================
@@ -2922,7 +2922,7 @@ static bool handleFloatFloatBinOp(EvalInfo &Info, const
BinaryOperator *E,
// If during the evaluation of an expression, the result is not
// mathematically defined [...], the behavior is undefined.
// FIXME: C++ rules require us to not conform to IEEE 754 here.
- if (LHS.isNaN()) {
+ if (!Info.getLangOpts().CPlusPlus23 && LHS.isNaN()) {
----------------
zahiraam wrote:
> My understanding of the intended design was that operations producing NaNs
> from non-NaNs in C++ constexpr cause the expression to not be constant. That
> said, NaNs are still in play because of `numeric_limits`.
>
> For `frexp` though, I think there being an "unspecified" value stored for the
> exponent is a problem where the design could use more clarity. I would think
> that the operation could still be performed, but the "unspecified" value
> should probably be considered indeterminate.
For C it looks like it's returning INT_MAX and INT__MIN. See comment/code here:
https://github.com/llvm/llvm-project/blob/main/llvm/include/llvm/ADT/APFloat.h#L1371
Do we want to do the same for C++23? If not what's exactly an "unspecified"
value?
https://github.com/llvm/llvm-project/pull/88978
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
