fix: Lazy evaluate consts in path_to_const
#14727
Conversation
rustbot
added
the
S-waiting-on-review
| @@ -118,9 +118,12 @@ pub(crate) fn infer_query(db: &dyn HirDatabase, def: DefWithBodyId) -> Arc<Infer | |||
| /// This is appropriate to use only after type-check: it assumes | |||
| /// that normalization will succeed, for example. | |||
| pub(crate) fn normalize(db: &dyn HirDatabase, trait_env: Arc<TraitEnvironment>, ty: Ty) -> Ty { | |||
| if !ty.data(Interner).flags.intersects(TypeFlags::HAS_PROJECTION) { | |||
There was a problem hiding this comment.
I would believe this to ruin performance, could you check that? Because if it does I don't think we want to merge this.
There was a problem hiding this comment.
I changed that, and per the CI time we are good again. But it is not an important part of the PR, we can remove that (at cost of not being able to find layout for struct Something { some_field: [Ty; SOME_CONST] }, the original issue works without this)
HKalbasi
force-pushed
the
mir
branch
2 times, most recently
from
a2e0a61 to
2719beb
Compare
|
@bors r+ |
|
☀️ Test successful - checks-actions |
lnicola
changed the title
path_to_constpath_to_const
|
This increased type mismatched on webrender by 9 |
|
Probably is the same problem of #14844 |
|
Can we revert this before tomorrow's release before until this is fixed? |
|
Type mismatch is an experimental diagnostic, and the problem is easily solvable (I just need to find some time to put on it) so I don't think we should revert this only for false positive in type mismatch. Does it have other negative effects? |
|
Hmmm. I didn't realize that this wasn't a "core" diagnostic. I can live with disabling this diagnostic till it gets fixed :) |
fix #14275