Resolve failures cause too many typechecking failures

[alexcrichton]

I noticed that on #31618 there was a resolve failure from this line, specifically:

../src/libstd/sys/unix/pipe.rs:94:13: 94:25 error: unresolved name `libc::select` [E0425]
../src/libstd/sys/unix/pipe.rs:94             libc::select(max + 1, &mut read, 0 as *mut _, 0 as *mut _,
                                              ^~~~~~~~~~~~
../src/libstd/sys/unix/pipe.rs:90:9: 96:13 note: in this expansion of try! (defined in <core macros>)
../src/libstd/sys/unix/pipe.rs:94:13: 94:25 help: run `rustc --explain E0425` to see a detailed explanation
../src/libstd/sys/unix/pipe.rs:90:14: 90:19 error: the trait `core::num::One` is not implemented for the type `()` [E0277]
../src/libstd/sys/unix/pipe.rs:90         try!(cvt_r(|| unsafe {
                                               ^~~~~
../src/libstd/sys/unix/pipe.rs:90:9: 96:13 note: in this expansion of try! (defined in <core macros>)
../src/libstd/sys/unix/pipe.rs:90:14: 90:19 help: run `rustc --explain E0277` to see a detailed explanation
../src/libstd/sys/unix/pipe.rs:90:14: 90:19 note: required by `sys::cvt_r`
../src/libstd/sys/unix/pipe.rs:90:14: 90:19 error: the trait `core::ops::Neg` is not implemented for the type `()` [E0277]
../src/libstd/sys/unix/pipe.rs:90         try!(cvt_r(|| unsafe {
                                               ^~~~~
../src/libstd/sys/unix/pipe.rs:90:9: 96:13 note: in this expansion of try! (defined in <core macros>)
../src/libstd/sys/unix/pipe.rs:90:14: 90:19 help: run `rustc --explain E0277` to see a detailed explanation
../src/libstd/sys/unix/pipe.rs:90:14: 90:19 note: required by `sys::cvt_r`
../src/libstd/sys/unix/pipe.rs:94:46: 94:57 error: casting `usize` as `*mut _` is invalid
../src/libstd/sys/unix/pipe.rs:94             libc::select(max + 1, &mut read, 0 as *mut _, 0 as *mut _,
                                                                               ^~~~~~~~~~~
../src/libstd/sys/unix/pipe.rs:90:9: 96:13 note: in this expansion of try! (defined in <core macros>)
../src/libstd/sys/unix/pipe.rs:94:59: 94:70 error: casting `usize` as `*mut _` is invalid
../src/libstd/sys/unix/pipe.rs:94             libc::select(max + 1, &mut read, 0 as *mut _, 0 as *mut _,
                                                                                            ^~~~~~~~~~~
../src/libstd/sys/unix/pipe.rs:90:9: 96:13 note: in this expansion of try! (defined in <core macros>)
../src/libstd/sys/unix/pipe.rs:95:26: 95:37 error: casting `usize` as `*mut _` is invalid
../src/libstd/sys/unix/pipe.rs:95                          0 as *mut _)
                                                           ^~~~~~~~~~~
../src/libstd/sys/unix/pipe.rs:90:9: 96:13 note: in this expansion of try! (defined in <core macros>)

The only real error here is that libc::select doesn't exist. All other errors are spawned off of this. This unfortunately makes it pretty hard to hunt through errors and figure out which ones need to get fixed, and at this point I'm just training myself to only look at the first error as all others are likely not too useful.

I attempted to minimize this as well:

mod foo {}

fn bar<T: Clone>() {}

fn main() {
    bar(foo::bar());
}

But this yielded a different error message:

foo.rs:6:9: 6:17 error: unresolved name `foo::bar` [E0425]
foo.rs:6     bar(foo::bar());
                 ^~~~~~~~
foo.rs:6:9: 6:17 help: run `rustc --explain E0425` to see a detailed explanation
foo.rs:6:5: 6:20 error: this function takes 0 parameters but 1 parameter was supplied [E0061]
foo.rs:6     bar(foo::bar());
             ^~~~~~~~~~~~~~~
foo.rs:6:5: 6:20 help: run `rustc --explain E0061` to see a detailed explanation
error: aborting due to previous error

Note that on stable this yields a message which is actionable and concise:

foo.rs:6:9: 6:17 error: unresolved name `foo::bar` [E0425]
foo.rs:6     bar(foo::bar());
                 ^~~~~~~~
foo.rs:6:9: 6:17 help: run `rustc --explain E0425` to see a detailed explanation
error: aborting due to previous error

[arielb1]

The type error is correct in your "minified" example - bar takes zero parameters and you passed one.

[alexcrichton]

Guh right, sorry!

