Make `#[track_caller] async fn` track the caller, not the poller/awaiter by theemathas · Pull Request #163396 · rust-lang/rust

theemathas · 2026-09-27T08:36:14Z

An LLM was used to locate relevant code, suggest ideas, and review the code. However, I manually wrote all code, and I take all responsibility for all code written and all decisions made.

Tracking issue (async_fn_track_caller): #110011

Related tracking issue (closure_track_caller): #87417

This PR is stacked on top of #163262 (first commit in this PR) and #163746 (the second commit in this PR, which was cherry-picked in). The third commit adds extensive tests for behavior touched in this PR. Commits 4-9 are the actual implementation, and also modification of the tests to reflect this new behavior.

This PR makes coroutines desugared from #[track_caller] async fn track the caller of the function, not the poller/awaiter of the coroutine. This is done as per T-lang's decision at #110011 (comment). The implementation modifies AST lowering to add an upvar to the coroutine, which stores the relevant caller location.

In both codegen and in consteval/miri, we read the relevant upvar from the coroutine at the point in time where we need access to the &Location reference (either when we call the caller_location intrinsic, when we panic, or when we call another #[track_caller] function). We only dereference this reference as needed, when we actually need the Location information.

This PR changes the behavior of async fn, but not of async closures, due to difficulties I've noted at #t-compiler/help > Help with `async_fn_track_caller` @ 💬

This PR's implementation assumes that the desired behavior in #110011 (comment) is the behavior that I think makes the most sense.

This PR corrects the behavior of #163406 when the async_fn_track_caller feature is enabled, but does not give a warning when the feature is disabled (resulting in the pre-existing broken behavior).

The tests pass on cranelift when I run ./x test ui --test-codegen-backend cranelift -- async-await/track-caller/panic-track-caller-no-unwind.rs. (Cranelift doesn't support unwinding.)

r? compiler

theemathas · 2026-09-29T12:00:58Z

+        if is_in_trait_impl {
+            // Check if we need to "inherit" #[track_caller] from the trait definition.
+            let Some(trait_item_def_id) =
+                self.get_partial_res(node_id).and_then(|r| r.expect_full_res().opt_def_id())
+            else {
+                self.dcx().span_delayed_bug(span, "could not resolve trait item being implemented");
+                return false;
+            };
+            return find_attr!(self.tcx, trait_item_def_id, TrackCaller(_));
+        }


I don't understand what this is doing, but I'm doing the same thing as this existing code:

rust/compiler/rustc_ast_lowering/src/item.rs

Lines 1148 to 1155 in c1070d6

let (effective_ident, impl_kind) = if is_in_trait_impl {

let trait_item_def_id = self

.get_partial_res(i.id)

.and_then(|r| r.expect_full_res().opt_def_id())

.ok_or_else(|| {

self.dcx()

.span_delayed_bug(span, "could not resolve trait item being implemented")

});

View changes since the review

theemathas · 2026-09-29T12:09:28Z

+///
+/// FIXME(async_fn_track_caller): What if the Location is stored inside a coroutine upvar?
 fn resolve_tracked_call_location(body: &Body, inherited: Option<MirConst>) -> MirConst {


What is this, and do I need to modify it?

View changes since the review

rustbot · 2026-09-29T12:19:45Z

miri is developed in its own repository. If the Miri part of this change can be broken out, consider making this change to rust-lang/miri instead. However, if Miri needs adjusting for rustc changes, just ignore this message.

cc @rust-lang/miri

Some changes occurred to the CTFE machinery

cc @RalfJung, @oli-obk, @lcnr

Some changes occurred in compiler/rustc_attr_ir

cc @jdonszelmann, @JonathanBrouwer

Some changes occurred to MIR optimizations

cc @rust-lang/wg-mir-opt

Some changes occurred to the intrinsics. Make sure the CTFE / Miri interpreter
gets adapted for the changes, if necessary.

cc @rust-lang/miri, @RalfJung, @oli-obk, @lcnr

Some changes occurred to the CTFE / Miri interpreter

cc @rust-lang/miri

theemathas · 2026-10-01T14:57:54Z

I split up the implementation into multiple commits, hopefully making reviewing easier. (And also hopefully making it less confusing for me.)

theemathas · 2026-10-07T15:35:17Z

I've reimplemented this PR in the Miri portion so that it doesn't walk the stack, but instead passes the caller location argument down the stack, as per @RalfJung's suggestion.

rustbot · 2026-10-07T15:48:32Z

rustc_codegen_cranelift is developed in its own repository. If possible, consider making this change to rust-lang/rustc_codegen_cranelift instead.

cc @bjorn3

…try> Make `#[track_caller] async fn` track the caller, not the poller/awaiter try-job: dist-x86_64-linux

theemathas · 2026-10-08T14:08:02Z

Added tests that are verified locally to pass for cranelift.

