Rollup of 11 pull requests - #161769
Closed
JonathanBrouwer wants to merge 51 commits into
Closed
Conversation
Generalize the recovery to match arguments positionally against the
generic parameters, skipping those that need no explicit argument:
elided early-bound lifetimes, synthetic `impl Trait` type parameters,
and trailing type parameters with defaults. Bail only when there are
more arguments than can be matched.
Omitting such a parameter leaves fewer arguments than parameters, so the
exact-count check made recovery feed `{type error}` for the anon const
while the real lowering later fed the true type, tripping a double-feed
ICE under the `-Zthreads=0` front-end race.
Each exercises the early `type_of` query under the `-Zthreads=0` front-end race with an omittable generic parameter: a synthetic `impl Trait` param, an elided early-bound lifetime, multiple explicit args, and a defaulted trailing type parameter.
`needs_coroutine_by_move_body_def_id` asks for `type_of`, and for a body owner nested inside a const argument's anon const that goes through `typeck` of the anon const, which needs the anon const's own type. That type is never computed, only fed while the enclosing body is type-checked. Asking for it in the same pass that type-checks the bodies lets the parallel front end reach the nested body owner first, computing and caching an error type for the anon const that then conflicts with the type fed later on. Since nothing reports an error in that case, the delayed bugs surface as an ICE. Split it into a second pass over the body owners so every body has been type-checked, and every const argument lowered, before any nested body owner is asked for its type.
Recovering the type from the HIR path meant re-deriving what the generic argument lowering already works out, and only covered the cases the partial classifier recognized: free function paths resolved to `DefKind::Fn`, with no late-bound lifetimes. Anything else still cached an error type and tripped the double-feed ICE, including associated functions, tuple struct constructors, inherent type-relative paths, and const arguments in type annotations, which are not path expressions at all and so cannot be classified this way. With the by-move bodies generated after typeck the anon const's type is always fed before anything asks for it, so drop the recovery along with the tests that only exercised its argument counting. The shapes they covered are folded into the main regression test.
A naive `f32::mul_add(a as f32, b as f32, c as f32) as f16` has insufficient precision
update dependencies
…when-reliable run `f16` and `f128` tests natively when reliable
port all variadic functions to strict signature checking
These non-obvious things are worth documenting.
run _Unwind_RaiseException test on Windows
unify shim_sig and shim_sig_variadic macros
add internal DSL for testing binders implements rust-lang/project-assumptions-on-binders#8 apologies to the inevitable swarm of people who get pinged for this... basically the only interesting diff is `compiler/rustc_hir_analysis/src/check/wfcheck.rs`, which actually implements the test. there's probably still features we want to add to this DSL (e.g. `RegionConstraint::AliasTyOutlivesViaEnv`), and there might be some bugs lurking, but this is at least a base to work off of. It's perma-unstable and for internal use only, so support and quality doesn't have to be incredibly high. For example, parser recovery and whatnot is just, bad, but whatever. r? BoxyUwU
…=nia-e Add custom allocators to `(try_)map` on `Box`, `Rc`, `Arc` Adds custom allocators to `(try_)map` on `Box`, `Rc`, `Arc`. Not on `UniqueArc`/`UniqueRc` because too much allocator-related API is missing atm and I didn't want to end up writing unsound garbage. The `else` branch in those needs to drop a weakref, I'll get to that later. cc rust-lang#160534 r? nia-e
Fix debugger visualizer tuple child ordering w/ PDB debug info see: llvm/llvm-project#218613 for an explanation of the issue. tl;dr, PDB fields reflect rustc's reordering, whereas DWARF fields are in source-code-order. This causes issues with `GetChildAtIndex` in cases where both DWARF- and PDB-based types can interact with them. This PR uses field names to retrieve field indexes, rather than blindly indexing into the field array. I did a quick scan over the rest of the `GetChildAtIndex` calls, and they all fall under a few safe cases 1. in DWARF or PDB specific providers (mainly enum discriminants, which can't be reordered anyway) 2. used to get field `0` of a struct with only 1 field 3. in the summary provider where we're typically already interacting with the synthetic value and synthetic children anyway cc rust-lang#161657, fixes `tests/debuginfo/tuple-struct.rs` on `windows-msvc` <details><summary>`windows-msvc` `tests/debuginfo` results with this change</summary> ``` failures: [debuginfo-lldb] tests\debuginfo\associated-types.rs [debuginfo-lldb] tests\debuginfo\borrowed-basic.rs [debuginfo-lldb] tests\debuginfo\borrowed-unique-basic.rs [debuginfo-lldb] tests\debuginfo\basic-types-globals.rs#no-lto [debuginfo-lldb] tests\debuginfo\borrowed-tuple.rs [debuginfo-lldb] tests\debuginfo\boxed-struct.rs [debuginfo-lldb] tests\debuginfo\captured-fields-2.rs [debuginfo-lldb] tests\debuginfo\cross-crate-spans.rs [debuginfo-lldb] tests\debuginfo\coroutine-locals.rs [debuginfo-lldb] tests\debuginfo\box.rs [debuginfo-lldb] tests\debuginfo\captured-fields-1.rs [debuginfo-lldb] tests\debuginfo\c-style-enum-in-composite.rs [debuginfo-lldb] tests\debuginfo\by-value-self-argument-in-trait-impl.rs [debuginfo-lldb] tests\debuginfo\destructured-for-loop-variable.rs [debuginfo-lldb] tests\debuginfo\dummy_span.rs [debuginfo-lldb] tests\debuginfo\evec-in-struct.rs [debuginfo-lldb] tests\debuginfo\destructured-fn-argument.rs [debuginfo-lldb] tests\debuginfo\destructured-local.rs [debuginfo-lldb] tests\debuginfo\generic-struct.rs [debuginfo-lldb] tests\debuginfo\issue-22656.rs [debuginfo-lldb] tests\debuginfo\basic-types-globals.rs#lto [debuginfo-lldb] tests\debuginfo\lexical-scope-in-if.rs [debuginfo-lldb] tests\debuginfo\lexical-scope-in-stack-closure.rs [debuginfo-lldb] tests\debuginfo\lexical-scope-in-unconditional-loop.rs [debuginfo-lldb] tests\debuginfo\lexical-scope-with-macro.rs [debuginfo-lldb] tests\debuginfo\lexical-scope-in-unique-closure.rs [debuginfo-lldb] tests\debuginfo\lexical-scope-in-if-let.rs [debuginfo-lldb] tests\debuginfo\lexical-scopes-in-block-expression.rs [debuginfo-lldb] tests\debuginfo\method-on-generic-struct.rs [debuginfo-lldb] tests\debuginfo\lexical-scope-in-while.rs [debuginfo-lldb] tests\debuginfo\lexical-scope-in-for-loop.rs [debuginfo-lldb] tests\debuginfo\name-shadowing-and-scope-nesting.rs [debuginfo-lldb] tests\debuginfo\no_mangle-info.rs [debuginfo-lldb] tests\debuginfo\packed-struct-with-destructor.rs [debuginfo-lldb] tests\debuginfo\pretty-slices.rs [debuginfo-lldb] tests\debuginfo\packed-struct.rs [debuginfo-lldb] tests\debuginfo\reference-debuginfo.rs [debuginfo-lldb] tests\debuginfo\simple-struct.rs [debuginfo-lldb] tests\debuginfo\simple-lexical-scope.rs [debuginfo-lldb] tests\debuginfo\shadowed-variable.rs [debuginfo-lldb] tests\debuginfo\simple-tuple.rs [debuginfo-lldb] tests\debuginfo\shadowed-argument.rs [debuginfo-lldb] tests\debuginfo\struct-with-destructor.rs [debuginfo-lldb] tests\debuginfo\strings-and-strs.rs [debuginfo-lldb] tests\debuginfo\struct-in-struct.rs [debuginfo-lldb] tests\debuginfo\pretty-std-collections.rs [debuginfo-lldb] tests\debuginfo\tuple-in-tuple.rs [debuginfo-lldb] tests\debuginfo\thread-names.rs#win [debuginfo-lldb] tests\debuginfo\vec-slices.rs [debuginfo-lldb] tests\debuginfo\var-captured-in-sendable-closure.rs [debuginfo-lldb] tests\debuginfo\var-captured-in-stack-closure.rs [debuginfo-lldb] tests\debuginfo\var-captured-in-nested-closure.rs test result: FAILED. 244 passed; 52 failed; 223 ignored; 0 measured; 0 filtered out; finished in 13.22s ``` </details>
miri subtree update I want to add proper vararg checks for Miri's shims but it turns out I'll have to adjust the rustc side for this. It'll be easier to just do the entire thing as a rustc patch, but that needs a sync first. --- Subtree update of `miri` to rust-lang/miri@40000c5. Created using https://github.com/rust-lang/josh-sync. r? @ghost
make trivial ABI check resilient against new repr There's currently no way to exhaustively match on "what basic `repr` is this", which is kind of fragile as it means a new repr will get whatever the `else` branch happens to be everywhere. Changing this would be a major refactor, but let's at least make it possible to `assert!` that we covered all the cases by adding an explicit way to check for `repr(Rust)`.
…ailable-suggestion, r=mu001999
Detect missing binding available: add a MaybeIncorrect suggestion
When suggesting that a missing binding is available in a pattern but not used, if it is behind a `..` suggest a MaybeIncorrect fix replacing `..` with `{binding}, ..`
This suggestion could be a `tool_only_suggestion`, or a `span_suggestion_hidden`. Other types of suggestions seem a bit too verbose for this simple case.
The `tool_only_suggestion` (first commit) does not add any suggestion in stderr, the `span_suggestion_hidden` (second commit) shows a minimal help line.
This work was extracted from PR rust-lang#156239.
…nded, r=tgross35 test `f16::mul_add` not double-rounding the result test the precision of `f16::mul_add`. The semantics of `mul_add` are that there should only be one rounding of the final result back into the storage type, i.e. the intermediate result of the multiplication should not be rounded. A naive implementation of `f16::mul_add` as `f32::mul_add(a as f32, b as f32, c as f32) as f16` has insufficient precision, see llvm/llvm-project#98389. Our `rustc_codegen_gcc` backend still used the `f32` approach, this PR changes it to instead (implicitly) use the fallback body from `core`, which uses `f64::mul_add` so that the result is correctly rounded.
…ts, r=jdonszelmann Add two comments relating to new-solver performance These non-obvious things are worth documenting. r? @jdonszelmann
Add codegen test for static table search loop unrolling Adds a test for reported case in rust-lang#44041. The follow up case in that issue is still open so this does not close it
…=Kobzol do not compress debuginfo for Cygwin when I updated Rust for MSYS, I got errors from gcc regarding -Wl,--compress-debug-sections=zlib flag being unsupported. so don't add it for all Cygwin hosts r? @Kobzol
rustbot
added
A-attributes
Member
Author
|
@bors r+ p=5 |
Contributor
rust-bors
Bot
added
S-waiting-on-bors
rust-bors
Bot
added
S-waiting-on-author
Contributor
|
This pull request was unapproved due to being closed. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
12 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Successful merges:
(try_)maponBox,Rc,Arc#161617 (Add custom allocators to(try_)maponBox,Rc,Arc)f16::mul_addnot double-rounding the result #161522 (testf16::mul_addnot double-rounding the result)r? @ghost
Create a similar rollup