Consider:

  #if 1 ? 1, 0: 3
  #error
  #endif

This seems to be de-jure well-formed, but implementations uniformly reject this (gcc and clang only in pedantic mode) due to a comma operator appearing in the controlling expression of the #if.

The specification status appears to be the same in C++ and C, so SG22 should offer an opinion whether it is actually a correct interpretation of the respective standard that both C and C++ reject this snippet, and if there is some appetite to rectify the specification (as opposed to fixing the bugs in the implementations).

See CWG3017 for details.