It is also happening on powerpc-linux-gnu:
http://gcc.gnu.org/ml/gcc-testresults/2008-05/msg02422.html

The tests fail on powerpc*-unknown-linux-gnu with 128-bit long doubles because __builtin_fpclassify claims that __LDBL_MAX__ is infinite, not normal.  Here's a small testcase minimized from type-generic-1.c:

#define FP_NAN 1
#define FP_INFINITE 2
#define FP_NORMAL 3
#define FP_SUBNORMAL 4
#define FP_ZERO 5
#define fpclassify(X) __builtin_fpclassify(FP_NAN, FP_INFINITE, FP_NORMAL, FP_SUBNORMAL, FP_ZERO, (X))

int
main ()
{
  volatile long double ld;

  ld = __LDBL_MAX__;
  if (fpclassify (ld) != FP_NORMAL)
    __builtin_abort ();

  return 0;
}

Subject: Bug 36332

Author: jakub
Date: Tue Sep  2 18:46:29 2008
New Revision: 139906

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=139906
Log:
	PR target/36332
	* real.c (real_maxval): Clear a lower bit to make real_maxval
	match get_max_float for IBM long double format.

	* gcc.c-torture/execute/ieee/pr36332.c: New test.

Added:
    trunk/gcc/testsuite/gcc.c-torture/execute/ieee/pr36332.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/real.c
    trunk/gcc/testsuite/ChangeLog

Subject: Bug 36332

Author: jakub
Date: Tue Sep  2 19:11:55 2008
New Revision: 139907

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=139907
Log:
	PR target/36332
	* real.c (real_maxval): Clear a lower bit to make real_maxval
	match get_max_float for IBM long double format.

	* gcc.c-torture/execute/ieee/pr36332.c: New test.

Added:
    branches/gcc-4_3-branch/gcc/testsuite/gcc.c-torture/execute/ieee/pr36332.c
Modified:
    branches/gcc-4_3-branch/gcc/ChangeLog
    branches/gcc-4_3-branch/gcc/real.c
    branches/gcc-4_3-branch/gcc/testsuite/ChangeLog

Fixed.