Change toolset name from gcc-3_4_3-sunos to gcc-3.4.3-sunos for BBv2

[SVN r33334]

status/explicit-failures-markup.xml

@@ -44,7 +44,7 @@
     <mark-toolset name="gcc-4.0.2-linux" status="required"/>
     <mark-toolset name="gcc-3_3-darwin" status="required"/>
     <mark-toolset name="gcc-4_0-darwin" status="required"/>
-    <mark-toolset name="gcc-3_4_3-sunos" status="required"/>
+    <mark-toolset name="gcc-3.4.3-sunos" status="required"/>
     <mark-toolset name="intel-win32-8_1" status="required"/>
     <mark-toolset name="intel-win32-9_0" status="required"/>
     <mark-toolset name="intel-8.1-linux" status="required"/>
@@ -1121,7 +1121,7 @@
 
         <mark-expected-failures>
             <test name="hash_float_test"/>
-            <toolset name="gcc-3_4_3-sunos"/>
+            <toolset name="gcc-3.4.3-sunos"/>
             <note author="Daniel James">
                 On this compiler the hash function is returning the same value
                 for <code>std::numeric_limits&lt;long double&gt;::max()</code>,
@@ -2434,7 +2434,7 @@ for more information.
         <mark-expected-failures>
             <test name="octonion_test"/>
             <test name="quaternion_test"/>
-            <toolset name="gcc-3_4_3-sunos"/>
+            <toolset name="gcc-3.4.3-sunos"/>
             <note author="Caleb Epstein">
               There appears to be a bug in gcc's <code>std::exp (long
               double)</code> on this platform.