@rustbot ready

Edit: Maybe it might be better if I use #[cfg(panic = "unwind")] instead? (Edit: Done.)

…ter. Part 1: AST lowering

…ter. Part 2: MIR building

…ter. Part 3: misc

…ter. Part 4: codegen

…ter. Part 5: Miri

…ter. Part 6: Cranelift

theemathas · 2026-10-09T14:34:09Z

Tested locally for cranelift by applying the following change and running ./x test ui --test-codegen-backend cranelift -- async-await/track-caller/panic-track-caller.rs

diff --git a/tests/ui/async-await/track-caller/panic-track-caller.rs b/tests/ui/async-await/track-caller/panic-track-caller.rs
index 99dec25fdca..248db622bf6 100644
--- a/tests/ui/async-await/track-caller/panic-track-caller.rs
+++ b/tests/ui/async-await/track-caller/panic-track-caller.rs
@@ -1,15 +1,15 @@
 // This test is duplicated (with changes) at
 // src/tools/miri/tests/pass/async-panic-track-caller.rs
-
+//@ no-prefer-dynamic
 //@ run-pass
 //@ edition:2024
 //@ revisions: nofeat afn cls afn_cls nofeat_opt afn_opt cls_opt afn_cls_opt
-//@[nofeat_opt] compile-flags: -O -Zinline-mir-hint-threshold=1000
-//@[afn_opt] compile-flags: -O -Zinline-mir-hint-threshold=1000
-//@[cls_opt] compile-flags: -O -Zinline-mir-hint-threshold=1000
-//@[afn_cls_opt] compile-flags: -O -Zinline-mir-hint-threshold=1000
+//@[nofeat_opt] compile-flags: -O -Zinline-mir-hint-threshold=1000 -Cpanic=abort
+//@[afn_opt] compile-flags: -O -Zinline-mir-hint-threshold=1000 -Cpanic=abort
+//@[cls_opt] compile-flags: -O -Zinline-mir-hint-threshold=1000 -Cpanic=abort
+//@[afn_cls_opt] compile-flags: -O -Zinline-mir-hint-threshold=1000 -Cpanic=abort
+//@ compile-flags: -Cpanic=abort
 // gate-test-async_fn_track_caller
-//
 #![feature(stmt_expr_attributes, coroutines, coroutine_trait, gen_blocks)]
 #![cfg_attr(any(afn, afn_cls, afn_opt, afn_cls_opt), feature(async_fn_track_caller))]
 #![cfg_attr(any(cls, afn_cls, cls_opt, afn_cls_opt), feature(closure_track_caller))]

theemathas force-pushed the async-track-caller-not-poller branch 3 times, most recently from f21a754 to 7e01c99 Compare September 28, 2026 15:47

This comment has been minimized.

Sign in to view

theemathas force-pushed the async-track-caller-not-poller branch from 7e01c99 to c87bff0 Compare September 29, 2026 11:52

theemathas commented Sep 29, 2026

View reviewed changes

rustbot assigned oli-obk Sep 29, 2026

theemathas marked this pull request as ready for review September 29, 2026 12:19

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Sep 29, 2026

RalfJung reviewed Sep 29, 2026

View reviewed changes

Comment thread compiler/rustc_const_eval/src/interpret/eval_context.rs Outdated

RalfJung reviewed Sep 29, 2026

View reviewed changes

Comment thread compiler/rustc_const_eval/src/interpret/operand.rs Outdated

theemathas force-pushed the async-track-caller-not-poller branch from c87bff0 to b14e6ed Compare September 29, 2026 13:07

RalfJung reviewed Sep 29, 2026

View reviewed changes

Comment thread compiler/rustc_const_eval/src/interpret/intrinsics.rs Outdated

This comment was marked as resolved.

