-
Notifications
You must be signed in to change notification settings - Fork 1.9k
fix: Fix mir evaluator not respecting block local trait impls #16299
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
| pub enum StatementKind { | ||
| TraitEnvBlockEnter(hir_def::BlockId), | ||
| TraitEnvBlockExit, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not too happy about having to add extra statements for this, but I can't think of a better way to encode this.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Sorry to be making comments from the sidelines since I don't have the whole context, but it's kind of surprising to me that there's any trait solving / method resolution happening during MIR eval at all 🤔
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It is surprising, but the reason for that is that we currently try to resolve the concrete method of a trait function "late". That is inference result only resolves the trait function, but MIR later wants the function of the actual trait implementation which happens in mir lowering/eval currrently (and in the IDE layer for goto def etc). I guess we should be able to move that logic into type check (and get rid of the lookup_impl_method_query).
Also comments like this are very much appreciated.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can we do it during MIR lowering?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I don't think we can get rid of the lookup_impl_method_query in MIR eval completely without changing the evaulation of dyn Trait which currently only stores the type id in the metadata part, and resolve the impl method on demand. Though I think we can switch to a more direct vtable representation, and then do all the method resolutions in the monomorphization.
Does the current implementation handle block local trait impls for dyn Trait? For example:
trait Foo {
fn foo(&self);
}
fn some_dyn_foo() -> Box<dyn Foo> {
struct St;
impl Foo for St { fn foo(&self) { println!("foo"); } }
Box::new(St)
}
fn main() {
let x = some_dyn_foo();
x.foo();
}If not, is it possible to handle it in future under current architecture?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Though I think we can switch to a more direct vtable representation, and then do all the method resolutions in the monomorphization.
Yes, IMO that's what we should do.
|
☔ The latest upstream changes (presumably #16329) made this pull request unmergeable. Please resolve the merge conflicts. |
No description provided.