Do not increase recursion depth for coroutine witness and rigid opaques when proving auto traits by adwinwhite · Pull Request #162275 · rust-lang/rust

adwinwhite · 2026-09-04T04:52:52Z

With deeply nested async calls, we can easily overflow evaluating auto trait goals as shown in #159228.
Users have no choice but increase the per-crate recursion_limit which is bad for compilation time/RSS. And downstream users may encounter the same warnings when calling library async functions.

We mitigate that by no longer increasing recursion depth for coroutine witness and rigid opaques when proving auto traits. See the comments for why it's okay to do so.
This fix actually makes the required depth a third of what it was before for async calls.

I've checked locally that bors and triagebot no longer have FCWs.

r? lcnr

lcnr · 2026-09-07T14:06:32Z

+                            ty::Alias(
+                                ty::IsRigid::Yes,
+                                ty::AliasTy { kind: ty::Opaque { .. }, .. },
+                            ) => LowerAvailableDepth::No,


hmm, this means we get

Vec<opaque>: Send 7

opaque: Send 7

hidden_ty: Send 6

instead of

Vec<opaque>: Send 7

opaque: Send 6

hidden_ty: Send 6

This feels slightly iffy as you can also prove opaque: Send using its item bounds 🤔

does just the coroutine witness stuff fix some of the fcws?

View changes since the review

Oh, I forgot about alias bound candidate.

The coroutine witness change alone fixes triagebot but not bors, unfortunately.

hmm, we could moe the step_kind computation into this function and then use a separate goal source for the opaque type auto trait leakage?

alternatively, hmm. want to merge just the coroutine witness change for now. I do just feel generally less confident about doing this for auto trait leakage 🤔

I really don't know what I want here :> I guess doing it for opaque auto trait leakage is also fine 🤷 it also doesn#t feel too relevant or principled, no matter what we do 😅

It's awkward to pass the auto trait leakage info to search graph since the goal source is set by probe_and_evaluate_goal_for_constituent_tys which is used by more than auto traits.
Is it okay to check whether we're gonna pick alias bound candidate in this helper?
We'll still have

Vec: Send 7

opaque: Send 7

hidden: Send 6

but only when it's proved via auto trait leakage.

rustbot · 2026-09-11T04:54:23Z

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.

adwinwhite · 2026-09-11T06:08:34Z

@bors try @rust-timer queue

Do not increase recursion depth for coroutine witness and rigid opaques when proving auto traits

rust-bors · 2026-09-11T07:07:24Z

☀️ Try build successful (CI)
Build commit: 3a50673 (3a506736b86b6ad0edf17206fe49ac364edaaaeb)
Base parent: 9e9544d (9e9544da4cd757c74428bad35059cc8122fdcd7d)

rust-timer · 2026-09-11T08:08:57Z

Finished benchmarking commit (3a50673): comparison URL.

Overall result: ❌ regressions - no action needed

Benchmarking means the PR may be perf-sensitive. Consider adding rollup=never if this change is not fit for rolling up.

@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)	0.2%	[0.2%, 0.3%]	3
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Max RSS (memory usage)

Results (primary -0.4%)

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

	mean	range	count
Regressions ❌ (primary)	1.3%	[0.4%, 5.9%]	7
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-4.4%	[-7.4%, -2.8%]	3
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-0.4%	[-7.4%, 5.9%]	10

Cycles

Results (primary -0.6%, secondary 6.8%)

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

	mean	range	count
Regressions ❌ (primary)	0.9%	[0.4%, 1.3%]	2
Regressions ❌ (secondary)	6.8%	[6.8%, 6.8%]	1
Improvements ✅ (primary)	-0.9%	[-2.0%, -0.5%]	10
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-0.6%	[-2.0%, 1.3%]	12

Binary size

Results (primary -0.0%)

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)	-	-	0
Improvements ✅ (primary)	-0.0%	[-0.0%, -0.0%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-0.0%	[-0.0%, -0.0%]	1

Bootstrap: 478.978s -> 478.032s (-0.20%)
Artifact size: 403.72 MiB -> 403.81 MiB (0.02%)

Current nightly fails without this. This change is undesirable for a number of reasons - see arti#2715. But it is required for now to allow building with nightly. We should revert it as soon as possible. (I am watching the upstream ticket and the MR rust-lang/rust/pull/162275 which may fix it.) Also, this is a very noisy change which affects many many crates, so we hope to revert it before the next release! THIS IS NOT THE ONLY COMMIT TO REVERT. When reverting, use git log -G 'recursion.*limit' to find appropriate commits an d git grep 'recursion.*limit' to check you're done.

lcnr · 2026-09-24T09:19:21Z

+                            ) => {
+                                // We only want to skip lowering depth when the proving is done via auto
+                                // trait leakage. If the goal can be proved via item bounds, we
+                                // should lower depth faithfully.


can you add a "FIXME: Ideally we'd instead lower the depth for the nested goal instead. Implementing this is a bit harder"

because you could have impl Sized: Send bound as a where-clause with TAIT/rtn. getting the recursion depth wrong isn't a big issue, but would still like to to not be slightly scuffed long-term

with this r=me

View changes since the review

…t proving

lcnr · 2026-09-24T15:53:13Z

@bors r+ rollup

rust-bors · 2026-09-24T15:53:17Z

📌 Commit caadd68 has been approved by lcnr

It is now in the queue for this repository.

