[WIP] rustdoc: Rewrite auto trait impl synthesis #149019

fmease · 2025-11-17T10:52:42Z

Rewrite its core from scratch using the next-gen trait solver and its ProofTreeVisitor.

Fixes #105199.
Fixes #110740.
Fixes #110741.
Fixes #111102 (tiny FIXME: need to wrap ACE RHS in { } if it's a Param).
Fixes #114097.
Fixes #116539.
Fixes #120606.
Fixes #127593.
Fixes #136778.
Fixes #139964 (no longer hangs, FIXME: now fails under VERIFY).
Fixes #144918.
Fixes #149281 (no longer ICEs, FIXME: we still don't propagate const Bounds).

FIXME: #123298 (still fails under VERIFY).
FIXME: #91380 (Jyn's code: fully fixed; original code: still fails under VERIFY).

Also aims to address rust-lang/trait-system-refactor-initiative#157 & rust-lang/trait-system-refactor-initiative#208 (yet to be tested). NB: Upper bound is #141564 (comment).

fmease · 2025-11-24T02:17:55Z

Not finished but let's get an initial reading, I should've done that a while ago.

@bors try @rust-timer queue

…sis, r=<try> [WIP] rustdoc: Rewrite auto trait impl synthesis

rust-bors · 2025-11-24T04:40:26Z

☀️ Try build successful (CI)
Build commit: 634512a (634512ace28481d8e463b64400a5788b037103de, parent: c23ed3ef28b3cc30658310f771ddc66e5b687cf4)

rust-timer · 2025-11-24T05:59:37Z

Finished benchmarking commit (634512a): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please do so in sufficient writing along with @rustbot label: +perf-regression-triaged. If not, please fix the regressions and do another perf run. If its results are neutral or positive, the label will be automatically removed.

@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)	0.2%	[0.2%, 0.2%]	1
Regressions ❌ (secondary)	20.2%	[0.2%, 40.2%]	2
Improvements ✅ (primary)	-3.6%	[-14.8%, -0.2%]	16
Improvements ✅ (secondary)	-2.3%	[-12.3%, -0.1%]	7
All ❌✅ (primary)	-3.4%	[-14.8%, 0.2%]	17

Max RSS (memory usage)

Results (primary 2.3%, secondary -3.8%)

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

	mean	range	count
Regressions ❌ (primary)	2.3%	[1.0%, 5.9%]	10
Regressions ❌ (secondary)	1.7%	[1.7%, 1.7%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-4.5%	[-6.5%, -2.2%]	8
All ❌✅ (primary)	2.3%	[1.0%, 5.9%]	10

Cycles

Results (primary -5.3%, secondary 3.9%)

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)	6.6%	[1.8%, 26.4%]	9
Improvements ✅ (primary)	-5.3%	[-10.3%, -3.1%]	7
Improvements ✅ (secondary)	-8.2%	[-10.0%, -6.3%]	2
All ❌✅ (primary)	-5.3%	[-10.3%, -3.1%]	7

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 471.926s -> 471.456s (-0.10%)
Artifact size: 388.27 MiB -> 386.31 MiB (-0.50%)

src/librustdoc/clean/auto_trait.rs

lcnr · 2025-11-26T09:16:14Z

src/librustdoc/clean/auto_trait.rs

+
+        if let Some(clause) = predicate.as_clause()
+            && let ty::ClauseKind::Projection(pred) = clause.kind().skip_binder()
+            && let ty::Param(_) = pred.self_ty().kind()


think this doesn't handle <T::Assoc as Trait>::Assoc

Yes, indeed. I did already implement that a while ago locally but didn't push any of it because it led me into a deep rabbit hole, I'll elaborate later.

I'll elaborate later.

Ok, so while I obviously need to recurse into the self type while it's a projection until it's a type param (and bail if it isn't) [A], presently I believe I also have to push any(?) AliasRelate(Equate) [B] as well as any trait & projection predicates that has_infer() assuming the goal result is ambiguous [C].

I'm still a bit lost in the weeds there though. Why do I need to do that? Well it's to account for a whole class of scenarios that feature projection predicates. See below:

I'm subsequently assuming the following "preamble":

#![feature(auto_traits)] pub auto trait Marker {} pub struct Outer<T>(Inner<T>); struct Inner<T>(T);

I'm also assuming we're "analyzing" <T> <Outer<T> as Marker>.

