https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90830
--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to zhonghao from comment #0)
> However, I tried clang and icc, and both compilers accept the code. Is the
> bug fully fixed?
They only accept it because your code doesn't instantiate the function
template. I've explained this to you before.
Try instantiating the template
template <int I>
constexpr int f ()
{
enum E { a = f<0> () };
return 0;
}
constexpr int i = f<1>();
Now Clang says:
90830.cc:4:16: error: expression is not an integral constant expression
enum E { a = f<0> () };
^~~~~~~
90830.cc:4:16: note: in instantiation of function template specialization
'f<0>' requested here
90830.cc:7:19: note: in instantiation of function template specialization
'f<1>' requested here
constexpr int i = f<1>();
^
90830.cc:4:16: note: undefined function 'f<0>' cannot be used in a constant
expression
enum E { a = f<0> () };
^
90830.cc:2:15: note: declared here
constexpr int f ()
^
1 error generated.
And EDG says:
"90830.cc", line 4: error: expression must have a constant value
enum E { a = f<0> () };
^
"90830.cc", line 4: note: this call cannot be evaluated because the target
function function "f<I>() [with I=0]" (declared at line 2) is not
constexpr or not completely defined yet
enum E { a = f<0> () };
^
detected during instantiation of "int f<I>() [with I=0]" at line 4
"90830.cc", line 7: warning: variable "i" was declared but never referenced
constexpr int i = f<1>();
^
1 error detected in the compilation of "90830.cc".
Apart from that, *obviously* the code is not valid because it would cause
infinite recursion. That should be obvious.
