Jakob Degen

After #733 this reproduces as ```rust #[test] fn chalk_688() { test! { program { #[lang(fn_once)] trait FnOnce { type Output; } struct Ordering {} closure foo(self,) {} impl FnOnce for...

> @JakobDegen just now finding some time to look at this and #733. Sorry for the delay. Shouldn't the repro for this end up with a hang, instead of an...

> Ah okay. For future reference, you can use `} yields[SolverChoice::recursive_default()] {` to restrict to only the recursive solver! That's good to know, thanks! > Are you up for trying...

@vringar if you're no longer working on this, can I give it a go?

The one thing I'd note about the perf report is that (if I'm reading things correctly) the size of compiled binaries seems to have increased. That seems concerning, because it...

I agree that this opt should be moved to codegen, for the following reasons: 1. We have another set of MIR opts, the simplify try related stuff (especially `SimplifyArmIdentity`) that...

@wesleywiser yeah, I don't mind merging it either - it doesn't really affect things either way. > I will note that MIR optimizations in general all suffer from the third...

Thanks for the suggestions! @rustbot ready

> At some point it might be worth adding a validation of cached information behind `-Zvalidate-mir`. Yes, definitely. At some point I want to add stateful analyses in the pass...

@bors r=tmiasko rollup=never Potentially perf sensitive