It has a single call site. This will enable an optimization in the next
commit.

Do a `NEEDS_CANONICAL` check on the entire input. We already have such
checks on the relevant parts, but this lets us entirely avoid building a
`Canonicalizer` when the check fails.

It causes it to intern or not intern. Instead we can just let the
call site that passes in `true` do its own interning.

`decode_error_kind` for Hermit only knew 15 errno values and returned
`Uncategorized` for all others. For example, removing a non-empty
directory failed with `Uncategorized` instead of `DirectoryNotEmpty`.

Hermit uses the Linux errno numbering, so use the same mapping as on
Unix. `hermit-abi` defines no `ENOTSUP` and `EWOULDBLOCK` is an alias
of `EAGAIN`, so both need no separate arm.

…ts should be skipped in Miri due to missing shims

It is represented as an ADT, so needs custom logic to be accepted. The point of this type is that it's compatible with its C counterpart, so the lint definitely should accept it.

```
error[E0364]: `f` is private, and cannot be re-exported
  --> $DIR/privacy-early.rs:10:13
   |
LL |     pub macro m() {
LL |         use f as g;
   |             ^^^^^^
...
LL |     foo::m!();
   |     --------- in this macro invocation
   |
note: consider marking `f` as `pub` in the imported module
  --> $DIR/privacy-early.rs:6:5
   |
LL |     fn f() {}
   |     ^^^^^^^^^
```
where it was
```
error[E0364]: `f` is private, and cannot be re-exported
  --> $DIR/privacy-early.rs:10:13
   |
LL |     pub macro m() {
LL |         use f as g;
   |             ^^^^^^
...
LL |     foo::m!();
   |     --------- in this macro invocation
   |
note: consider marking `f` as `pub` in the imported module
  --> $DIR/privacy-early.rs:10:13
   |
LL |         use f as g;
   |             ^^^^^^
...
LL |     foo::m!();
   |     --------- in this macro invocation
```

Use span when suggesting making module public:

```
note: consider declaring module `foo` with `pub`
  --> $DIR/E0365.rs:1:1
   |
LL | / mod foo {
LL | |     pub const X: u32 = 1;
LL | | }
   | |_^
```

Show container when talking about enclosing module:

```
note: consider marking `f` as `pub` in the imported module
  --> $DIR/privacy-early.rs:6:5
   |
LL | mod foo {
LL |     fn f() {}
   |     ^^^^^^^^^
```

Use accurate descriptor for enclosing item:

```
note: consider declaring struct `S1` with `pub`
  --> $DIR/crate-private-reexport.rs:3:1
   |
LL | / struct S1 {
LL | |     #[rustfmt::skip]
LL | |     bar: i32,
LL | | }
   | |_^
```
```
note: consider marking `Full` as `pub` in the imported variant
  --> $DIR/issue-46209-private-enum-variant-reexport.rs:26:9
   |
LL |     enum Lieutenant {
LL |         JuniorGrade,
LL |         Full,
   |         ^^^^
```

like commit cbdd3f3 did for stable.

…e: field is not printed

* In irrefutable patterns, the FieldPat is skipped, which means the name: field is not printed;
since FieldPat can be listed out of order and the list can be incomplete, it is not generally
possible to associate a given pattern within a Leaf or Variant with the actual field.

the slightly janky indentation is consistent with the way the other PatKinds do it
(with no comma)
* avoid referring to std in THIR tests
* review feedback (test too long)
* review feedback (naming, explanation, test too long)
* additional review feedback (test still too long)

…r=maxdexh

std: use skip_unsupported_miri_tests to control when tests should be skipped in Miri due to missing shims

We currently run some subset of std tests in Miri, skipping some because Miri doesn't support the operations they need. That is based on filters in the makefile. People don't know what is actually being tested and add random `cfg(miri)` in places to disable tests that were never run. This makes it quite hard to track which tests are being skipped and which operations we should add in Miri to get 100% std coverage. So let's replace that string-based filter with a cfg flag, `skip_unsupported_miri_tests`, that can be annotated at the tests.

Unfortunately I can't figure out how to disable all doctests in a module based on a cfg flag, so this entirely disables std doctests being run in Miri in our CI. (We still run them in https://github.com/rust-lang/miri-test-libstd.)

See prior discussion [on Zulip](https://rust-lang.zulipchat.com/#narrow/channel/219381-t-libs/topic/Miri.20vs.20tests.20in.20std/with/635913743).
Docs PR: rust-lang/std-dev-guide#86

r? @tgross35
try-job: test-x86_64-gnu-aux

…beta, r=Kobzol

Normalize "beta release channel" in `const-supertrait-unstable` test

like commit cbdd3f3 did for stable.

