HighCommander4 wrote: > The failures are related to new diagnostics that are issued as a side effect > of running additional checks in BuildDeclarationNameExpr().
A closer investigation has revealed that this diagnosis wasn't quite accurate. The three `SemaCXX` failures were not in fact caused by new diagnostics issued as part of doing extra semantic analysis in `BuildDeclarationNameExpr()`. Rather, the original version of the patch had the unintended effect of having [this overload](https://searchfox.org/llvm/rev/9b832b726c9c9bb0672c5f0912f6f131e3e27a10/clang/lib/Sema/SemaExpr.cpp#3431) of `BuildDeclarationNameExpr()` construct an `UnresolvedLookupExpr` even when the only lookup result was an invalid decl, where previously it would fail in that scenario. In my updated patch, I made an adjustment to this function to behave similarly to the other overload, i.e. wrap the resulting expression in a `RecoveryExpr` in the invalid-decl case. This fixed all three `SemaCXX` failures. The `OpenMP` failure remains, though there too it's not the case that the extra diagnostic is **directly** produced by the extra checks that now run in `BuildDeclarationNameExpr()`. Rather, I suspect that the extra semantic analysis we're doing **indirectly** leads to extra diagnostics. I haven't yet tracked down the chain of causation there. https://github.com/llvm/llvm-project/pull/81662 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
