check object safety of generic constants #78365
Conversation
rust-highfive
added
the
S-waiting-on-review
lcnr
added
A-const-generics
lcnr
force-pushed
the
const-eval-obj-safety
branch
from
2a7f342 to
0e419ef
Compare
src/test/ui/const-generics/const_evaluatable_checked/object-safety-ok-infer-err.stderr
Show resolved
Hide resolved
|
I think it's fine to merge this as is without waiting on the FCP on that PR to finish |
| false | ||
| fn visit_const(&mut self, ct: &ty::Const<'tcx>) -> bool { | ||
| // First check if the type of this constant references `Self`. | ||
| if self.visit_ty(ct.ty) { |
There was a problem hiding this comment.
Can you also add a test for this? That will need the const generics feature gate so you can have non-usize constants, but that's fine.
There was a problem hiding this comment.
we already check this, [u8; bar::<Self>] only fails because of visit_ty as bar is Const { ty: fn bar with substs [Self], ct: ZST }.
If we did not add this visit_ty here we would never fail for constants, which took me a while to find out
There was a problem hiding this comment.
wait... what? as far as I can tell you only have bar::<Self>(), which is a constant of usize type. Even if its body contains other stuff.
There was a problem hiding this comment.
yeah, meant bar::<Self>(). The issue here is that bar::<Self> is a ZST of type fn() -> usize { bar<Self> }, so we have to look at the type of constants to see if they reference Self. The value of a const cannot reference Self afaik
There was a problem hiding this comment.
I get that, but I don't understand how the constant we are visiting is ever anything but a constant with ty: usize, where the constant's value may be Unevaluated(DefIdOfbar, [Self]), but that value is essentially irrelevant for the visit_ty here. Or is this during the visiting of the MIR of the usize typed constant? I don't think it ever is that, as the visitor we are currently talking about is visiting the signature of the function
There was a problem hiding this comment.
So we are visiting [u8; bar::<Self>()] and then the abstract const for the array length which is
[Leaf(Const { ty: fn() -> usize {bar::<Self>}, val: Value(Scalar(<ZST>)) }), FunctionCall(n0, [])]
So Self is only mentioned in the type of the function object.
Or is this during the visiting of the MIR of the usize typed constant?
During the visiting of the AbstractConst of the usize typed constant
There was a problem hiding this comment.
During the visiting of the AbstractConst of the usize typed constant
Oh, now I realize where my confusion is coming from. It's actually both!
So...
struct Foo<T, const X: T>(T);isn't possible, thus no
trait Foo {
fn bar(&self) -> Foo<Self, something()>;
}
is possible either.
lcnr
force-pushed
the
const-eval-obj-safety
branch
from
c0ab1b1 to
8546a80
Compare
|
@bors r+ |
|
📌 Commit 60bcc58 has been approved by |
bors
added
S-waiting-on-bors
Rollup of 10 pull requests Successful merges: - rust-lang#78152 (Separate unsized locals) - rust-lang#78297 (Suggest calling await on method call and field access) - rust-lang#78351 (Move "mutable thing in const" check from interning to validity) - rust-lang#78365 (check object safety of generic constants) - rust-lang#78379 (Tweak invalid `fn` header and body parsing) - rust-lang#78391 (Add const_fn in generics test) - rust-lang#78401 (resolve: private fields in tuple struct ctor diag) - rust-lang#78408 (Remove tokens from foreign items in `TokenStripper`) - rust-lang#78447 (Fix typo in comment) - rust-lang#78453 (Fix typo in comments) Failed merges: r? `@ghost`
As
Selfcan only be effectively used in constants withconst_evaluatable_checkedthis should not matter outside of it.Implements the first item of #72219
r? @oli-obk for now cc @nikomatsakis