Change type of async context parameter after state transform. by cjgillot · Pull Request #157166 · rust-lang/rust

cjgillot · 2026-05-30T17:36:29Z

Future::poll expects a &mut std::task::Context<'_>. Meanwhile, async coroutines use std::future::ResumeTy as resume parameter. This is meant to workaround the limitations of borrowck, which cannot prove that coroutines implement for<'a, 'b> CoroutineTrait<&'a mut Context<'b>>.

In the coroutine state transform, we need to change the signature from ResumeTy to the proper &mut Context<'_>. The current code attempts to find locals that have type ResumeTy to change their type. This is needlessly complex and relies on undocumented behaviour of the MIR builder.

Instead, this PR proposes to replace the ResumeTy argument with a new local, with value ResumeTy(transmute(context)).

Based on #156875.

rustbot · 2026-05-30T17:36:32Z

This PR changes MIR

cc @oli-obk, @RalfJung, @JakobDegen, @vakaras

Some changes occurred to MIR optimizations

cc @rust-lang/wg-mir-opt

rustbot · 2026-05-30T17:36:35Z

r? @tiif

rustbot has assigned @tiif.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

Owners of files modified in this PR: compiler
compiler expanded to 73 candidates
Random selection from 17 candidates

rustbot · 2026-06-04T12:12:55Z

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

oli-obk · 2026-06-09T15:36:02Z

+/// HIR uses `get_context` to unwrap a `&mut Context<'_>` from a `ResumeTy`.
+/// Both types are just a single pointer, but liveness analysis does not know that and
+/// supposes that the operand and the destination are live at the same time.
+/// Forcibly inline those calls to avoid this.


preexisting, but: it feels a bit funny to generate a function call in ast lowering only to desugar it here manually. The function body is effectively dead code.

View changes since the review

oli-obk

@bors r+

View changes since this review

rust-bors · 2026-06-09T15:36:44Z

📌 Commit b7e26ac has been approved by oli-obk

It is now in the queue for this repository.

Change type of async context parameter after state transform. `Future::poll` expects a `&mut std::task::Context<'_>`. Meanwhile, async coroutines use `std::future::ResumeTy` as resume parameter. This is meant to workaround the limitations of borrowck, which cannot prove that coroutines implement `for<'a, 'b> CoroutineTrait<&'a mut Context<'b>>`. In the coroutine state transform, we need to change the signature from `ResumeTy` to the proper `&mut Context<'_>`. The current code attempts to find locals that have type `ResumeTy` to change their type. This is needlessly complex and relies on undocumented behaviour of the MIR builder. Instead, this PR proposes to replace the `ResumeTy` argument with a new local, with value `ResumeTy(transmute(context))`. Based on #156875.

jhpratt · 2026-06-09T21:51:41Z

yielding to a rollup=never PR

@bors yield

rust-bors · 2026-06-09T21:51:44Z

Auto build was cancelled. Cancelled workflows:

https://github.com/rust-lang/rust/actions/runs/27238086714

The next pull request likely to be tested is #129543.

Rollup of 17 pull requests Successful merges: - #157166 (Change type of async context parameter after state transform.) - #157335 (bootstrap: Handle dotted table keys when parsing bootstrap.toml) - #157503 (Disable `tests/debuginfo/pretty-std.rs` `OsString` cdb check) - #157571 (Remove ProcMacro enum from proc macro ABI) - #148183 (rustdoc: Test & document `test_harness` code block attribute) - #153847 (Fix marker trait winnowing depending on impl order) - #156067 (Fix async drop glue for Box<T>) - #156399 (fix improper ctypes in Znext solver) - #157338 (Make `Literal::byte_character_value` work with bytes as well) - #157410 (Implement rustc_public::CrateDef{,Type} for FieldDef) - #157605 (Arg splat experiment - syntax impl) - #157630 (Add multibyte JSON diagnostic regression test) - #157633 (Reorder `impl` restriction rendering and add bottom margin) - #157642 (Report duplicate relaxed bounds during ast lowering) - #157652 (fix doc for unicode normalization faq on `casefold` APIs) - #157661 (Update to ar_archive_writer v0.5.2) - #157668 (Add test for matches in `rustc_must_match_exhaustively`) Failed merges: - #157670 (Rename `errors.rs` file to `diagnostics.rs` (4/N))

rust-bors · 2026-06-10T02:25:28Z

⌛ Testing commit b7e26ac with merge 39465ed...

