Rework unsizing coercions in the new solver #147840

jdonszelmann · 2025-10-18T08:55:27Z

Replaces #141926, contains:

a commit adding tests that fail before this work
the two commits from the previous PR
a commit in which these tests are fixed
finally, a fixup for an in my opinion rather large regression in diagnostics. It's still not perfect, but better?

I hope this is roughly what you had in mind

Fixes rust-lang/trait-system-refactor-initiative#241 and rust-lang/trait-system-refactor-initiative#238, adding tests for both

r? @lcnr

rustbot · 2025-10-18T08:55:30Z

Some changes occurred to the core trait solver

cc @rust-lang/initiative-trait-system-refactor

changes to inspect_obligations.rs

cc @compiler-errors, @lcnr

BoxyUwU · 2025-10-18T09:08:48Z

can you split the old solver custom fulfillment loop out into its own function? named something that makes it clear that its old solver specific too

lcnr

please add comments to the added tests explaining why we've added them and also link to the relevant issues

View changes since this review

lcnr · 2025-10-20T07:55:10Z

tests/ui/traits/next-solver/unsize-goal-mismatch.rs

+fn foo<'a>(x: Box<dyn Trait<'a>>) -> Box<dyn Super<'a>> {
+    x
+    //~^ ERROR type annotations needed: cannot satisfy `dyn Trait<'_>: Unsize<dyn Super<'_>>
+}


why does this one not use revisions?

good question 🤔 it should get some :)

rustbot · 2025-10-20T11:53:53Z

This PR was rebased onto a different master 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.

compiler/rustc_hir_typeck/src/coercion.rs

jdonszelmann · 2025-10-27T11:01:41Z

@rustbot review

lcnr · 2025-10-27T11:08:24Z

compiler/rustc_hir_typeck/src/coercion.rs

+                    && let ty::Infer(ty::TyVar(vid)) = *pred.self_ty().skip_binder().kind()
+                    && self.fcx.type_var_is_sized(vid)
+                {
+                    // Here we detected that an unszing to `dyn Trait` is possible, yay!


this comment does not quite seem right 🤔

So what's going on is that we end up with a coercion from ?is_sized to some dyn Trait. The only way that can succeed is via unsizing, so we just unsize here. If we don't know whether the infer var has to be Sized, this could also just end up being an ordinary assignment, in which case trying to coerce would be wrong.

As the error messages when coercing where we shouldn't are worse than the error messages when not coercing when we should, we fall back to equality/subtyping if it's unclear whether we should coerce

hmm, I think I understand what you mean here, and I think the comment explains that better now

lcnr · 2025-10-27T11:09:00Z

pls link to the fixed trait-system-refactor-initiative issues 😁

jdonszelmann · 2025-10-27T11:27:33Z

@rustbot review

lcnr · 2025-10-27T11:40:12Z

@bors r+

bors · 2025-10-27T11:40:15Z

📌 Commit 1756622 has been approved by lcnr

It is now in the queue for this repository.

…=lcnr Rework unsizing coercions in the new solver Replaces rust-lang#141926, contains: - a commit adding tests that fail before this work - the two commits from the previous PR - a commit in which these tests are fixed - finally, a fixup for an in my opinion rather large regression in diagnostics. It's still not perfect, but better? I hope this is roughly what you had in mind Fixes rust-lang/trait-system-refactor-initiative#241 and rust-lang/trait-system-refactor-initiative#238, adding tests for both r? `@lcnr`

Rollup of 6 pull requests Successful merges: - #147840 (Rework unsizing coercions in the new solver) - #148139 (Add `coverage` scope for controlling paths in code coverage) - #148154 (Add a mailmap entry) - #148158 (ci: loongarch64: use medium code model to avoid relocation overflows) - #148172 (rustc-dev-guide subtree update) - #148175 (Fix typos: duplicate words in comments) r? `@ghost` `@rustbot` modify labels: rollup

…=lcnr Rework unsizing coercions in the new solver Replaces rust-lang#141926, contains: - a commit adding tests that fail before this work - the two commits from the previous PR - a commit in which these tests are fixed - finally, a fixup for an in my opinion rather large regression in diagnostics. It's still not perfect, but better? I hope this is roughly what you had in mind Fixes rust-lang/trait-system-refactor-initiative#241 and rust-lang/trait-system-refactor-initiative#238, adding tests for both r? ``@lcnr``

…=lcnr Rework unsizing coercions in the new solver Replaces rust-lang#141926, contains: - a commit adding tests that fail before this work - the two commits from the previous PR - a commit in which these tests are fixed - finally, a fixup for an in my opinion rather large regression in diagnostics. It's still not perfect, but better? I hope this is roughly what you had in mind Fixes rust-lang/trait-system-refactor-initiative#241 and rust-lang/trait-system-refactor-initiative#238, adding tests for both r? ```@lcnr```

Rollup of 14 pull requests Successful merges: - #144936 (CFI: Fix types that implement Fn, FnMut, or FnOnce) - #147185 (repr(transparent): do not consider repr(C) types to be 1-ZST) - #147840 (Rework unsizing coercions in the new solver) - #147915 (Update target maintainers android.md) - #148013 (1.91.0 release notes) - #148044 (compiletest: show output in debug logging) - #148057 (tests/ui/sanitizer/hwaddress.rs: Run on aarch64 and remove cgu hack) - #148139 (Add `coverage` scope for controlling paths in code coverage) - #148154 (Add a mailmap entry) - #148158 (ci: loongarch64: use medium code model to avoid relocation overflows) - #148166 (Re-enable macro-stepping test for AArch64) - #148172 (rustc-dev-guide subtree update) - #148175 (Fix typos: duplicate words in comments) - #148186 (rustdoc-search: add an integration test for CCI) Failed merges: - #147935 (Add LLVM realtime sanitizer) r? `@ghost` `@rustbot` modify labels: rollup