Here, we want to collect <T as Iterator>, Proj(Iterator::Item, [T]) == ?0 and <?0 as Copy> (omitting sized preds obv):

impl<T, U> Marker for Inner<T> where T: Iterator<Item = U>, U: Copy, // or `(U,): Copy` etc. {}

That I can easily manage. But now consider this:

impl<T> Marker for Inner<T> where T: Iterator<Item: Iterator<Item = ()>> {}

Here the proof tree contains AliasRelate(Proj(…), …)s and it gets rather hairy (all those intermediary type variables obviously, etc.). I was able to hack something together locally but it isn't beautiful. I don't have a concrete question and I'm not sure if I'm seeking guidance, I just got a bit overwhelmed ^^'

In any case, that made me realize that it's probably more than okay if the algorithm remains rather incomplete in this PR, an MVP if you will, as it's more important to migrate to the next solver than to perfect SATI (thereby conservatively reporting unimplemented in more complicated cases for now).

After all, the SATI@main is even more incomplete: It filters out projection predicates that contain infer vars (even though those can be turned into extra synthetic parameters like I do in this PR (in the old SATI the synthetic impl has to have the same params as the ADT which is ofc super restrictive))! What's worse, it unconditionally validates the resulting set of predicates which of course leads to all of these ICEs due to its incompleteness! ^^'

fmease · 2025-11-28T22:36:51Z

@bors try parent=last @rust-timer queue

fmease · 2025-11-28T22:57:02Z

src/librustdoc/clean/auto_trait.rs

+        if nested_goals.is_empty() {
+            return ControlFlow::Break(());
+        }


This feels like a hack to me. Namely, breaking & considering the auto trait unimplemented if (the goal failed and) there is a candidate, a single one, but it doesn't have any goals (so it's kind of "unproductive" or "fake"). Otherwise I only ever break if there's no candidate at all which makes total sense.

Why did I add this check? Well, to correctly determine that <'a> <Outer<'a> as Marker> is unimplemented here (due to a universe mismatch):

#![feature(auto_traits)] pub auto trait Marker {} pub struct Outer<'a>(fn(&'a ())); impl Marker for for<'a> fn(&'a ()) {}

DEBUG PROOF TREE

(G) Err(NoSolution) Misc #C=1 for[] TraitPredicate(<Outer<'a> as Marker>, polarity:Positive) (C#0) Err(NoSolution) TraitCandidate/BuiltinImpl(Misc) (G) Err(NoSolution) ImplWhereBound #C=1 for[] TraitPredicate(<fn(&'a ()) as Marker>, polarity:Positive) (C#0) Err(NoSolution) TraitCandidate/Impl(DefId(0:8 ~ rm[1917]::{impl#0}))

Is that how you're "supposed" to detect this?

rust-bors · 2025-11-29T00:53:14Z

☀️ Try build successful (CI)
Build commit: 94939e6 (94939e6e3ab893597fa00b53eef0f474a0bb2ea7, parent: c23ed3ef28b3cc30658310f771ddc66e5b687cf4)

rust-timer · 2025-11-29T02:13:10Z

Finished benchmarking commit (94939e6): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please do so in sufficient writing along with @rustbot label: +perf-regression-triaged. If not, please fix the regressions and do another perf run. If its results are neutral or positive, the label will be automatically removed.

@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)	40.0%	[40.0%, 40.0%]	1
Improvements ✅ (primary)	-3.9%	[-15.3%, -0.2%]	16
Improvements ✅ (secondary)	-2.3%	[-12.7%, -0.1%]	8
All ❌✅ (primary)	-3.9%	[-15.3%, -0.2%]	16

Max RSS (memory usage)

Results (primary 1.6%, secondary -4.4%)

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

	mean	range	count
Regressions ❌ (primary)	2.0%	[1.2%, 5.1%]	7
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-1.1%	[-1.1%, -1.1%]	1
Improvements ✅ (secondary)	-4.4%	[-6.6%, -2.2%]	8
All ❌✅ (primary)	1.6%	[-1.1%, 5.1%]	8

Cycles

Results (primary -5.4%, secondary 2.9%)

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)	7.1%	[2.2%, 25.9%]	8
Improvements ✅ (primary)	-5.4%	[-11.2%, -2.4%]	8
Improvements ✅ (secondary)	-5.5%	[-11.0%, -1.7%]	4
All ❌✅ (primary)	-5.4%	[-11.2%, -2.4%]	8

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 471.926s -> 470.104s (-0.39%)
Artifact size: 388.27 MiB -> 386.27 MiB (-0.51%)

…sis, r=<try> [WIP] rustdoc: Rewrite auto trait impl synthesis

* avoid looking for user-written impls (there can't ever be any) * avoid copying & interning clauses * avoid processing region obligations

…sis, r=<try> [WIP] rustdoc: Rewrite auto trait impl synthesis

fmease · 2025-11-29T19:22:53Z

@bors try parent=last

rust-bors · 2025-11-29T19:23:02Z

⌛ Trying commit 574042c with merge 978c108…

(The previously running try build was automatically cancelled.)

To cancel the try build, run the command @bors try cancel.

Workflow: https://github.com/rust-lang/rust/actions/runs/19788265666

…sis, r=<try> [WIP] rustdoc: Rewrite auto trait impl synthesis

fmease · 2025-11-29T22:40:29Z

@rust-timer build 978c108 profiles=doc

rust-timer · 2025-11-30T00:19:40Z

Finished benchmarking commit (978c108): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please do so in sufficient writing along with @rustbot label: +perf-regression-triaged. If not, please fix the regressions and do another perf run. If its results are neutral or positive, the label will be automatically removed.

@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)	40.0%	[40.0%, 40.0%]	1
Improvements ✅ (primary)	-3.9%	[-15.3%, -0.2%]	16
Improvements ✅ (secondary)	-3.7%	[-12.7%, -0.4%]	5
All ❌✅ (primary)	-3.9%	[-15.3%, -0.2%]	16

Max RSS (memory usage)

Results (primary 1.7%, secondary -2.1%)

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

	mean	range	count
Regressions ❌ (primary)	2.4%	[1.4%, 5.2%]	4
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-1.1%	[-1.1%, -1.1%]	1
Improvements ✅ (secondary)	-2.1%	[-2.1%, -2.1%]	1
All ❌✅ (primary)	1.7%	[-1.1%, 5.2%]	5

Cycles

Results (primary -5.6%, secondary 3.8%)

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)	25.4%	[25.4%, 25.4%]	1
Improvements ✅ (primary)	-5.6%	[-11.4%, -2.4%]	8
Improvements ✅ (secondary)	-7.0%	[-10.8%, -3.1%]	2
All ❌✅ (primary)	-5.6%	[-11.4%, -2.4%]	8

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 471.926s -> 471.234s (-0.15%)
Artifact size: 388.27 MiB -> 386.24 MiB (-0.52%)

fmease added the rla-silenced Silences rust-log-analyzer postings to the PR it's added on. label Nov 17, 2025

fmease force-pushed the rustdoc-rewrite-auto-trait-impl-synthesis branch 2 times, most recently from 15d9b7b to ea10870 Compare November 17, 2025 11:43

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 Nov 17, 2025

This was referenced Nov 18, 2025

rustdoc: ICE: synthetic auto trait impls: unable to fulfill trait due to ambiguity #144918

Open

Your Rust source code & software isn't licensed to the public FlexiFormal/FLAMS#59

Closed

fmease removed the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Nov 19, 2025

This was referenced Nov 19, 2025

rustdoc: ICE: synthetic auto trait impls: unexpected result when selecting […] #135363

Closed

rustdoc: ICE: synthetic auto trait impls: unexpected result when selecting […] obligation #105781

Closed

This comment has been minimized.

Sign in to view

rust-bors bot added a commit that referenced this pull request Nov 24, 2025

Auto merge of #149019 - fmease:rustdoc-rewrite-auto-trait-impl-synthe…

634512a

…sis, r=<try> [WIP] rustdoc: Rewrite auto trait impl synthesis

This comment has been minimized.

Sign in to view

rustbot added the perf-regression Performance regression. label Nov 24, 2025

fmease assigned lcnr Nov 25, 2025

This comment was marked as outdated.

