lcnr
lcnr
@bors try @rust-timer queue
see the PR description and crater triage https://github.com/rust-lang/rust/pull/129021#issuecomment-2297702647. This fixes (the non-higher ranked) cases of our most famous unsoundness without causing any unintended regressions😊 @rfcbot merge
> I tried this at first, but it resulted in too many errors. Can you share an example snippet which failed? I would assume that we still constrain the original...
this is surprising to me :thinking: while the expected type in `check_expr_array` starts out as an infer var, shouldn't we replace it with a new one at the start of...
hmm, I don't feel like I've got a good grasp of the way this works rn (and it's definitely not in cache for me rn) I do think treating array...
ah wait, can you actually try to perfectly match the if/match behavior :thinking: it feels weird to weaken inference without going all the way here. We can crater and look...
@rfcbot concern try going all the way, no expectation
as a general question: should we mention the decls at all in the proof trace? I am not that happy about using `..` for empty fields but also don't like...
> @lcnr the decls should only have to be output once in the trace of a proof, because they shouldn't change. So I guess we could just print them in...