Mistaken change in GCC (fwd)

[Gerald Pfeifer <pfeifer at dbai dot tuwien dot ac dot at>]

I am forwarding this comment which RMS sent to the SC.

Neil, Zack, I think that's most probably your area of expertise?

Gerald

---------- Forwarded message ----------
A recent change in GCC introduced this problem:

      $ cat k.c
      #if #cpu (i386) || #cpu (sparc) || #cpu (m68k) || #cpu (alpha)
      #define GC_SETJMP_WORKS 1
      #endif
      $ gcc -E k.c > /dev/null
      $ gcc -traditional -E k.c > /dev/null
      k.c:1: Invalid token in expression
      [Exit 1]

This change should be undone, because it was based on a mistaken idea
of the meaning of --traditional.

The purpose of --traditional is to make valid old-style code work.  It
should not make newer constructs fail, unless that is the only way to
make valid old-style code work.  However, it is ok for --pedantic
when used with --traditional to cause warnings for them.

This is not just wrong in the abstract, it also cause trouble for
Emacs.  Hence I am cc'ing the Emacs maintainer.