Always record reference to binding in match if guards #78393
Conversation
|
r? @lcnr (rust_highfive has picked a reviewer for you, use r? to override) |
rust-highfive
added
the
S-waiting-on-review
|
r? @tmandry maybe |
| tcx.mk_region(ty::RegionKind::ReErased), | ||
| ty::TypeAndMut { ty, mutbl: hir::Mutability::Not }, | ||
| ); | ||
| self.record(ref_ty, scope, Some(expr), expr.span, guard_borrowing_from_pattern); |
There was a problem hiding this comment.
I think this is the only call to self.record that actually needs to ever pass true for guard_borrowing_from_pattern but I didn't want to make the recording of types any less cautious. If there are perf reasons to not want to record too many types in the generator witness we could rethink that.
There was a problem hiding this comment.
Hmm, I think you're right. We could try disabling the others in a follow-up.
| let tcx = self.fcx.tcx; | ||
| let ref_ty = tcx.mk_ref( | ||
| // Use `ReErased` as `resolve_interior` is going to replace all the regions anyway. | ||
| tcx.mk_region(ty::RegionKind::ReErased), |
There was a problem hiding this comment.
Maybe someone with better knowledge about lifetimes in the compiler should double-check this makes sense.
|
Looks good, thanks for this! @bors r+ By the way, if you're interested in helping out more, I'd be delighted if you came by the wg-async-foundations stream on Zulip. |
|
📌 Commit 8bf9abb has been approved by |
bors
added
S-waiting-on-bors
|
☀️ Test successful - checks-actions |
When encountering a binding from a
matchpattern in itsifguard when computing a generator's interior types, we must always record the type of a reference to the binding because of howifguards are lowered to MIR. This was missed in #75213 because the binding in that test case was autorefed and we recorded that adjusted type anyway.Fixes #78366