redcode wrote: > I think your check is correct. I think your comment is not. Dependent types > can have layout T&& for example.
Ooopss... :sweat: > However, I suspect this is a bug in SEMA. It appears that the lambda stays > LDK_Unknown. If so, Sema probably should treat it as dependent, and this > checker guard could become a defensive backstop. A FIXME pointing at the > issue inside ActOnLambdaExpressionAfterIntroducer would be good. As well as a > github issue maybe? Thanks for the review. I'll modify the comment and add a FIXME in `ActOnLambdaExpressionAfterIntroducer` pointing at the issue, and open a separate issue for the Sema bug. One thing: the guard skips records with dependent fields silently. That means it can also mask future Sema bugs that leave a record non-dependent but with a dependent field, turning a loud crash into a silently missed warning. Should I add an `assert` alongside the guard, so this still fires in assert-enabled builds while the guard keeps release builds from crashing? Or is the FIXME plus a separate issue enough tracking for now? https://github.com/llvm/llvm-project/pull/232986 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
