Fix #12939: reporting of ignored tests in rustdoc

[ipetkov]

librustdoc: instead of skipping ignored tests, pass them to libtest
so it can report them as such. If a test is marked as notrust,
however, it will not show up in the final report.

Fix #12939

[alexcrichton]

This condition should change slightly I think. It should consider ignore as "print this during the test run, but don't run it" and notrust as "ignore this entirely for testing".

[ipetkov]

That's a really good point, hadn't thought about that!

[alexcrichton]

I'm a little confused about the changes to libtest, could you provide a before/after of what you changed? For example, this is what I get today:

#[test] fn a() {}
#[ignore] #[test] fn b() {}

$ ./foo                     

running 2 tests
test b ... ignored
test a ... ok

test result: ok. 1 passed; 0 failed; 1 ignored; 0 measured

[ipetkov]

Sorry I should have made this clearer: this change affects running documentation tests under rustdoc, e.g.

// foo.rs
#![crate_id="foo"]

//! This is a small crate

/**
```rust
assert!(true);
```

```ignore
assert!(!false);
```
*/
/// foo always returns 5
pub fn foo() -> int {
    5
}

$ rustdoc --test foo.rs 

running 1 test
test foo_0 ... ok

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured

This is the result I get after building from master, when it should really output test result: 1 passed; 0 failed; 1 ignored; 0 measured for the last line.

I'm a little confused about the changes to libtest, could you provide a before/after of what you changed?

Looking over the diff again I just realized my misunderstanding of some code in libtest. I (wrongly) believed that libtest was also dropping tests marked as ignored, and tried to "improve" the code. The main problem for this issue is just rustdoc ignoring the tests. I'll remove my changes to libtest and update the PR branch.

[ipetkov]

Updated the PR branch, and re-rebased on master