upvar: Make sure that typeck succeeded before analyzing a closure

[barosl]

If a type that cannot be called as a function is called, and the result is assigned to a variable, the "unexpected callee type" ICE occurs.

Fixes #20714.
Fixes #20842.
Fixes #20862.

[rust-highfive]

r? @pcwalton

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

[nikomatsakis]

From IRC:

[11:48:39] <nmatsakis> barosl: I have mixed feelings; we used to do this, but it made it hard to know when errors would be reported, in turn making it hard to write reliable test cases
[11:48:57] <nmatsakis> in general I've been modifying instead to have things return Err() in theoretically impossible cases

[alexcrichton]

re-r? @nikomatsakis

[nikomatsakis]

I'm still not particularly keen on this patch. I'd rather make EUV more tolerant somehow. I'm torn between just returning rather than calling span_bug or possibly having the delegate have a fn tolerate_errors(&self) -> bool method that is only true during typeck. That seems nicer. @barosl what do you think of that?

[nikomatsakis]

I apologize for the extreme delay in responding on this bug. Believe it or not I am digging myself out of a github notification hole that has accumulated and only saw @alexcrichton's comments now.

[alexcrichton]

ping @nikomatsakis

[tamird]

20714 is already fixed fwiw

[nikomatsakis]

It seems that two out of the three relevant issues have been fixed. Given how old and inactive this PR is, I'm going to close it. Feel free to re-open with a newer version if desired. :)