r? @nrc

With this PR we can save-analysis on code with errors, essential foundation work for IDE support.

Some of this is scary stuff. Probably time to lint against this.

Found with `make --warn-undefined-variables`.

r? @alexcrichton

Also add a note about the necessary simplicity of the conversion.
Related issue: #29349

r? @steveklabnik

This PR refactors away `Module`'s `external_module_children` and instead puts `extern crate` declarations in `children` like other items, simplifying duplicate checking and name resolution.

This PR also allows values to share a name with extern crates, which are only defined in the type namespace. Other than that, it is a pure refactoring.

r? @nrc

Currently any compilation to MIPS spits out the warning:

    'generic' is not a recognized processor for this target (ignoring processor)

Doesn't make for a great user experience! We don't encounter this in the normal
bootstrap because the cpu/feature set are set by the makefiles. Instead let's
just propagate these to the defaults for the entire target all the time (still
overridable from the command line) and prevent warnings from being emitted by
default.

r? @nrc

Some of this is scary stuff. Probably time to lint against this.

Found with `make --warn-undefined-variables`.

```
enum Univariant {
    X = 17
}
```

Fixes #10292

This reverts PR #30324, fixing bug #30159 in which a public a glob import makes public any preceding imports that share a name with an item in the module being glob imported from.

For example,
```rust
pub fn f() {}
pub mod foo {
    fn f() {}
}

mod bar {
    use f;
    use f as g;
    pub use foo::*; // This makes the first import public but does not affect the second import.
}
```

This is a [breaking-change].

The target was meant as a modern generic `armv7` option, therefore a few changes were necessary:

- gcc's `-march=armv7` was causing compilation failures on modern linux systems
- rust codegen defaulted to `cortex-a7` causing illegal instruction crashes on previous `armv7-a` processors (e.g, cortex-a5, cortex-a8)

Fixes #31207

by removing abort_if_new_errors

This target covers MIPS devices that run the trunk version of OpenWRT.

The x86_64-unknown-linux-musl target always links statically to C libraries. For
the mips(el)-unknown-linux-musl target, we opt for dynamic linking (like most of
other targets do) to keep binary size down.

As for the C compiler flags used in the build system, we use the same flags used
for the mips(el)-unknown-linux-gnu target.

r? @alexcrichton

fixes failed test (std::os::raw::tests::unix) in x86_64-musl target

Revert PR #30324, fixing a bug in which a public glob import can make preceding imports public (fixes #30159).

r? @steveklabnik

cf #31303

This is a PR for #21195. It changes the way unspecified `help` and `ǹote` messages are handled in compile-fail tests as suggested by @oli-obk in the issue: if there are some `note` or `help` annotations, there must be annotations for all `help` or `note` messages of this test. Maybe it makes also sense to add an option to specify that the this test should fail if there are unspecified `help` or `note` messages.

With this change, the following tests fail:

    [compile-fail] compile-fail/changing-crates.rs
    [compile-fail] compile-fail/default_ty_param_conflict_cross_crate.rs
    [compile-fail] compile-fail/lifetime-inference-give-expl-lifetime-param.rs
    [compile-fail] compile-fail/privacy1.rs
    [compile-fail] compile-fail/svh-change-lit.rs
    [compile-fail] compile-fail/svh-change-significant-cfg.rs
    [compile-fail] compile-fail/svh-change-trait-bound.rs
    [compile-fail] compile-fail/svh-change-type-arg.rs
    [compile-fail] compile-fail/svh-change-type-ret.rs
    [compile-fail] compile-fail/svh-change-type-static.rs
    [compile-fail] compile-fail/svh-use-trait.rs

I'll add the missing annotations if we decide to accept this change.

the previous code generated a temporary of the inner type and assigned the box-memory to it. So if you did `let x: Box<usize> = box 5;` you got a

```rust
let var0: Box<usize>; // x
let mut tmp0: Box<usize>;
let mut tmp1: usize;
...
tmp1 = Box(usize);
(*tmp1) = const 5;
tmp0 = tmp1;
var0 = tmp0;
```

r? @nagisa

I don't believe these test cases have served any purpose in years.

The shootout benchmarks are now upstreamed. A new benchmark suite
should rather be maintained out of tree.

r? @nikomatsakis

- Successful merges: #31099, #31244, #31280, #31290, #31292, #31294, #31295, #31296
- Failed merges:

Rustdoc will automatically wrap things in main, but this doesn't work
here.

Fixes #31249

These are free functions in the text, but methods in the standard
library.

Fixes #31266

This code was refactored, but the words were not

Fixes #31284

This test has been deadlocking and causing problems on the bots basically since
its inception. Some memory safety issues were fixed in 987dc84b, but the
deadlocks remained afterwards unfortunately.

After some investigation, I've concluded that this is just a situation where OSX
is not guaranteed to run destructors. The fix in 987dc84b observed that OSX was
rewriting the backing TLS memory to its initial state during destruction while
we weren't looking, and this would have the effect of canceling the destructors
of any other initialized TLS slots.

While very difficult to pin down, this is basically what I assume is happening
here, so there doesn't seem to really be anythig we can do to ensure the test
robustly passes on OSX, so just ignore it for now.

We forgot to pass down the `-m64` flag to gcc, so we were actually compiling
powerpc code which would then later fail to link!

Perhaps, it makes more sense to link straight to [Installing on Linux or Mac](https://doc.rust-lang.org/book/getting-started.html#installing-on-linux-or-mac) and [Installing on Windows](https://doc.rust-lang.org/book/getting-started.html#installing-on-windows) sections. Platform Support section could be skipped in case of *quick* start.

This commit refactors the field `Module::children` from mapping `Name` -> `NameBindings` to mapping `(Name, Namespace)` -> `NameBinding` and refactors the field `Module::import_resolutions` from mapping `Name` -> `ImportResolutionPerNamespace` to mapping `(Name, Namespace)` -> `ImportResolution`.

This allows the duplicate checking code to be refactored so that `NameBinding` no longer needs ref-counting or a RefCell (removing the need for `NsDef`).

r? @nikomatsakis