Noticed in 2025-03-01 (which at time of writing isn't on godbolt yet) that the MIR for <= for a generated type still contains

 bb1: { _4 = move ((_3 as Some).0: std::cmp::Ordering); StorageLive(_5); _5 = (copy _4,); _0 = <fn(std::cmp::Ordering) -> bool {std::cmp::Ordering::is_le} as FnOnce<(std::cmp::Ordering,)>>::call_once(std::cmp::Ordering::is_le, move _5) -> [return: bb2, unwind continue]; }

https://play.rust-lang.org/?version=nightly&mode=release&edition=2024&gist=57c2829c1cf0a8778926bd87a0378feb

I can workaround that particular case, but calls to trivial functions like is_le really ought to MIR-inline, even if done through the voldemort function item like this.