Workflow: https://github.com/rust-lang/rust/actions/runs/27248950736

Change type of async context parameter after state transform. `Future::poll` expects a `&mut std::task::Context<'_>`. Meanwhile, async coroutines use `std::future::ResumeTy` as resume parameter. This is meant to workaround the limitations of borrowck, which cannot prove that coroutines implement `for<'a, 'b> CoroutineTrait<&'a mut Context<'b>>`. In the coroutine state transform, we need to change the signature from `ResumeTy` to the proper `&mut Context<'_>`. The current code attempts to find locals that have type `ResumeTy` to change their type. This is needlessly complex and relies on undocumented behaviour of the MIR builder. Instead, this PR proposes to replace the `ResumeTy` argument with a new local, with value `ResumeTy(transmute(context))`. Based on #156875.

jhpratt · 2026-06-10T02:31:08Z

same thing

@bors yield

rust-bors · 2026-06-10T02:31:11Z

Auto build was cancelled. Cancelled workflows:

https://github.com/rust-lang/rust/actions/runs/27248950736

The next pull request likely to be tested is #157683.

Rollup of 17 pull requests Successful merges: - #157166 (Change type of async context parameter after state transform.) - #157335 (bootstrap: Handle dotted table keys when parsing bootstrap.toml) - #157503 (Disable `tests/debuginfo/pretty-std.rs` `OsString` cdb check) - #157571 (Remove ProcMacro enum from proc macro ABI) - #148183 (rustdoc: Test & document `test_harness` code block attribute) - #153847 (Fix marker trait winnowing depending on impl order) - #156067 (Fix async drop glue for Box<T>) - #156399 (fix improper ctypes in Znext solver) - #157338 (Make `Literal::byte_character_value` work with bytes as well) - #157410 (Implement rustc_public::CrateDef{,Type} for FieldDef) - #157605 (Arg splat experiment - syntax impl) - #157630 (Add multibyte JSON diagnostic regression test) - #157633 (Reorder `impl` restriction rendering and add bottom margin) - #157642 (Report duplicate relaxed bounds during ast lowering) - #157652 (fix doc for unicode normalization faq on `casefold` APIs) - #157661 (Update to ar_archive_writer v0.5.2) - #157668 (Add test for matches in `rustc_must_match_exhaustively`) Failed merges: - #157670 (Rename `errors.rs` file to `diagnostics.rs` (4/N))

rust-bors · 2026-06-10T02:37:33Z

⌛ Testing commit b7e26ac with merge e1e4985...

Workflow: https://github.com/rust-lang/rust/actions/runs/27249369474

Change type of async context parameter after state transform. `Future::poll` expects a `&mut std::task::Context<'_>`. Meanwhile, async coroutines use `std::future::ResumeTy` as resume parameter. This is meant to workaround the limitations of borrowck, which cannot prove that coroutines implement `for<'a, 'b> CoroutineTrait<&'a mut Context<'b>>`. In the coroutine state transform, we need to change the signature from `ResumeTy` to the proper `&mut Context<'_>`. The current code attempts to find locals that have type `ResumeTy` to change their type. This is needlessly complex and relies on undocumented behaviour of the MIR builder. Instead, this PR proposes to replace the `ResumeTy` argument with a new local, with value `ResumeTy(transmute(context))`. Based on #156875.

jhpratt · 2026-06-10T02:42:45Z

repeated network failures; I've closed the tree elsewhere

@bors yield

rust-bors · 2026-06-10T02:42:48Z

Auto build was cancelled. Cancelled workflows:

https://github.com/rust-lang/rust/actions/runs/27249369474

The next pull request likely to be tested is #157683.

