We've already removed non-sysroot proc macro server, which effectively
removed support for Rust <1.64.0, so this removal of fallback path
shouldn't be problem at this point.
More APIs for `la_arena::IdxRange`
```rust
impl<T> ExactSizeIterator for IdxRange<T>;
impl<T> Arena<T> {
pub fn alloc_many<II: IntoIterator<Item = T>>(&mut self, iter: II) -> IdxRange<T>;
}
```
1. There are no currently ways to get `IdxRange` without manually offseting `Idx`. Providing a method for multiple-allocation simplifies this process and makes it less error-prone.
2. `IdxRange: ExactSizeIterator` makes `iter.zip(range).rev()` possible. Since `Zip: DoubleEndedIterator` requires all its arguments to be `ExactSizeIterator`. It also ease the usage for, eg. `len()`.
3. Fixed a typo.
I noticed that `IdxRange::end` may be invalid. Is it good to return `Idx` instead of `RawIdx`?
Fix libs publish branch filter
line-index didn't actually get published from #14733, probably because the branch filter was for main but the main branch is called master here. This fixes the workflow file
I also tweaked the libs readme mostly just so the paths filter would pick up the changes.
Make line-index a lib, use nohash_hasher
These seem like they are not specific to rust-analyzer and could be pulled out to their own libraries. So I did.
https://github.com/azdavis/millet/issues/31
fix: ignore impls with `#[rustc_reservation_impl]`
Fixes#12247Fixes#14279
Currently core has two blanket impls for `From`: `impl<T> From<T> for T` and `impl<T> From<!> for T`. These are conflicting and thus chalk cannot uniquely solve `S: From<?0>` for any type `S`.
The latter impl is actually a reservation impl and should not be considered during trait selection. More generally, impls attributed with perma-unstable `#[rustc_reservation_impl]` attribute should be disregarded except for coherence checks. See rust-lang/rust#64631 and rust-lang/rust#64715 for details.
I chose to entirely ignore them in hir-ty because we don't do coherence checks.