Do not increase recursion depth for coroutine witness and rigid opaques when proving auto traits With deeply nested async calls, we can easily overflow evaluating auto trait goals as shown in rust-lang#159228. Users have no choice but increase the per-crate `recursion_limit` which is bad for compilation time/RSS. And downstream users may encounter the same warnings when calling library async functions. We mitigate that by no longer increasing recursion depth for coroutine witness and rigid opaques when proving auto traits. See the comments for why it's okay to do so. This fix actually makes the required depth a third of what it was before for async calls. I've checked locally that `bors` and `triagebot` no longer have FCWs. r? lcnr

Rollup of 9 pull requests Successful merges: - #162228 (tests: Run more pauth tests in CI and make them pass) - #162493 (Add support for -Zsanitizer-cfi-minimal-runtime) - #163271 (add a leak check test) - #163282 (cleanup: clean up more dependencies that are unused) - #157562 (Avoid computing layout of enums with non-int discriminants) - #162275 (Do not increase recursion depth for coroutine witness and rigid opaques when proving auto traits) - #163037 (fix and test `va_arg` on `f128` on `x86`) - #163114 (Add regression test for hang on mutually recursive trait impls) - #163255 (Check the entire library and cg_clif workspaces for permitted deps in tidy)

Rollup of 8 pull requests Successful merges: - #162228 (tests: Run more pauth tests in CI and make them pass) - #163271 (add a leak check test) - #163282 (cleanup: clean up more dependencies that are unused) - #157562 (Avoid computing layout of enums with non-int discriminants) - #162275 (Do not increase recursion depth for coroutine witness and rigid opaques when proving auto traits) - #163037 (fix and test `va_arg` on `f128` on `x86`) - #163114 (Add regression test for hang on mutually recursive trait impls) - #163255 (Check the entire library and cg_clif workspaces for permitted deps in tidy)

Rollup merge of #162275 - adwinwhite:half-depth, r=lcnr Do not increase recursion depth for coroutine witness and rigid opaques when proving auto traits With deeply nested async calls, we can easily overflow evaluating auto trait goals as shown in #159228. Users have no choice but increase the per-crate `recursion_limit` which is bad for compilation time/RSS. And downstream users may encounter the same warnings when calling library async functions. We mitigate that by no longer increasing recursion depth for coroutine witness and rigid opaques when proving auto traits. See the comments for why it's okay to do so. This fix actually makes the required depth a third of what it was before for async calls. I've checked locally that `bors` and `triagebot` no longer have FCWs. r? lcnr

Current nightly fails without this. This change is undesirable for a number of reasons - see arti#2715. But it is required for now to allow building with nightly. We should revert it as soon as possible. (I am watching the upstream ticket and the MR rust-lang/rust/pull/162275 which may fix it.) Also, this is a very noisy change which affects many many crates, so we hope to revert it before the next release! THIS IS NOT THE ONLY COMMIT TO REVERT. When reverting, use git log -G 'recursion.*limit' to find appropriate commits an d git grep 'recursion.*limit' to check you're done.

rustbot assigned lcnr Sep 4, 2026

adwinwhite marked this pull request as ready for review September 4, 2026 06:13

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 4, 2026

lcnr reviewed Sep 7, 2026

View reviewed changes

michaelkaplan13 mentioned this pull request Sep 8, 2026

Fix Lint commonwarexyz/constantinople#56

Merged

Kobzol mentioned this pull request Sep 10, 2026

Tracking issue for future-compatibility warning recursion_depth_exceeding_limit #159228

Open

adwinwhite force-pushed the half-depth branch from 855ea9e to cf1f8ac Compare September 11, 2026 04:54

This comment has been minimized.

Sign in to view

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Sep 11, 2026

This comment has been minimized.

Sign in to view

rust-bors Bot pushed a commit that referenced this pull request Sep 11, 2026

Auto merge of #162275 - adwinwhite:half-depth, r=<try>

3a50673

Do not increase recursion depth for coroutine witness and rigid opaques when proving auto traits

This comment has been minimized.

Sign in to view

rustbot removed the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Sep 11, 2026

lcnr reviewed Sep 24, 2026

View reviewed changes

dont lower depth for coroutine witness and rigid opaques in auto trai…

caadd68

…t proving

adwinwhite force-pushed the half-depth branch from cf1f8ac to caadd68 Compare September 24, 2026 10:47

rust-bors Bot added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Sep 24, 2026

jhpratt mentioned this pull request Sep 24, 2026

Rollup of 9 pull requests #163287

Closed

jhpratt mentioned this pull request Sep 25, 2026

Rollup of 8 pull requests #163300

Merged

rust-bors Bot merged commit 396431e into rust-lang:main Sep 25, 2026
13 checks passed

rustbot added this to the 1.100.0 milestone Sep 25, 2026

github-actions Bot mentioned this pull request Oct 3, 2026

Toolchain upgrade to nightly-2026-10-06 flux-rs/flux#1811

Draft

Uh oh!

Conversation

adwinwhite commented Sep 4, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

lcnr Sep 7, 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!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

adwinwhite Sep 11, 2026 • edited 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 11, 2026

Uh oh!

adwinwhite commented Sep 11, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

rust-bors Bot commented Sep 11, 2026

Uh oh!

This comment has been minimized.

rust-timer commented Sep 11, 2026

Overall result: ❌ regressions - no action needed

Uh oh!

lcnr Sep 24, 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!

lcnr commented Sep 24, 2026

Uh oh!

rust-bors Bot commented Sep 24, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

adwinwhite commented Sep 4, 2026 •

edited by rustbot

Loading

lcnr Sep 7, 2026 •

edited by rustbot

Loading

adwinwhite Sep 11, 2026 •

edited

Loading

lcnr Sep 24, 2026 •

edited by rustbot

Loading