Normalize anon consts in new solver #112183
Conversation
rustbot
added
S-waiting-on-review
|
Some changes occurred to the core trait solver cc @rust-lang/initiative-trait-system-refactor |
compiler-errors
changed the title
| @@ -672,7 +672,10 @@ impl<'a, 'b, 'tcx> TypeFolder<TyCtxt<'tcx>> for AssocTypeNormalizer<'a, 'b, 'tcx | |||
| #[instrument(skip(self), level = "debug")] | |||
| fn fold_const(&mut self, constant: ty::Const<'tcx>) -> ty::Const<'tcx> { | |||
| let tcx = self.selcx.tcx(); | |||
| if tcx.lazy_normalization() || !needs_normalization(&constant, self.param_env.reveal()) { | |||
| if tcx.features().generic_const_exprs | |||
| || tcx.trait_solver_next() | |||
There was a problem hiding this comment.
I don't know if it makes sense to do this- we dont stop fold_ty from normalizing things if trait_solver_next is enabled so I wouldn't think it makes sense to avoid normalization of consts.
There was a problem hiding this comment.
It doesn't hurt, though? Like, the only reason we have eager normalization of types currently is because it's difficult to disable because of alias-relate ambiguities (like due to lifetimes primarily).
I could revert this, but it does seem to move us closer to actually-lazy-norm at least for consts?
There was a problem hiding this comment.
I guess i'm a bit nervous about having types/consts diverge especially when it comes to implementing lazy norm since if we only do it for consts it seems more likely that bugs would go hidden since usage of const generics is way rarer than usage of type generics.
There was a problem hiding this comment.
I think i'd want @lcnr to be okay with doing this if this is going to get merged as is
There was a problem hiding this comment.
That's fair, I can revert it and we can discuss "if and how we want to get to actually-lazy-norm-for-consts" in the new solver separately then.
compiler-errors
force-pushed
the
new-solver-anon-ct
branch
2 times, most recently
from
5062967 to
6c1284d
Compare
rust-cloud-vms
bot
force-pushed
the
new-solver-anon-ct
branch
from
ec94d60 to
3144a57
Compare
…IATs, don't partially support const projection for impls
rust-cloud-vms
bot
force-pushed
the
new-solver-anon-ct
branch
from
3144a57 to
84196f3
Compare
|
r=me if CI passes |
|
@bors r=BoxyUwU |
bors
added
S-waiting-on-bors
…mpiler-errors Rollup of 6 pull requests Successful merges: - rust-lang#109609 (Separate AnonConst from ConstBlock in HIR.) - rust-lang#112166 (bootstrap: Rename profile = user to profile = dist) - rust-lang#112168 (Lower `unchecked_div`/`_rem` to MIR's `BinOp::Div`/`Rem`) - rust-lang#112183 (Normalize anon consts in new solver) - rust-lang#112211 (pass `--lib` to `x doc`) - rust-lang#112223 (Don't ICE in new solver when auto traits have associated types) r? `@ghost` `@rustbot` modify labels: rollup
We don't do any of that
expand_abstract_constsstuff so this isn't sufficient to make GCE work, but it does allow, e.g.[(); 1]: Default, to solve.r? @BoxyUwU