[Bug-gsl] gsl expects isnan(nan) to return 1

[sisyphus1]

Hi,

I have this test failure in gsl-1.16 - using a mingw64 port of gcc-4.8.2 
(x64):

###########################
==================================
  gsl 1.16: sys/test-suite.log
==================================

# TOTAL: 1
# PASS:  0
# SKIP:  0
# XFAIL: 0
# FAIL:  1
# XPASS: 0
# ERROR: 0

.. contents:: :depth: 2

FAIL: test
==========

FAIL: gsl_isnan(nan) (-1 observed vs 1 expected) [310]
###########################

This failure occurs because math.h's isnan() has returned "-1" instead of 
the expected "1".
However, I believe (perhaps mistakenly) that this gsl test should be testing 
for "true" not "1".

Interestingly, isnan() on the other mingw64 compilers I have at my disposal 
*does* return "1" for a nan.
(And these other compilers include a 32-bit version of gcc-4.8.2 from the 
same vendor.)

Cheers,
Rob