Sign in to view

lcnr reviewed Nov 26, 2025

View reviewed changes

src/librustdoc/clean/auto_trait.rs Show resolved Hide resolved

lcnr reviewed Nov 26, 2025

View reviewed changes

fmease force-pushed the rustdoc-rewrite-auto-trait-impl-synthesis branch from ea10870 to a5fdf11 Compare November 28, 2025 21:44

rustbot added the WG-trait-system-refactor The Rustc Trait System Refactor Initiative (-Znext-solver) label Nov 28, 2025

fmease force-pushed the rustdoc-rewrite-auto-trait-impl-synthesis branch from a5fdf11 to 107f5fe Compare November 28, 2025 22:34

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 Nov 28, 2025

fmease commented Nov 28, 2025

View reviewed changes

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 Nov 29, 2025

rustdoc: Rewrite auto trait impl synthesis

52980e2

fmease force-pushed the rustdoc-rewrite-auto-trait-impl-synthesis branch 2 times, most recently from b292701 to 4b20fe6 Compare November 29, 2025 19:07

rust-bors bot added a commit that referenced this pull request Nov 29, 2025

Auto merge of #149019 - fmease:rustdoc-rewrite-auto-trait-impl-synthe…

915912e

…sis, r=<try> [WIP] rustdoc: Rewrite auto trait impl synthesis

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Nov 29, 2025

fmease added 2 commits November 29, 2025 20:20

rustdoc: Perform less work when synth'ing Sized impls

f3b0947

* avoid looking for user-written impls (there can't ever be any) * avoid copying & interning clauses * avoid processing region obligations

rustdoc: Properly clean middle::ty negative bounds

574042c

fmease force-pushed the rustdoc-rewrite-auto-trait-impl-synthesis branch from 4b20fe6 to 574042c Compare November 29, 2025 19:21

rust-lang deleted a comment from rust-timer Nov 29, 2025

rust-lang deleted a comment from rust-bors bot Nov 29, 2025

rust-bors bot added a commit that referenced this pull request Nov 29, 2025

Auto merge of #149019 - fmease:rustdoc-rewrite-auto-trait-impl-synthe…

7a9d368

…sis, r=<try> [WIP] rustdoc: Rewrite auto trait impl synthesis

rust-lang deleted a comment from rust-timer Nov 29, 2025

rust-bors bot added a commit that referenced this pull request Nov 29, 2025

Auto merge of #149019 - fmease:rustdoc-rewrite-auto-trait-impl-synthe…

978c108

…sis, r=<try> [WIP] rustdoc: Rewrite auto trait impl synthesis

fmease removed the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Nov 29, 2025

This comment has been minimized.

Sign in to view

fmease mentioned this pull request Nov 30, 2025

rustdoc doesn't render non-trivial built-in Sized impls (and calls Sized impls auto trait impls) #149467

Open

rust-lang deleted a comment from rust-bors bot Nov 30, 2025

[WIP] rustdoc: Rewrite auto trait impl synthesis #149019

Are you sure you want to change the base?

[WIP] rustdoc: Rewrite auto trait impl synthesis #149019

Conversation

fmease commented Nov 17, 2025 • 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.

fmease commented Nov 24, 2025

Uh oh!

This comment has been minimized.

This comment has been minimized.

rust-bors bot commented Nov 24, 2025

Uh oh!

This comment has been minimized.

rust-timer commented Nov 24, 2025

Overall result: ❌✅ regressions and improvements - please read the text below

Uh oh!

This comment was marked as outdated.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

fmease Nov 26, 2025 • 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!

fmease Nov 28, 2025 • 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!

fmease Nov 28, 2025 • 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!

fmease commented Nov 28, 2025

Uh oh!

This comment has been minimized.

fmease Nov 28, 2025 • 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!

rust-bors bot commented Nov 29, 2025

Uh oh!

This comment has been minimized.

rust-timer commented Nov 29, 2025

Overall result: ❌✅ regressions and improvements - please read the text below

Uh oh!

fmease commented Nov 29, 2025

Uh oh!

rust-bors bot commented Nov 29, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

fmease commented Nov 29, 2025

Uh oh!

This comment has been minimized.

rust-timer commented Nov 30, 2025

Overall result: ❌✅ regressions and improvements - please read the text below

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

fmease commented Nov 17, 2025 •

edited

Loading

fmease Nov 26, 2025 •

edited

Loading

fmease Nov 28, 2025 •

edited

Loading

fmease Nov 28, 2025 •

edited

Loading

fmease Nov 28, 2025 •

edited

Loading

rust-bors bot commented Nov 29, 2025 •

edited

Loading