================ Comment at: lib/Sema/SemaExprCXX.cpp:5783-5784 @@ -5769,1 +5782,4 @@ +// If the variable's initializer is dependent, analyze it for any constructs +// that we can use to conclude that the initializer can never be a constexpr. + ---------------- Richard Smith wrote: > This isn't quite correct: you need to avoid walking into subexpressions that > might not be evaluated (operands of `?:`, RHS of `&&` and `||`, ....). > > Please take this out of the patch for now, we can review this diagnostic/QoI > improvement separately once we have the main functionality landed. (This > seems like useful functionality for other diagnostics too.) Aah - Yes - will expunge it. I agree - it would be nice to have this functionality - perhaps we can work on this afterwards!
http://llvm-reviews.chandlerc.com/D2035 _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
