xplor-nih testsuite regressions with gcc 4.4

[Jack Howarth]

   I noticed this evening that when building xplor-nih (http://nmr.cit.nih.gov/xplor-nih/)
with current gcc trunk that we now have 22 additional testsuite failures in xplor-nih (out
of 133 tests) compared to the the results when xplor-nih is built with gcc 4.3.2. These failures
all seem to be related to the inability of xplor-nih is process its scripting language when
built with gcc 4.4. For example, the xtarget.inp testscript produces the error...

@@ -62,59 +62,75 @@
  XREFIN: allocating space for    1000 reflections.
  XREFINE> reflections 
  REFLection> INDEx    0   2   0  Fobs  4171.8999 
- XRRR2:      729 new reflections have been added.
- XREFINE> 
- XREFINE>   fwindow 0.01 100000. 
- XREFINE>   eval ($lr=10.) 
+ %REFLection-ERR: unrecognized command:
+    fwindow 
+    ^^^^^^^
+ %REFLection-ERR: unrecognized command:
+    fwindow 0.01 
+            ^^^^

compared to a log of the normal xtarget.inp testsuite run. I'll see if I can pinpoint
where this is happening in xplor-nih but suspect the problem is some flaw introduced
into the libgfortran io routines.
                Jack