systemgenerator replacement, will there be a “if everything is SMIRNOFF then use the OpenFF Toolkit/Interchange for everything” condition or flag here?
|