`take` + `into_iter` + pattern matching is nicer than `drain` + `map` +
`collect`.

In rust-lang#119972 the code should have become `E0123` rather than `0123`. This
fix doesn't affect the outcome because the proc macro errors out before
the type of the code is checked, but the fix makes the test's code
consistent with other similar code elsewhere.

It contains an `i128`, but when creating them we convert any number
outside the range -100..100 to a string, because Fluent uses an `f64`.
It's all a bit strange.

This commit changes the `i128` to an `i32`, which fits safely in
Fluent's `f64`, and removes the -100..100 range check. This means that
only integers outside the range of `i32` will be converted to strings.

When there are two possibilities, both of which use a `String`, it's
nicer to use a struct than an enum. Especially when mapping the contents
into a tuple.

Some of these tests were originally written as part of a custom `run-make`
test, so at that time they weren't able to use the normal compiletest header
directive parser.

Now that they're properly integrated, there's no need for them to use
`compile-flags` to specify the edition, since they can use `edition` instead.

* Get rid of a typo in a function name
* Rename `currently_processing_generics`: The old name confused me at first since
  I assumed it referred to generic *parameters* when it was in fact referring to
  generic *arguments*. Generics are typically short for generic params.
* Get rid of a few unwraps by properly leveraging slice patterns

To match `buffer_error`.

It has no effect. Note that `infcx.set_tainted_by_errors()` is still
called, so taintedness is still being propagated.

`BorrowckErrors` stores a mix of error and non-error diags in
`buffered`. As a result, it downgrades `DiagnosticBuilder`s to
`Diagnostic`s, losing the emission guarantees, and so has to use a
`tainted_by_errors` field to record whether an error has occurred.

This commit splits `buffered` into `buffered_errors` and
`buffered_non_errors`, keeping them as `DiagnosticBuilder`s and
preserving the emission guarantees.

This also requires fixing a bunch of incorrect lifetimes on
`DiagnosticBuilder` use points.

It now has a single call site.

And some related things. Because it can hold non-error diagnostics.

`DiagnosticBuilderInner` was removed some time ago.

- `emitted_at` isn't used outside the crate.
- `code` and `messages` are public fields, so there's no point have
  trivial getters/setters for them.
- `suggestions` is public, so the comment about "functionality on
  `Diagnostic`" isn't needed.

`emit_future_breakage` calls
`self.dcx().take_future_breakage_diagnostics()` and then passes the
result to `self.dcx().emit_future_breakage_report(diags)`. This commit
removes the first of these and lets `emit_future_breakage_report` do the
taking.

It also inlines and removes what is left of `emit_future_breakage`,
which has a single call site.

It's only ever used with a reference to `OwnerInfo` as an argument.

Co-authored-by: Michael Goulet <michael@errs.io>

add avx512fp16 to x86 target features

std_detect avx512fp16: rust-lang/stdarch#1508

…ark-Simulacrum

Release notes for 1.76

Cargo, library stabilizations and some cleanups, particularly to future compat, still pending.

cc `@cuviper` `@rust-lang/release`

Revert unsound libcore changes

fixes rust-lang#120537

these were introduced in rust-lang#119911

coverage: Use normal `edition:` headers in coverage tests

Some of these tests were originally written as part of a custom `run-make` test, so at that time they weren't able to use the normal compiletest header directive parser.

Now that they're properly integrated, there's no need for them to use `compile-flags` to specify the edition, since they can use `edition` instead.

In most cases the `.cov-map` snapshot isn't affected at all, but in a few cases we add or remove a line, which slightly disturbs the first line number in each instrumented function.

… r=compiler-errors

Suggest changing type to const parameters if we encounter a type in the trait bound position

The first commit is just drive-by cleanup.

Provide a structured suggestion if the user forgot to prefix a “const parameter” with `const`, e.g., in `struct Tagged<TAG: u64>;`. This happens to me from time to time. Maybe C++ devs are also prone to this mistake given template syntax looks like `template<typename T, uint32_t N>`.

…i-obk

Miscellaneous diagnostics cleanups

All found while working on some speculative, invasive changes, but worth doing in their own right.

r? `@oli-obk`

…ted_by_errors, r=oli-obk

Remove `BorrowckErrors::tainted_by_errors`

This PR removes one of the `tainted_by_errors` occurrences, replacing it with direct use of `ErrorGuaranteed`.

r? `@oli-obk`

Remove unnecessary `.to_string()`/`.as_str()`s

hir: Remove the generic type parameter from `MaybeOwned`

It's only ever used with a reference to `OwnerInfo` as an argument.

Follow up to rust-lang#120346.

…ef-field-access, r=compiler-errors

Fix ICE on field access on a tainted type after const-eval failure

Fixes rust-lang#120615.

r? oli-obk or compiler