[nikomatsakis]

I think this is a roughly minimized example:

trait TheTrait { }

fn closure<F, T>(x: F) -> Result<T, ()>
    where F: FnMut() -> T, T: TheTrait,
{
    unimplemented!()
}

fn foo() {
    try!(closure(|| bar(0 as *mut _))) // bar intentionally undefined
}

fn main() { }

For me it gets:

/home/nmatsakis/tmp/issue-31997.rs:10:21: 10:24 error: unresolved name `bar` [E0425]
/home/nmatsakis/tmp/issue-31997.rs:10     try!(closure(|| bar(0 as *mut _))) // bar intentionally undefined
                                                          ^~~
/home/nmatsakis/tmp/issue-31997.rs:10:5: 10:39 note: in this expansion of try! (defined in <std macros>)
/home/nmatsakis/tmp/issue-31997.rs:10:21: 10:24 help: run `rustc --explain E0425` to see a detailed explanation
<std macros>:5:8: 6:42 error: mismatched types:
 expected `()`,
    found `core::result::Result<_, _>`
(expected (),
    found enum `core::result::Result`) [E0308]
<std macros>:5 return $ crate:: result:: Result:: Err (
<std macros>:6 $ crate:: convert:: From:: from ( err ) ) } } )
/home/nmatsakis/tmp/issue-31997.rs:10:5: 10:39 note: in this expansion of try! (defined in <std macros>)
<std macros>:5:8: 6:42 help: run `rustc --explain E0308` to see a detailed explanation
/home/nmatsakis/tmp/issue-31997.rs:10:10: 10:17 error: the trait `TheTrait` is not implemented for the type `()` [E0277]
/home/nmatsakis/tmp/issue-31997.rs:10     try!(closure(|| bar(0 as *mut _))) // bar intentionally undefined
                                               ^~~~~~~
/home/nmatsakis/tmp/issue-31997.rs:10:5: 10:39 note: in this expansion of try! (defined in <std macros>)
/home/nmatsakis/tmp/issue-31997.rs:10:10: 10:17 help: run `rustc --explain E0277` to see a detailed explanation
/home/nmatsakis/tmp/issue-31997.rs:10:10: 10:17 note: required by `closure`
/home/nmatsakis/tmp/issue-31997.rs:10:25: 10:36 error: casting `usize` as `*mut _` is invalid
/home/nmatsakis/tmp/issue-31997.rs:10     try!(closure(|| bar(0 as *mut _))) // bar intentionally undefined
                                                              ^~~~~~~~~~~
/home/nmatsakis/tmp/issue-31997.rs:10:5: 10:39 note: in this expansion of try! (defined in <std macros>)
error: aborting due to 3 previous errors

[nikomatsakis]

Note in particular that this includes:

/home/nmatsakis/tmp/issue-31997.rs:10:10: 10:17 error: the trait `TheTrait` is not implemented for the type `()` [E0277]

which is pretty wacky, since () is not really a useful anything there.

[nikomatsakis]

(Making some progress on this.)

[alexcrichton]

Another small example yields the errors:

foo.rs:16:19: 16:31 error: failed to resolve. Use of undeclared type or module `HashMap` [E0433]
foo.rs:16     let mut map = HashMap::new();
                            ^~~~~~~~~~~~
foo.rs:16:19: 16:31 help: run `rustc --explain E0433` to see a detailed explanation
foo.rs:16:19: 16:31 error: unresolved name `HashMap::new` [E0425]
foo.rs:16     let mut map = HashMap::new();
                            ^~~~~~~~~~~~
foo.rs:16:19: 16:31 help: run `rustc --explain E0425` to see a detailed explanation
foo.rs:33:25: 33:55 error: the type of this value must be known in this context
foo.rs:33         let mut entry = map.entry(instance.to_owned()).or_insert(Instance {
                                  ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
foo.rs:40:21: 40:30 error: the type of this value must be known in this context
foo.rs:40             assert!(entry.end.is_none());
                              ^~~~~~~~~
foo.rs:40:13: 40:42 note: in this expansion of assert! (defined in <std macros>)
error: aborting due to 2 previous errors

The first error:

failed to resolve. Use of undeclared type or module `HashMap`

is the only real error, everything else is just spurious after that

[nikomatsakis]

My PR in #32258 doesn't really help those cases; I think I could get rid of those "type of this value must be known" errors easily enough, at least.

[nikomatsakis]

@alexcrichton pushed a commit that changes the output for that test too:

issue-31997.rs:30:19: 30:31 error: failed to resolve. Use of undeclared type or module `HashMap` [E0433]
issue-31997.rs:30     let mut map = HashMap::new();
                                    ^~~~~~~~~~~~
issue-31997.rs:30:19: 30:31 help: run `rustc --explain E0433` to see a detailed explanation
issue-31997.rs:30:19: 30:31 error: unresolved name `HashMap::new` [E0425]
issue-31997.rs:30     let mut map = HashMap::new();
                                    ^~~~~~~~~~~~
issue-31997.rs:30:19: 30:31 help: run `rustc --explain E0425` to see a detailed explanation
error: aborting due to 2 previous errors

[alexcrichton]

Thanks @nikomatsakis! I think that's basically where we were at in ancient times, I've seen that double error for things like foo::bar for ages (cc @jseyfried, feeling ambitious?)

[jseyfried]

@alexcrichton Fixed, see #32789.

[alexcrichton]

Awesome, thanks @jseyfried!