bjoernager/rust
bjoernager/rust
1
Fork 0
Code Activity

Rollup merge of #130273 - lcnr:overflow-no-constraints, r=compiler-errors

Browse source 
more eagerly discard constraints on overflow

We always discard the results of overflowing goals inside of the trait solver. We previously did so when instantiating the response in `evaluate_goal`. Canonicalizing results only to later discard them is also  inefficient 🤷

It's simpler and nicer to debug to eagerly discard constraints inside of the query itself.

r? ``@compiler-errors``
This commit is contained in:
Matthias Krüger 2024-09-12 19:03:43 +02:00 • committed by GitHub
commit cb1d80d1e5
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
5 changed files with 27 additions and 35 deletions
Download patch file Download diff file Expand all files Collapse all files

2
compiler/rustc_trait_selection/src/solve/fulfill.rs
View file

@ -302,7 +302,7 @@ fn fulfillment_error_for_stalled<'tcx>(
Ok((_, Certainty::Maybe(MaybeCause::Overflow { suggest_increasing_limit }))) => (
FulfillmentErrorCode::Ambiguity { overflow: Some(suggest_increasing_limit) },
// Don't look into overflows because we treat overflows weirdly anyways.
// In `instantiate_response_discarding_overflow` we set `has_changed = false`,
// We discard the inference constraints from overflowing goals, so
// recomputing the goal again during `find_best_leaf_obligation` may apply
// inference guidance that makes other goals go from ambig -> pass, for example.
//