[Bug preprocessor/23827] New: standard C++ should not have hex float preprocessing tokens

["jsm28 at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>]

Standard C++ does not include hexadecimal floating point numbers.  The
preprocessor correctly pedwarns for C++ if such a preprocessing token is
converted to a token, but it incorrectly allows p+ and p- in preprocessing
tokens in standard C++ mode in the first place.  The problem is probably the table

static const struct lang_flags lang_defaults[] =
{ /*              c99 c++ xnum std  //   digr  */
  /* CXX98  */  { 0,  1,  1,   1,   1,   1     },

in libcpp/init.c: xnum should be 0 not 1.  The valid C++ testcases this affects
(compile with -std=c++98 -pedantic-errors, should execute OK, instead aborts)
are along the lines of gcc.dg/c90-hexfloat-2.c:

#define f (
#define l )
#define str(x) #x
#define xstr(x) str(x)

/* C90 and C++: "0x1p+( 0x1p+)"; C99: "0x1p+f 0x1p+l" */
const char *s = xstr(0x1p+f 0x1p+l);

extern "C" void abort (void);
extern "C" int strcmp (const char *, const char *);

int
main (void)
{
  if (strcmp (s, "0x1p+( 0x1p+)"))
    abort ();
  else
    return 0; /* Correct C90 and C++ behavior.  */
}

-- 
           Summary: standard C++ should not have hex float preprocessing
                    tokens
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: preprocessor
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: jsm28 at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23827