Before:

![rustdoc-intra-doc-link-summary-before](https://user-images.githubusercontent.com/1593513/122623069-9d995e80-d04f-11eb-8d46-ec2ec126bb5e.png)

After:

![rustdoc-intra-doc-link-summary](https://user-images.githubusercontent.com/1593513/122623076-a4c06c80-d04f-11eb-967a-f5916871c34b.png)

It's much more complicated than just going through the TyCtxt.

This makes it consistent with rustc_session::Options.

…tsakis

Update array_into_iter lint for 1.53 and edition changes.

This updates the array_into_iter lint for Rust 1.53 and the edition changes.

See rust-lang#84513

r? `@estebank`

…ohnTitor

Add regression test for issue rust-lang#52025

Closes rust-lang#52025

Took the test from rust-lang#52025

rustdoc: add optional woff2 versions of Source Serif and Source Code

This provides woff2 versions of Source Serif and Source Code similar to how rust-lang#82545 provides woff2 versions of Fira Sans. The total byte count for the six files (three for each font family) is reduced by 25% from 476 KiB to 358 KiB.

…oc-link-summary, r=CraftSpider

Resolve intra-doc links in summary desc

Before:

![rustdoc-intra-doc-link-summary-before](https://user-images.githubusercontent.com/1593513/122623069-9d995e80-d04f-11eb-8d46-ec2ec126bb5e.png)

After:

![rustdoc-intra-doc-link-summary](https://user-images.githubusercontent.com/1593513/122623076-a4c06c80-d04f-11eb-967a-f5916871c34b.png)

Cleanup handling of `crate_name` for doctests

- Remove unnecessary reallocation
- Remove unnecessary messing around with `queries` for the crate name; it's simpler and faster to go through the TyCtxt instead
- Rename `cratename` -> `crate_name` for consistency with the rest of the compiler

…eSeulArtichaut

Fix `unused_unsafe` around `await`

Enables `unused_unsafe` lint for `unsafe { future.await }`.

The existing test for this is `unsafe { println!() }`, so I assume that `println!` used to contain compiler-generated unsafe but this is no longer true, and so the existing test is broken. I replaced the test with `unsafe { ...await }`. I believe `await` is currently the only instance of compiler-generated unsafe.

Reverts some parts of rust-lang#85421, but the issue predates that PR.

…ts-check-pass, r=JohnTitor

Mark some edition tests as check-pass

## Overview
This helps with rust-lang#62277. In short, there are some tests that were marked as `build-pass` when it was unclear whether `check-pass` might be more appropriate. This PR marks some of those tests as `compile-pass`, in addition to making some incidental formatting improvements.

## A brief explanation of why this is correct
These tests fall into a few buckets.

`src/test/ui/dyn-keyword/dyn-2015-edition-keyword-ident-lint.rs`
`src/test/ui/dyn-keyword/dyn-2015-idents-in-decl-macros-unlinted.rs`
`src/test/ui/dyn-keyword/dyn-2015-idents-in-macros-unlinted.rs`
`src/test/ui/dyn-keyword/dyn-2015-no-warnings-without-lints.rs`
`src/test/ui/dyn-keyword/issue-56327-dyn-trait-in-macro-is-okay.rs`

These test a lint for a keyword added in a new edition and the corresponding changes in keyword rules.

`src/test/ui/editions/edition-feature-ok.rs`
This checks that a feature related to an edition transition is valid.

`src/test/ui/editions/edition-imports-virtual-2015-ambiguity.rs`
This checks that imports between editions work correctly.

`src/test/ui/editions/edition-keywords-2015-2015-expansion.rs`
`src/test/ui/editions/edition-keywords-2018-2015-expansion.rs`
This checks the interaction between a change in keyword status over editions and macros.

All of the things being tested come before linking and codegen, so it is safe to use `check-pass` for them.