https://llvm.org/bugs/show_bug.cgi?id=21837
I hit this recently and decided to investigate a little bit.
I'm still not 100% sure if we can arrive here with Access == AS_none, though.
Details
Details
Diff Detail
Diff Detail
Event Timeline
SemaAccess.cpp | ||
---|---|---|
1361–1362 | Please add an assert that the Target is marked invalid here. I would also think that assuming invalid declarations are accessible is a better default; we don't want additional follow-on errors after the initial diagnostic that resulted in the entity being marked as invalid. (And this comment also applies to the other change.) |
Comment Actions
Hrmm, the assertion you suggested to add (Target is marked invalid) fails, so I think this requires a different fix. I'll try to investigate futher over the next days.
Please add an assert that the Target is marked invalid here. I would also think that assuming invalid declarations are accessible is a better default; we don't want additional follow-on errors after the initial diagnostic that resulted in the entity being marked as invalid.
(And this comment also applies to the other change.)