Fix async fn lowering ICE with APIT.
Fixes#60518.
This PR fixes an ICE where simple bindings (which aren't replaced with replacement arguments during async fn lowering) were not being visited in the def collector and thus caused an ICE during HIR lowering for types that use their `DefId` at that point - such as `impl Trait`.
r? @cramertj
Add rustfmt toml
This commit adds an rustfmt.toml for using for **new** code.
Old code should continut to use old style, until we put automated
style checks in place.
See
https://internals.rust-lang.org/t/running-rustfmt-on-rust-lang-rust-and-other-rust-lang-repositories/8732/81
for the reason why we deviate from the default formatting. The TL;DR
is that currently compiler uses a pretty condensed style of code, and
default settings both create a huge diff and inflate the number of
lines. use_small_heuristics=Max fixes that.
version=Two is required for bug-fixes, which technically can't be made
to the stable first version
cc @Centril: I guess adding config file with agreed-upon formatting doesn't hurt :-)
EDIT: Zullip discussion: https://rust-lang.zulipchat.com/#narrow/stream/131828-t-compiler/topic/x.2Epy.20fmt
Remove -Z borrowck=compare flag
This is the start of the work that needs to be done on #59193. It just removes the flag and updates
the tests.
r? @matthewjasper
Place related refactors
Meanwhile I was working on Place 2 I'm finding some little things that I had on my branch but preferred to land a separate PR for things that are simpler to merge.
r? @oli-obk
Constrain all regions in the concrete type for an opaque type
`push_outlives_components` skips some regions in a type, notably the signature
of a closure is ignored. Most of the time this is OK, but for opaque types the
concrete type is used when checking auto-trait bounds in other functions.
cc @nikomatsakis @pnkfelix
Closes#57464Closes#60127
Account for paths in incorrect pub qualifier help
Handle case where incorrect pub qualifier with a mod path is used and provide the same help given for all other incorrect qualifiers by making the `pub(crate)` parse check more specific.
This distinction is fairly abstract, but in practice, the main advantage
here is that LLVM's triple code considers WASI to be an OS, so this
makes rustc agree with that.
Multi-variant layouts for generators
This allows generators to overlap fields using variants, but doesn't do any such overlapping yet. It creates one variant for every state of the generator (unresumed, returned, panicked, plus one for every yield), and puts every stored local in each of the yield-point variants.
Required for optimizing generator layouts (#52924).
There was quite a lot of refactoring needed for this change. I've done my best in later commits to eliminate assumptions in the code that only certain kinds of types are multi-variant, and to centralize knowledge of the inner mechanics of generators in as few places as possible.
This change also emits debuginfo about the fields contained in each variant, as well as preserving debuginfo about stored locals while running in the generator.
Also, fixes#59972.
Future work:
- Use this change for an optimization pass that actually overlaps locals within the generator struct (#52924)
- In the type layout fields, don't include locals that are uninitialized for a particular variant, so miri and UB sanitizers can check our memory (see https://github.com/rust-lang/rust/issues/59972#issuecomment-483058172)
- Preserve debuginfo scopes across generator yield points
rustc: factor out most of hir::def::Def's variants into DefKind, and rename to Res.
The first two commits are about introducing `DefKind`, both to simplify/orthogonalize `hir::def::Def`, and to allow reasoning about the kind of a definition without dealing with the redundant `DefId`.
(There are likely more changes to be made, such as adding enough `DefKind` variants for `tcx.def_kind(def_id)` to return just `DefKind`, not `Option<DefKind>`, but this is pretty big as-is)
The third commit frees up the `Def` name (which we may want to use in the future for "definitions", in the sense of "entities with a `DefId`") by renaming `hir::def::Def` to `Res` ("resolution").
IMO this fits, as it represents all the possible name resolution results, not just "definitions (with a `DefId`)".
Quick examples:
```rust
// Before:
if tcx.describe_def(def_id) == Some(Def::Struct(def_id)) {...}
if let Def::Struct(def_id) = path.def {...}
```
```rust
// After:
if tcx.def_kind(def_id) == Some(DefKind::Struct) {...}
if let Res::Def(DefKind::Struct, def_id) = path.res {...}
```
r? @petrochenkov cc @rust-lang/compiler
This commit fixes an ICE where simple bindings (which aren't replaced
with replacement arguments during async fn lowering) were not being
visited in the def collector and thus caused an ICE during HIR lowering
for types that use their `DefId` at that point - such as `impl Trait`.
Currently when linking an artifact rustc will only conditionally call
the `Linker::export_symbols` function, but this causes issues on some
targets, like WebAssembly, where it means that executable outputs will
not have the same symbols exported that cdylib outputs have. This commit
sinks the conditional call to `export_symbols` inside the various
implementations of the function that still need it, and otherwise the
wasm linker is configured to always pass through symbol visibility
lists.
Fix potential integer overflow in SGX memory range calculation.
Thanks to Eduard Marin and David Oswald at the University of Burmingham, and Jo Van Bulck at KU Leuven for discovering this issue.
This commit adds an rustfmt.toml for using for **new** code.
Old code should continut to use old style, until we put automated
style checks in place.
See
https://internals.rust-lang.org/t/running-rustfmt-on-rust-lang-rust-and-other-rust-lang-repositories/8732/81
for the reason why we deviate from the default formatting. The TL;DR
is that currently compiler uses a pretty condensed style of code, and
default settings both create a huge diff and inflate the number of
lines. use_small_heuristics=Max fixes that.
version=Two is required for bug-fixes, which technically can't be made
to the stable first version
Rollup of 12 pull requests
Successful merges:
- #59928 (Make deprecation lint `ambiguous_associated_items` deny-by-default)
- #60220 (report fatal errors during doctest parsing)
- #60373 (Tidy: ensure lang features are sorted by since)
- #60388 (Disallow non-explicit elided lifetimes in async fn)
- #60393 ( Do not suggest incorrect syntax on pattern type error due to borrow)
- #60401 (Rename `RUST_LOG` to `RUSTC_LOG`)
- #60409 (Require a trait in the bounds of existential types)
- #60455 (Resolve match arm ty when arms diverge)
- #60457 (Const prop refactoring)
- #60467 (Avoid repeated interning of static strings.)
- #60478 (minor compiler doc tweaks)
- #60501 (Propagate mutability from arguments to local bindings in async fn)
Failed merges:
r? @ghost
Avoid repeated interning of static strings.
`file_metadata_raw` interns the strings `"<unknown>"` and `""` very
frequently. This commit avoids that, which reduces the number of symbols
interned significantly and reduces instruction counts by up to 0.5% on
some workloads.
Const prop refactoring
This is rebased on top of #60428 so only the top commit is new.
This is the refactoring to remove the `mir` field from `ConstPropagator` which is necessary before we can begin to actually propagate constants.
r? @oli-obk