Rollup of 17 pull requests Successful merges: - #157166 (Change type of async context parameter after state transform.) - #157335 (bootstrap: Handle dotted table keys when parsing bootstrap.toml) - #157503 (Disable `tests/debuginfo/pretty-std.rs` `OsString` cdb check) - #157571 (Remove ProcMacro enum from proc macro ABI) - #148183 (rustdoc: Test & document `test_harness` code block attribute) - #153847 (Fix marker trait winnowing depending on impl order) - #156067 (Fix async drop glue for Box<T>) - #156399 (fix improper ctypes in Znext solver) - #157338 (Make `Literal::byte_character_value` work with bytes as well) - #157410 (Implement rustc_public::CrateDef{,Type} for FieldDef) - #157605 (Arg splat experiment - syntax impl) - #157630 (Add multibyte JSON diagnostic regression test) - #157633 (Reorder `impl` restriction rendering and add bottom margin) - #157642 (Report duplicate relaxed bounds during ast lowering) - #157652 (fix doc for unicode normalization faq on `casefold` APIs) - #157661 (Update to ar_archive_writer v0.5.2) - #157668 (Add test for matches in `rustc_must_match_exhaustively`) Failed merges: - #157670 (Rename `errors.rs` file to `diagnostics.rs` (4/N))

Rollup merge of #157166 - cjgillot:retype-context-after, r=oli-obk Change type of async context parameter after state transform. `Future::poll` expects a `&mut std::task::Context<'_>`. Meanwhile, async coroutines use `std::future::ResumeTy` as resume parameter. This is meant to workaround the limitations of borrowck, which cannot prove that coroutines implement `for<'a, 'b> CoroutineTrait<&'a mut Context<'b>>`. In the coroutine state transform, we need to change the signature from `ResumeTy` to the proper `&mut Context<'_>`. The current code attempts to find locals that have type `ResumeTy` to change their type. This is needlessly complex and relies on undocumented behaviour of the MIR builder. Instead, this PR proposes to replace the `ResumeTy` argument with a new local, with value `ResumeTy(transmute(context))`. Based on #156875.

Rollup of 17 pull requests Successful merges: - rust-lang/rust#157166 (Change type of async context parameter after state transform.) - rust-lang/rust#157335 (bootstrap: Handle dotted table keys when parsing bootstrap.toml) - rust-lang/rust#157503 (Disable `tests/debuginfo/pretty-std.rs` `OsString` cdb check) - rust-lang/rust#157571 (Remove ProcMacro enum from proc macro ABI) - rust-lang/rust#148183 (rustdoc: Test & document `test_harness` code block attribute) - rust-lang/rust#153847 (Fix marker trait winnowing depending on impl order) - rust-lang/rust#156067 (Fix async drop glue for Box<T>) - rust-lang/rust#156399 (fix improper ctypes in Znext solver) - rust-lang/rust#157338 (Make `Literal::byte_character_value` work with bytes as well) - rust-lang/rust#157410 (Implement rustc_public::CrateDef{,Type} for FieldDef) - rust-lang/rust#157605 (Arg splat experiment - syntax impl) - rust-lang/rust#157630 (Add multibyte JSON diagnostic regression test) - rust-lang/rust#157633 (Reorder `impl` restriction rendering and add bottom margin) - rust-lang/rust#157642 (Report duplicate relaxed bounds during ast lowering) - rust-lang/rust#157652 (fix doc for unicode normalization faq on `casefold` APIs) - rust-lang/rust#157661 (Update to ar_archive_writer v0.5.2) - rust-lang/rust#157668 (Add test for matches in `rustc_must_match_exhaustively`) Failed merges: - rust-lang/rust#157670 (Rename `errors.rs` file to `diagnostics.rs` (4/N))

JonathanBrouwer · 2026-06-13T10:42:25Z

@rust-timer build 0b580da

rust-timer · 2026-06-13T16:41:57Z

Finished benchmarking commit (0b580da): comparison URL.