Sign in to view

theemathas commented Sep 30, 2026

View reviewed changes

Comment thread compiler/rustc_codegen_ssa/src/mir/block.rs

theemathas force-pushed the async-track-caller-not-poller branch from b14e6ed to e1d818f Compare October 1, 2026 13:57

This comment has been minimized.

Sign in to view

theemathas force-pushed the async-track-caller-not-poller branch from e1d818f to 6fcdc09 Compare October 1, 2026 14:56

theemathas force-pushed the async-track-caller-not-poller branch 2 times, most recently from 6fcdc09 to 473dbf0 Compare October 1, 2026 15:01

theemathas changed the title ~~Make #[track_caller] async fn track the caller, not the poller~~ Make #[track_caller] async fn track the caller, not the poller/awaiter Oct 2, 2026

theemathas force-pushed the async-track-caller-not-poller branch from 473dbf0 to 57abba7 Compare October 2, 2026 11:02

theemathas added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Oct 7, 2026

This comment has been minimized.

Sign in to view

theemathas force-pushed the async-track-caller-not-poller branch from 2f38293 to 4598eb9 Compare October 7, 2026 15:48

This comment has been minimized.

Sign in to view

theemathas force-pushed the async-track-caller-not-poller branch from 4598eb9 to 9c81c0d Compare October 7, 2026 16:01

This comment was marked as resolved.

Sign in to view

rustbot added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Oct 7, 2026

This comment was marked as resolved.

Sign in to view

rust-bors Bot pushed a commit that referenced this pull request Oct 8, 2026

Auto merge of #163396 - theemathas:async-track-caller-not-poller, r=<…

f292875

…try> Make `#[track_caller] async fn` track the caller, not the poller/awaiter try-job: dist-x86_64-linux

This comment was marked as resolved.

Sign in to view

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Oct 8, 2026

Improve tests for async track caller

b1f4ad1

theemathas force-pushed the async-track-caller-not-poller branch from 822ca16 to aa2bf51 Compare October 9, 2026 12:00

theemathas added 6 commits October 9, 2026 20:12

Make #[track_caller] async fn track the caller, not the poller/awai…

6cdbc6a

…ter. Part 1: AST lowering

Make #[track_caller] async fn track the caller, not the poller/awai…

7f2b3f5

…ter. Part 2: MIR building

Make #[track_caller] async fn track the caller, not the poller/awai…

2fe505a

…ter. Part 3: misc

Make #[track_caller] async fn track the caller, not the poller/awai…

b7648f3

…ter. Part 4: codegen

Make #[track_caller] async fn track the caller, not the poller/awai…

173c37f

…ter. Part 5: Miri

Make #[track_caller] async fn track the caller, not the poller/awai…

f857e66

…ter. Part 6: Cranelift

theemathas force-pushed the async-track-caller-not-poller branch from aa2bf51 to f857e66 Compare October 9, 2026 13:13

	let (effective_ident, impl_kind) = if is_in_trait_impl {
	let trait_item_def_id = self
	.get_partial_res(i.id)
	.and_then(\|r\| r.expect_full_res().opt_def_id())
	.ok_or_else(\|\| {
	self.dcx()
	.span_delayed_bug(span, "could not resolve trait item being implemented")
	});

Conversation

theemathas commented Sep 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment has been minimized.

This comment has been minimized.

theemathas Sep 29, 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!

Uh oh!

theemathas Sep 29, 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!

rustbot commented Sep 29, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

This comment was marked as resolved.

Uh oh!

This comment has been minimized.

theemathas commented Oct 1, 2026

Uh oh!

theemathas commented Oct 7, 2026

Uh oh!

This comment has been minimized.

rustbot commented Oct 7, 2026

Uh oh!

This comment has been minimized.

This comment was marked as resolved.

This comment was marked as resolved.

This comment was marked as resolved.

This comment was marked as resolved.

This comment was marked as resolved.

This comment was marked as resolved.

theemathas commented Oct 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

theemathas commented Oct 9, 2026

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

theemathas commented Sep 27, 2026 •

edited

Loading

theemathas Sep 29, 2026 •

edited by rustbot

Loading

theemathas Sep 29, 2026 •

edited by rustbot

Loading

theemathas commented Oct 8, 2026 •

edited

Loading