aaron.ballman added a comment. I'm seeing precommit CI failures:
Failed Tests (1): Clang :: SemaCXX/constant-expression-cxx2b.cpp ================ Comment at: clang/lib/Sema/SemaDeclCXX.cpp:1905 + if (SemaRef.LangOpts.CPlusPlus2b) { + if (!VD->getType()->isLiteralType(SemaRef.Context)) + SemaRef.Diag(VD->getLocation(), ---------------- cor3ntin wrote: > hubert.reinterpretcast wrote: > > This seems to trigger even when the type is dependent: > > ``` > > <stdin>:1:36: warning: definition of a variable of non-literal type in a > > constexpr function is incompatible with C++ standards before C++2b > > [-Wpre-c++2b-compat] > > auto qq = [](auto x) { decltype(x) n; }; > > ^ > > 1 warning generated. > > ``` > > > > This also seems to emit even when `Kind` is not > > `Sema::CheckConstexprKind::Diagnose` (unlike the `static`/`thread_local` > > case above). Is the `CheckLiteralType` logic not reusable for this? > You are right, thanks for noticing that, it was rather bogus. > The reason I'm not using CheckLiteralType is to avoid duplicating a > diagnostics message, as CheckLiteralType doesn't allow us to pass parameter > to the diagnostic message. > > It leaves us with an uncovered scenario though: We do not emit the warning on > template instantiation, and I don't think there is an easy way to do that. > The reason I'm not using CheckLiteralType is to avoid duplicating a > diagnostics message, as CheckLiteralType doesn't allow us to pass parameter > to the diagnostic message. Huh? ``` static bool CheckLiteralType(Sema &SemaRef, Sema::CheckConstexprKind Kind, SourceLocation Loc, QualType T, unsigned DiagID, Ts &&...DiagArgs) { ... } ``` I would hope `DiagArgs` should do exactly that? :-) Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D122249/new/ https://reviews.llvm.org/D122249 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits