Avoid cat_expr Erred notes when already in error state. #22767
Conversation
Also, to ensure we do not let the dropck get skipped, ICE if cat_expr errors when *not* in error state. This is not known to be a breaking change (i.e. I do not know of a current case that causes the new ICE to be exercised).
|
r? @Aatch (rust_highfive has picked a reviewer for you, use r? to override) |
|
wait, that bit about "this is not known to be a breaking change" is a clear lie: the original test case that this was filed against (in #22265) exercises the ICE. |
|
(closing; I want to investigate the case that is ICE'ing before trying to land this.) |
|
I am mildly concerned that this might fail in some cases where there are unresolved types that have not been reported by the time regionck runs, but if/when we observe that we should probably fix by just caching and reporting those cases earlier. I agree that the potential for bad things to slip through the cracks is high and should be avoided. |
|
⌛ Testing commit 92bc3ea with merge f8baa57... |
|
⌛ Testing commit 92bc3ea with merge ec865de... |
|
💔 Test failed - auto-win-32-nopt-t |
|
@bors: retry |
Avoid `cat_expr Erred` notes when already in error state. Also, to ensure we do not let the dropck get skipped, ICE if `cat_expr` errors when *not* in error state. This is not known to be a breaking change (i.e. I do not know of a current case that causes the new ICE to be exercised). Fix #22265
Avoid
cat_expr Errednotes when already in error state.Also, to ensure we do not let the dropck get skipped, ICE if
cat_exprerrors when not in error state.This is not known to be a breaking change (i.e. I do not know of a current case that causes the new ICE to be exercised).
Fix #22265