Rollup of 6 pull requests Successful merges: - #147840 (Rework unsizing coercions in the new solver) - #148139 (Add `coverage` scope for controlling paths in code coverage) - #148154 (Add a mailmap entry) - #148158 (ci: loongarch64: use medium code model to avoid relocation overflows) - #148172 (rustc-dev-guide subtree update) - #148175 (Fix typos: duplicate words in comments) r? `@ghost` `@rustbot` modify labels: rollup

Rollup merge of #147840 - jdonszelmann:unsizing-coercions, r=lcnr Rework unsizing coercions in the new solver Replaces #141926, contains: - a commit adding tests that fail before this work - the two commits from the previous PR - a commit in which these tests are fixed - finally, a fixup for an in my opinion rather large regression in diagnostics. It's still not perfect, but better? I hope this is roughly what you had in mind Fixes rust-lang/trait-system-refactor-initiative#241 and rust-lang/trait-system-refactor-initiative#238, adding tests for both r? ````@lcnr````

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. WG-trait-system-refactor The Rustc Trait System Refactor Initiative (-Znext-solver) labels Oct 18, 2025

rustbot assigned lcnr Oct 18, 2025

This comment has been minimized.

Sign in to view

lcnr changed the title ~~Reowork unsizing coercions in the new solver~~ Rework unsizing coercions in the new solver Oct 19, 2025

lcnr reviewed Oct 20, 2025

View reviewed changes

jdonszelmann force-pushed the unsizing-coercions branch from 9846ace to 316479a Compare October 20, 2025 11:40

This comment has been minimized.

Sign in to view

jdonszelmann force-pushed the unsizing-coercions branch from 316479a to 4b66d2b Compare October 20, 2025 11:53

This comment has been minimized.

Sign in to view

jdonszelmann and others added 3 commits October 20, 2025 14:35

add tests for unsizing coercions (both ICE)

491554d

Rename nit

14a2770

Rework unsizing coercions in new solver

5689477

jdonszelmann force-pushed the unsizing-coercions branch from 4b66d2b to 29f04e0 Compare October 20, 2025 12:36

This comment has been minimized.

Sign in to view

jdonszelmann force-pushed the unsizing-coercions branch from 29f04e0 to f2d7236 Compare October 20, 2025 13:30

lcnr reviewed Oct 24, 2025

View reviewed changes

compiler/rustc_hir_typeck/src/coercion.rs Outdated Show resolved Hide resolved

lcnr reviewed Oct 24, 2025

View reviewed changes

compiler/rustc_hir_typeck/src/coercion.rs Outdated Show resolved Hide resolved

jdonszelmann added 2 commits October 27, 2025 12:00

move old solver coercion code to separate function

349dbec

prove both newly added tests are fixed by the changes

3d9cb04

jdonszelmann force-pushed the unsizing-coercions branch from f2d7236 to 95200bf Compare October 27, 2025 11:01

lcnr reviewed Oct 27, 2025

View reviewed changes

find the right error source when we can't unsize

1756622

jdonszelmann force-pushed the unsizing-coercions branch from 95200bf to 1756622 Compare October 27, 2025 11:27

bors 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 Oct 27, 2025

matthiaskrgr mentioned this pull request Oct 27, 2025

Rollup of 6 pull requests #148178

Open

Zalathar mentioned this pull request Oct 28, 2025

Rollup of 13 pull requests #148195

Closed

Zalathar mentioned this pull request Oct 28, 2025

Rollup of 14 pull requests #148202

Merged

bors merged commit 1cf3dc6 into rust-lang:master Oct 28, 2025
11 checks passed

rustbot added this to the 1.93.0 milestone Oct 28, 2025

This was referenced Oct 28, 2025

unsized coercion: select vs trait goal mismatch rust-lang/trait-system-refactor-initiative#241

Closed

non-fatal overflow for slightly different goals in MIR borrowck results in ICE rust-lang/trait-system-refactor-initiative#238

Closed

Uh oh!

Rework unsizing coercions in the new solver #147840

Rework unsizing coercions in the new solver #147840

Uh oh!

Conversation

jdonszelmann commented Oct 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rustbot commented Oct 18, 2025

Uh oh!

This comment has been minimized.

BoxyUwU commented Oct 18, 2025

Uh oh!

lcnr 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!

lcnr Oct 20, 2025

Choose a reason for hiding this comment

Uh oh!

jdonszelmann Oct 20, 2025

Choose a reason for hiding this comment

Uh oh!

This comment has been minimized.

This comment has been minimized.

rustbot commented Oct 20, 2025

Uh oh!

This comment has been minimized.

This comment has been minimized.

Uh oh!

Uh oh!

jdonszelmann commented Oct 27, 2025

Uh oh!

lcnr Oct 27, 2025

Choose a reason for hiding this comment

Uh oh!

jdonszelmann Oct 27, 2025

Choose a reason for hiding this comment

Uh oh!

lcnr commented Oct 27, 2025

Uh oh!

jdonszelmann commented Oct 27, 2025

Uh oh!

lcnr commented Oct 27, 2025

Uh oh!

bors commented Oct 27, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

jdonszelmann commented Oct 18, 2025 •

edited

Loading

lcnr left a comment •

edited by rustbot

Loading