Add patch to fix compiler error: narrowing conversion

On ARM, char is unsigned by default

gdcm-2.6-fix-narrowing-error.patch

@@ -0,0 +1,11 @@
+--- gdcm-2.6.3.orig/Testing/Source/DataStructureAndEncodingDefinition/Cxx/TestElement1.cxx	2016-01-27 14:25:39.000000000 +0100
++++ gdcm-2.6.3/Testing/Source/DataStructureAndEncodingDefinition/Cxx/TestElement1.cxx	2016-02-01 22:34:34.559716794 +0100
+@@ -88,7 +88,7 @@
+ 
+ int TestUL()
+ {
+-  const char array[4] = {-78, 1, 0, 0}; // 434
++  const signed char array[4] = {-78, 1, 0, 0}; // 434
+   {
+   Element<VR::UL, VM::VM1> a;
+   // reinterpret_cast< const Element<VR::UL, VM::VM1>& > ( array );

gdcm.spec

@@ -21,6 +21,7 @@ Patch3:	gdcm-2.4.0-no-versioned-dir.patch
 #Patch4: gdcm-0005-support-vtk6.patch
 Patch5: gdcm-2.4.0-find-python27.patch
 Patch6: gdcm-2.6-fix-cmake-config-paths.patch
+Patch7: gdcm-2.6-fix-narrowing-error.patch
 
 BuildRequires:	CharLS-devel >= 1.0
 BuildRequires:	cmake
@@ -120,6 +121,7 @@ used this library with python
 %patch2 -p 1
 %patch3 -p 1
 %patch6 -p 1
+%patch7 -p 1
 
 # Remove bundled utilities (we use Fedora's ones)