[expr.prim.req.nested] misuse of term "unevaluated operand"

[zygoloid]

[expr.prim.req.nested]p2: "A local parameter shall only appear as an unevaluated operand (7.2) within the constraint-expression."

This allows requires (T a) { requires sizeof(a) == 4; } but does not allow requires (T a) { requires sizeof(f(a)) == 4; }, because the term "unevaluated operand" only applies to the immediate operand of various contexts. I believe the intent here is "An unqualified-id naming a local parameter shall not be potentially evaluated." or similar.

[zygoloid]

This also seems wrong:

[expr.prim.req.general]p2: "Expressions appearing within a requirement-body are unevaluated operands (7.2)."

Taken literally that would presumably transitively cover all expressions lexically within a requirement-body, allowing requires (int n) { array<int, n>(); }, because n here would be an unevaluated operand.

Even if we interpret this to only apply at the top level and not transitively (which I assume was the intent), this also breaks [expr.prim.req.nested]p2: requires (bool b) { requires b; } is apparently valid, because (the second occurrence of) b here is an unevaluated operand per [expr.prim.req.general]p2 and hence is permitted by [expr.prim.req.nested]p2.

I think we need an explicit list of the places that are treated as unevaluated operands: I think the expression in a simple-requirement or compound-requirement should be, and expressions appearing elsewhere in a requires-expression should receive no special treatment.