Comments mess with rustdoc's auto-test insertion behavior

[steveklabnik]

Simplest to see in action:

C:\Users\steve\tmp> cat foo.rs
/// ```
/// #![feature(box_syntax)]
///
/// let five = box 5;
/// ```
///
pub fn passes() {}

/// ```
/// // a comment
/// #![feature(box_syntax)]
///
/// let five = box 5;
/// ```
///
pub fn fails() {}
C:\Users\steve\tmp> rustdoc --test foo.rs

running 2 tests
test foo.rs - fails (line 9) ... FAILED
test foo.rs - passes (line 1) ... ok

failures:

---- foo.rs - fails (line 9) stdout ----
error[E0658]: box expression syntax is experimental; you can call `Box::new` instead. (see issue #49733)
 --> foo.rs:13:12
  |
6 | let five = box 5;
  |            ^^^^^
  |
  = help: add #![feature(box_syntax)] to the crate attributes to enable

thread 'foo.rs - fails (line 9)' panicked at 'couldn't compile the test', librustdoc\test.rs:325:17
note: Run with `RUST_BACKTRACE=1` for a backtrace.


failures:
    foo.rs - fails (line 9)

test result: FAILED. 1 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out

including the comment means that everything gets wrapped in a main, which means the feature flag is wrapped in a main, and therefore doesn't turn on. this is very confusing.

[carols10cents]

I believe this was fixed by #54861; I can no longer reproduce this.

(I was looking into how rustdoc's fn main wrapping works for book reasons and found this issue and that PR :))