Overall result: ❌ regressions - please read:

Benchmarking means the PR may be perf-sensitive. It's automatically marked not fit for rolling up. Overriding is possible but disadvised: it risks changing compiler perf.

Next, please: If you can, justify the regressions found in this try perf run in writing along with @rustbot label: +perf-regression-triaged. If not, fix the regressions and do another perf run. Neutral or positive results will clear the label automatically.

@bors rollup=never
@rustbot label: -S-waiting-on-perf +perf-regression

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	0.6%	[0.2%, 1.3%]	17
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Max RSS (memory usage)

This perf run didn't have relevant results for this metric.

Cycles

Results (secondary 2.7%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	2.7%	[2.7%, 2.7%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Binary size

Results (primary 0.1%, secondary 3.2%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	0.1%	[0.1%, 0.1%]	4
Regressions ❌ (secondary)	3.2%	[0.4%, 5.5%]	14
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.1%	[0.1%, 0.1%]	4

Bootstrap: 525.47s -> 517.162s (-1.58%)
Artifact size: 400.80 MiB -> 400.89 MiB (0.02%)

JonathanBrouwer · 2026-06-13T16:46:16Z

Caused the perf regressions in #157683
Is there anything we can do about this?

@oli-obk

Desugar async blocks in HIR instead of MIR Implements MCP rust-lang/compiler-team#997 Based on #157166 In the current implementation, `gen`/`async`/`async gen` blocks and closures have type `Coroutine(..)` and `CoroutineClosure(..)`. Those types implement `Iterator`, `Future` or `AsyncIterator` depending on the initial desugaring. This creates a lot of complexity: - trait solvers must check which kind of coroutine each time; - MIR StateTransform needs to fixup types depending on the coroutine kind. I propose to change the desugaring for coroutines to: - `gen { .. }` becomes `CoroutineIterator::from_coroutine(#[coroutine] { .. })`; - `async { .. }` becomes `CoroutineFuture::from_coroutine(#[coroutine] { .. })`; - `async gen { .. }` becomes `CoroutineAsyncIterator::from_coroutine(#[coroutine] { .. })`. This way, all coroutines implement `std::ops::Coroutine` and `core` is responsible for translating this to user-friendly traits. All the complexity is pushed to error-reporting code, which is not soundness-critical. Coroutine closures are a little more complex, as we need to keep the `CoroutineClosure` type for borrow-checking. Main design point: I create two methods on `TyCtxt` that are meant to do the back-and-forth between wrapped and unwrapped coroutines. `coroutine_desugared_type` wraps a coroutine inside the adapter struct. `try_unwrap_desugared_coroutine` unwraps it. r? @oli-obk cc @lcnr @RalfJung cc @estebank as I modify quite a lot of diagnostic code

@oli-obk

Desugar async blocks in HIR instead of MIR Implements MCP rust-lang/compiler-team#997 Based on #157166 In the current implementation, `gen`/`async`/`async gen` blocks and closures have type `Coroutine(..)` and `CoroutineClosure(..)`. Those types implement `Iterator`, `Future` or `AsyncIterator` depending on the initial desugaring. This creates a lot of complexity: - trait solvers must check which kind of coroutine each time; - MIR StateTransform needs to fixup types depending on the coroutine kind. I propose to change the desugaring for coroutines to: - `gen { .. }` becomes `CoroutineIterator::from_coroutine(#[coroutine] { .. })`; - `async { .. }` becomes `CoroutineFuture::from_coroutine(#[coroutine] { .. })`; - `async gen { .. }` becomes `CoroutineAsyncIterator::from_coroutine(#[coroutine] { .. })`. This way, all coroutines implement `std::ops::Coroutine` and `core` is responsible for translating this to user-friendly traits. All the complexity is pushed to error-reporting code, which is not soundness-critical. Coroutine closures are a little more complex, as we need to keep the `CoroutineClosure` type for borrow-checking. Main design point: I create two methods on `TyCtxt` that are meant to do the back-and-forth between wrapped and unwrapped coroutines. `coroutine_desugared_type` wraps a coroutine inside the adapter struct. `try_unwrap_desugared_coroutine` unwraps it. r? @oli-obk cc @lcnr @RalfJung cc @estebank as I modify quite a lot of diagnostic code

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels May 30, 2026

rustbot assigned tiif May 30, 2026

This comment has been minimized.

