7

The following code compiles in both GCC and Clang:

long double operator""_a(long double);     
auto x = 0e1_a+0; // OK

But not this (replacing _a with _e):

long double operator""_e(long double);
auto y = 0e1_e+0; // Error: unable to find numeric literal operator 'operator""_e+0'

OTOH, this code compiles:

auto z = 0e1_e +0; // OK

What's going on?

(This question is inspired by this GCC bug report.)

T.C.
  • 133,968
  • 17
  • 288
  • 421

1 Answers1

14

Maximal munch strikes again.

[lex.pptoken]/p3:

If the input stream has been parsed into preprocessing tokens up to a given character:

  • [two exceptions not relevant here]
  • Otherwise, the next preprocessing token is the longest sequence of characters that could constitute a preprocessing token, even if that would cause further lexical analysis to fail, except that a header-name (2.8) is only formed within a #include directive (16.2).

The problem is that 0e1_e+0, unlike 0e1_a+0, is a valid preprocessing number ([lex.ppnumber]):

pp-number:
    digit
    . digit
    pp-number digit
    pp-number identifier-nondigit
    pp-number ’ digit
    pp-number ’ nondigit
    pp-number e sign
    pp-number E sign
    pp-number .

As a result, 0e1_e+0 is parsed as a single pp-number preprocessing token, and then explodes later because it cannot be converted to a valid token (for obvious reasons).

0e1_a+0, on the other hand, is parsed as three tokens, 0e1_a, +, and 0, and all is well.

T.C.
  • 133,968
  • 17
  • 288
  • 421