mangling of array new

[jicama]

I notice that the mangling of new-expressions is specified as

           ::= [gs] nw <expression>* _ <type> E                     # new (expr-list) type
           ::= [gs] nw <expression>* _ <type> <initializer>         # new (expr-list) type (init)
           ::= [gs] na <expression>* _ <type> E                     # new[] (expr-list) type
           ::= [gs] na <expression>* _ <type> <initializer>         # new[] (expr-list) type (init)

The new[] mangling doesn't clarify how the length of the array allocation is encoded. GCC encodes it into the type by building an array type; Clang seems to not encode it at all. So, for

template <unsigned> struct A { };
template<typename T>
auto foo() -> A<sizeof(new T[42])>
{ return {}; }
int main() { foo<int>(); }

GCC mangles foo as _Z3fooIiE1AIXszna_A42_T_EEEv,
Clang as _Z3fooIiE1AIXszna_T_EEEv

Any objection to clarifying that the GCC mangling is correct?

[rjmccall]

Abstractly, it probably ought to be mangled as a separate operand, but since GCC's choice is perfectly functional here, I agree that the ABI should just bless it. Looks like GCC does properly follow the array-type mangling for instantiation-independent vs. -dependent size expressions. Mind writing up a patch?