Sign in to view

cjgillot force-pushed the retype-context-after branch from 8447533 to 8c16d8a Compare May 30, 2026 17:54

This comment has been minimized.

Sign in to view

cjgillot force-pushed the retype-context-after branch from 8c16d8a to ab5076b Compare May 30, 2026 21:12

This comment has been minimized.

Sign in to view

cjgillot force-pushed the retype-context-after branch from ab5076b to 769048f Compare May 31, 2026 08:02

This comment has been minimized.

Sign in to view

cjgillot force-pushed the retype-context-after branch from 769048f to 89aaf42 Compare May 31, 2026 09:46

This comment has been minimized.

Sign in to view

cjgillot force-pushed the retype-context-after branch from 89aaf42 to 93879f5 Compare June 1, 2026 21:36

This comment has been minimized.

Sign in to view

cjgillot mentioned this pull request Jun 2, 2026

Desugar async blocks in HIR instead of MIR #157309

Open

oli-obk assigned oli-obk and unassigned tiif Jun 2, 2026

This comment has been minimized.

Sign in to view

cjgillot force-pushed the retype-context-after branch from 93879f5 to 323effe Compare June 3, 2026 21:40

This comment has been minimized.

Sign in to view

cjgillot added 2 commits June 4, 2026 08:10

Change type of async context parameter after state transform.

f60cb02

Bless type sizes tests.

b7e26ac

cjgillot force-pushed the retype-context-after branch from 323effe to b7e26ac Compare June 4, 2026 12:12

oli-obk reviewed Jun 9, 2026

View reviewed changes

oli-obk approved these changes Jun 9, 2026

View reviewed changes

rust-bors Bot merged commit e295994 into rust-lang:main Jun 10, 2026
12 of 13 checks passed

rustbot added this to the 1.98.0 milestone Jun 10, 2026

cjgillot deleted the retype-context-after branch June 10, 2026 10:52

This comment has been minimized.

Sign in to view

rustbot added the perf-regression Performance regression. label Jun 13, 2026

Uh oh!

Conversation

cjgillot commented May 30, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rustbot commented May 30, 2026

Uh oh!

rustbot commented May 30, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

rustbot commented Jun 4, 2026

Uh oh!

oli-obk Jun 9, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

oli-obk left a comment • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rust-bors Bot commented Jun 9, 2026

Uh oh!

jhpratt commented Jun 9, 2026

Uh oh!

rust-bors Bot commented Jun 9, 2026

Uh oh!

rust-bors Bot commented Jun 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jhpratt commented Jun 10, 2026

Uh oh!

rust-bors Bot commented Jun 10, 2026

Uh oh!

rust-bors Bot commented Jun 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jhpratt commented Jun 10, 2026

Uh oh!

rust-bors Bot commented Jun 10, 2026

Uh oh!

Uh oh!

JonathanBrouwer commented Jun 13, 2026

Uh oh!

This comment has been minimized.

rust-timer commented Jun 13, 2026

Overall result: ❌ regressions - please read:

Instruction count

Max RSS (memory usage)

Cycles

Binary size

Uh oh!

JonathanBrouwer commented Jun 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

cjgillot commented May 30, 2026 •

edited by rustbot

Loading

oli-obk Jun 9, 2026 •

edited by rustbot

Loading

oli-obk left a comment •

edited by rustbot

Loading

rust-bors Bot commented Jun 10, 2026 •

edited

Loading

rust-bors Bot commented Jun 10, 2026 •

edited

Loading

JonathanBrouwer commented Jun 13, 2026 •

edited

Loading