Type errors with for and do expressions could be cleverer

[ben0x539]

Calling functions meant to be used in for/do expressions in the respective other kind of expression gives type errors. While that's correct, it's not immediately clear where the error is (not even the squiggly lines under the code excerpts are very clear), and I think it should be yelling at me directly for using the wrong syntax sugar instead of outlining how it noticed that I screwed up. Here is some examples that I found a bit misleading:

foo.rs:2:14: 2:36 error: a `loop` function's last argument should return `bool`, not `()`
foo.rs:2     for task::spawn { return true; }
                       ^~~~~~~~~~~~~~~~~~~~~~

I'm returning a bool, what's the matter! I expect rustc is pointing at the type of task::spawn here even though my closure is underlined and the mismatch is really between for and the function type.

Similarly, in the inverse case,

foo.rs:2:15: 2:17 error: mismatched types: expected `bool` but found `()` (expected bool but found ())
foo.rs:2     do 5.times {}
                        ^~

Well, okay then, let me just return true...

foo.rs:2:24: 2:28 error: mismatched types: expected `()` but found `bool` (expected () but found bool)
foo.rs:2     do 5.times { return true; }
                                 ^~~~

Argh! Which is it! ;-)

[ben0x539]

I assume this got handled in #4141.

[ben0x539]

Looking closer, I think the first example could still do with mentioning that I should use do. Or maybe I'm too thick-headed for the loop protocol.

I still feel a bit weird that do 5.times { false } compiles, too.

[catamorphism]

I'll take a look at this.

[catamorphism]

There are definitely a couple things to fix here. First, the error message itself verges on misleading. In my branch, I've improved it quite a bit. Second, there are multiple errors. This is a case where my work on suppressing derived errors wasn't doing the trick; I've extended it a bit, but the output still doesn't look as good as I'd hope. Third, the spans (which are what tell the pretty-printer where to put the squiggly lines) are wrong. (I have no idea why).

I'm working on it!

[catamorphism]

Fixed in 5b64c79 -- please start a new issue if you find other confusing error messages like this! I love improving error messages.