I think we still have issues with closures somehow, maybe fallout from the Universes PR. In particular @Matthew Jasper I was hoping #60449 would have fixed #58840 as a duplicate of #57464 since its error seemed so similar to your minimization (playground still ICEing on stable but not nightly) but the non-minimized repro project still ICEs on the latest nightly unfortunately. I'll try and minimize it if no one gets to that first.
I've pushed a version which is a bit more reduced but still far from minimal
I feel this error is still another strand in the ball of "unexpected region in query response" errors
this specific piece of code started ICEing in January,
nightly-2019-01-04, but back then it was the older ICE in
(IIRC this is the first nightly with Universes)
I dout it'll be especially useful to bisect for the new location of the ICE for this code ?
(in case it is helpful, it's
nightly-2019-02-26 and I think it's probably the delay_span_bug from #58649 moving the ICE from the canonicalizer to universal_regions)
funny that all these happened with
I'll try and keep reducing the code more and more, with juniper + hyper + futures + tokio it's still a bit big cough cough
ah nice, jethrogb has reduced it already ! playground
We're skipping the new check in this case.
is the "we can just skip ahead" comment now incorrect because of universes ?
It's incorrect because the relation that it's talking about is just a normal outlives predicate, which is not the same as what we're now doing.
good find :thumbs_up:
so #60449 _should_ have fixed it, as expected
@Matthew Jasper :tada: on fixing it :)
I manually checked on the bigger repro on the latest nightly and we have the expected error instead of an ICE :thumbs_up:
what will juniper experience next !