ANAMASGARD wrote: @TPPPP72 The assertion was based on an invariant that holds during normal lambda parsing, but not during delayed parsing. When this default argument is reparsed, the lambda’s LambdaScopeInfo has already been popped from S.FunctionScopes, while the declaration context still refers to the lambda call operator. Therefore, reaching the end of FunctionScopes is a valid and reproducible state, not an internal compiler error. The guard handles that state, preserves the existing behavior when the scope is present, and allows Clang to continue and emit the normal diagnostics instead of crashing. The regression test verifies this exact behavior.
https://github.com/llvm/llvm-project/pull/213556 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