It failed on beta in <rust-lang#165713 (comment)>.

r? @Kobzol

…, r=programmerjake

Wrap proc-macro `EscapeError` enum into a struct

As discussed in rust-lang#151973.

Part of rust-lang#165750.

r? @programmerjake

Tweak private re-export errors to point at correct span and have more context

Use the correct span when talking about imported module that should be marked as public:
```
error[E0364]: `f` is private, and cannot be re-exported
  --> $DIR/privacy-early.rs:10:13
   |
LL |     pub macro m() {
LL |         use f as g;
   |             ^^^^^^
...
LL |     foo::m!();
   |     --------- in this macro invocation
   |
note: consider marking `f` as `pub` in the imported module
  --> $DIR/privacy-early.rs:6:5
   |
LL |     fn f() {}
   |     ^^^^^^^^^
```
where it was:
```
error[E0364]: `f` is private, and cannot be re-exported
  --> $DIR/privacy-early.rs:10:13
   |
LL |     pub macro m() {
LL |         use f as g;
   |             ^^^^^^
...
LL |     foo::m!();
   |     --------- in this macro invocation
   |
note: consider marking `f` as `pub` in the imported module
  --> $DIR/privacy-early.rs:10:13
   |
LL |         use f as g;
   |             ^^^^^^
...
LL |     foo::m!();
   |     --------- in this macro invocation
```

Use span when suggesting making module public:

```
note: consider declaring module `foo` with `pub`
  --> $DIR/E0365.rs:1:1
   |
LL | / mod foo {
LL | |     pub const X: u32 = 1;
LL | | }
   | |_^
```

Show container when talking about enclosing module:

```
note: consider marking `f` as `pub` in the imported module
  --> $DIR/privacy-early.rs:6:5
   |
LL | mod foo {
LL |     fn f() {}
   |     ^^^^^^^^^
```

Use accurate descriptor for enclosing item:

```
note: consider declaring struct `S1` with `pub`
  --> $DIR/crate-private-reexport.rs:3:1
   |
LL | / struct S1 {
LL | |     #[rustfmt::skip]
LL | |     bar: i32,
LL | | }
   | |_^
```
```
note: consider marking `Full` as `pub` in the imported variant
  --> $DIR/issue-46209-private-enum-variant-reexport.rs:26:9
   |
LL |     enum Lieutenant {
LL |         JuniorGrade,
LL |         Full,
   |         ^^^^
```

…names, r=adwinwhite

In irrefutable patterns, the FieldPat is skipped, which means the name: field is not printed

since FieldPat can be listed out of order and the list can be incomplete, it is not generally possible to associate a given pattern within a Leaf or Variant with the actual field.

the slightly janky indentation is consistent with the way the other PatKinds do it

…-opts, r=ShoyuVanilla

Canonicalization cleanups

I saw some moderate perf improvements locally with these changes but on CI they had no effect. Maybe due to PGO? Most of the changes are actually refactorings which I think are still worth merging.

r? @ShoyuVanilla

…, r=JonathanBrouwer

Remove unnecessary mut from HirIdValidator::check_nested_id

Remove unnecessary `mut` from `HirIdValidator::check_nested_id`, as this method doesn't need a mutated self. See [#t-compiler/help > The check_nested_id in HirIdValidator  doesn't need mut](https://rust-lang.zulipchat.com/#narrow/channel/182449-t-compiler.2Fhelp/topic/The.20check_nested_id.20in.20HirIdValidator.20.20doesn.27t.20need.20mut/with/639210467)

The changes compile successfully and pass the `./x check compiler/rustc_passes` check.

…, r=beetrees

make improper_ctypes accept `x87_f80`

tracking issue: rust-lang#163982

Like rust-lang#165274, but for `x87_f80`.

It is represented as an ADT, so needs custom logic to be accepted. The point of this type is that it's compatible with its C counterpart, so the lint definitely should accept it.

r? beetrees

std: map more errno values to ErrorKind on Hermit

`decode_error_kind` for Hermit only knew 15 errno values and returned `Uncategorized` for all others. For example, removing a non-empty directory failed with `Uncategorized` instead of `DirectoryNotEmpty`.

Hermit uses (more or less) the Linux errno numbering, so use the same mapping as on Unix. `hermit-abi` defines no `ENOTSUP` and `EWOULDBLOCK` is an alias of `EAGAIN`, so both need no separate arm.