Don't use usub.with.overflow intrinsic
The canonical form of a usub.with.overflow check in LLVM are separate sub + icmp instructions, rather than a usub.with.overflow intrinsic. Using usub.with.overflow will generally result in worse optimization potential.
The backend will attempt to form usub.with.overflow when it comes to actual instruction selection. This is not fully reliable, but I believe this is a better tradeoff than using the intrinsic in IR.
Fixes#103285.
ci: Bring back ninja for dist builders
The primary reason for this is that make can result in a substantial under utilization of parallelism (noticed while testing on a workstation), mostly due to the submake structure preventing good dependency tracking and scheduling.
In f758c7b2a78 (Debian 6 doesn't have ninja, so use make for the dist builds) llvm.ninja was disabled due to lack of distro package. This is no longer the case with the CentOS 7 base, so bring ninja back for a performance boost.
Clean up tests and add documentation for GATs related stuff
This is a follow-up PR for #13494.
- addresses https://github.com/rust-lang/rust-analyzer/pull/13494#discussion_r1006774897
- documents the ordering constraint on `Binders` and `Substitution` (which is not really follow-up for the previous PR, but it was introduced to support GATs and I strongly feel it's worth it)
Introduce UnordMap, UnordSet, and UnordBag (MCP 533)
This is the start of implementing [MCP 533](https://github.com/rust-lang/compiler-team/issues/533).
I followed `@eddyb's` suggestion of naming the collection types `Unord(Map/Set/Bag)` which is a bit easier to type than `Unordered(Map/Set/Bag)`
r? `@eddyb`
privacy: Rename "accessibility levels" to "effective visibilities"
And a couple of other naming and comment tweaks.
Related to https://github.com/rust-lang/rust/issues/48054
For `enum Level` I initially used naming `enum EffectiveVisibilityLevel`, but it was too long and inconvenient because it's used pretty often.
So I shortened it to just `Level`, if it needs to be used from some context where this name would be ambiguous, then it can be imported with renaming like `use rustc_middle::privacy::Level as EffVisLevel` or something.
poll_fn and Unpin: fix pinning
See [IRLO](https://internals.rust-lang.org/t/surprising-soundness-trouble-around-pollfn/17484) for details: currently `poll_fn` is very subtle to use, since it does not pin the closure, so creating a `Pin::get_unchcked(&mut capture)` inside the closure is unsound. This leads to actual miscompilations with `futures::join!`.
IMO the proper fix is to pin the closure when the future is pinned, which is achieved by changing the `Unpin` implementation. This is a breaking change though. 1.64.0 was *just* released, so maybe this is still okay?
The alternative would be to add some strong comments to the docs saying that closure captures are *not pinned* and doing `Pin::get_unchecked` on them is unsound.
Fix line numbers for MIR inlined code
`should_collapse_debuginfo` detects if the specified span is part of a
macro expansion however it does this by checking if the span is anything
other than a normal (non-expanded) kind, then the span sequence is
walked backwards to the root span.
This doesn't work when the MIR inliner inlines code as it creates spans
with expansion information set to `ExprKind::Inlined` and results in the
line number being attributed to the inline callsite rather than the
normal line number of the inlined code.
Fixes#103068
feat: type inference for generic associated types
This PR implements type inference for generic associated types. Basically, this PR lowers generic arguments for associated types in valid places and creates `Substitution`s for them.
I focused on the inference for correct Rust programs, so there are cases where we *accidentally* manage to infer things that are actually invalid (which would then be reported by flycheck so I deem them non-fatal). See the following tests and FIXME notes on them: `gats_with_dyn`, `gats_with_impl_trait`.
The added tests are rather arbitrary. Let me know if there are cases I'm missing or I should add.
Closes#9673
Remove `commit_if_ok` probe from NLL type relation
It was not really necessary to add the `commit_if_ok` in #100092 -- I added it to protect us against weird inference error messages due to recursive RPIT calls, but we are always on the error path when this happens anyways, and I can't come up with an example that makes this manifest.
Fixes#103599
r? `@oli-obk` since you reviewed #100092, feel free to re-roll.
🅱️📢 beta-nominating this since it's on beta (which forks in ~a week~ two days 😨) -- worst case we could revert the original PR on beta and land this on nightly, to give it some extra soak time...
Add Target Tier Policy notification.
This adds a notification posted to PRs when they add/modify a target spec.
This was hard-coded in highfive. I forgot to include this in https://github.com/rust-lang/rust/pull/103492.
Switch to upstream `positionEncoding`
Closes#13481
This drops support for the custom extension, but that's probably fine.
Draft because it's not tested yet.
fix: Test all generic args for trait when finding matching impl
Addresses https://github.com/rust-lang/rust-analyzer/pull/13463#issuecomment-1287816680
When finding matching impl for a trait method, we've been testing the unifiability of self type. However, there can be multiple impl of a trait for the same type with different generic arguments for the trait. This patch takes it into account and tests the unifiability of all type arguments for the trait (the first being the self type) thus enables rust-analyzer to find the correct impl even in such cases.
