Do not ICE when combining unsized locals and async

[estebank]

Fix#61335.

[rust-highfive]

r? @matthewjasper

(rust_highfive has picked a reviewer for you, use r? to override)

[estebank]

We could add an if ty.is_generator() here, which would ensure the correct error is emitted, and just ICE in any other (in theory unreachable) case.

[Centril]

Uhm; const_prop is too my knowledge an optimization transform not necessary for correctness so why are we emitting an error here?

[oli-obk]

Yea I think it's totally fine to just keep doing what the old code did and ignore errors. They may be occuring because we're in a generic function and may be bogus for specific concrete types

[estebank]

Should we close this PR then?

[Centril]

Well; we cannot merge this PR as-is I think but maybe we can morph the PR into something else. You can always open a new one tho... up to you. :)

[rust-highfive]

The job mingw-check of your PR failed (pretty log, raw log). Through arcane magic we have determined that the following fragments from the build log may contain information about the problem.

Click to expand the log.

2019-09-16T23:05:06.9297783Z ##[command]git remote add origin https://github.com/rust-lang/rust
2019-09-16T23:05:06.9489989Z ##[command]git config gc.auto 0
2019-09-16T23:05:06.9562880Z ##[command]git config --get-all http.https://github.com/rust-lang/rust.extraheader
2019-09-16T23:05:06.9621225Z ##[command]git config --get-all http.proxy
2019-09-16T23:05:06.9766400Z ##[command]git -c http.extraheader="AUTHORIZATION: basic ***" fetch --force --tags --prune --progress --no-recurse-submodules --depth=2 origin +refs/heads/*:refs/remotes/origin/* +refs/pull/64527/merge:refs/remotes/pull/64527/merge
---
2019-09-16T23:14:09.3306147Z Checking rustc_mir v0.0.0 (/checkout/src/librustc_mir)
2019-09-16T23:14:09.4901056Z error: this file contains an un-closed delimiter
2019-09-16T23:14:09.4902525Z --> src/librustc_mir/transform/const_prop.rs:824:3
2019-09-16T23:14:09.4904053Z |
2019-09-16T23:14:09.4904635Z 651 | impl<'mir, 'tcx> MutVisitor<'tcx> for ConstPropagator<'mir, 'tcx> {
2019-09-16T23:14:09.4905236Z | - un-closed delimiter
2019-09-16T23:14:09.4905641Z ...
2019-09-16T23:14:09.4906147Z 690 | if self.should_const_prop() {
2019-09-16T23:14:09.4906879Z | - this delimiter might not be properly closed...
2019-09-16T23:14:09.4907665Z 696 | }
2019-09-16T23:14:09.4908158Z | - ...as it matches this but it has different indentation
2019-09-16T23:14:09.4908499Z ...
2019-09-16T23:14:09.4908879Z 824 | }
2019-09-16T23:14:09.4908879Z 824 | }
2019-09-16T23:14:09.4909249Z | ^
2019-09-16T23:14:09.4909389Z 2019-09-16T23:14:09.4928369Z error: expected one of `.`, `;`, `?`, `}`, or an operator, found `=>`
2019-09-16T23:14:09.4929713Z |
2019-09-16T23:14:09.4929713Z |
2019-09-16T23:14:09.4930176Z 700 | Err(LayoutError::Unsized(_ty)) => {
2019-09-16T23:14:09.4930725Z | ^^ expected one of `.`, `;`, `?`, `}`, or an operator here
2019-09-16T23:14:09.4951336Z error: unexpected `self` parameter in function
2019-09-16T23:14:09.4951939Z --> src/librustc_mir/transform/const_prop.rs:715:9
2019-09-16T23:14:09.4952355Z |
2019-09-16T23:14:09.4952914Z 715 | &mut self,
---
2019-09-16T23:14:11.5114772Z == clock drift check ==
2019-09-16T23:14:11.5134534Z local time: Mon Sep 16 23:14:11 UTC 2019
2019-09-16T23:14:11.6657494Z network time: Mon, 16 Sep 2019 23:14:11 GMT
2019-09-16T23:14:11.6658014Z == end clock drift check ==
2019-09-16T23:14:12.4254934Z ##[error]Bash exited with code '1'.
2019-09-16T23:14:12.4286848Z ##[section]Starting: Checkout
2019-09-16T23:14:12.4288502Z ==============================================================================
2019-09-16T23:14:12.4288568Z Task : Get sources
2019-09-16T23:14:12.4288609Z Description : Get sources from a repository. Supports Git, TfsVC, and SVN repositories.

I'm a bot! I can only do what humans tell me to, so if this was not helpful or you have suggestions for improvements, please ping or otherwise contact @TimNN. (Feature Requests)

[bjorn3]

An unsized type isnt generic, so returning TooGeneric doesnt make sense.

[RalfJung]

But that's also why this entire approach looks odd... a generic type will eventually be monomorphized, so that's a program that cannot be executed now but can be executed later. But if the type is unsized now it will always be unsized, right, so why does it make any sense to delay the error?

[oli-obk]

yea that should just be reported via the machinery below, which should simplify the rest

[bors]

☔ The latest upstream changes (presumably #64419) made this pull request unmergeable. Please resolve the merge conflicts.

[JohnCSimon]

Ping from triage
@estebank Can you please address the merge conflicts and address the comments?
CC: @matthewjasper
Thanks.

[JohnCSimon]

Pinging again from triage
@estebank Can you please address the merge conflicts and address the comments?
CC: @matthewjasper
Thanks.

[Centril]

This hasn't seen any movement for some time so I'm going to close it out as inactive.
As always, feel free to reopen whenever (but ideally with a more sound approach ^^).