FFado 2.0~rc1 FTBFS on IA64

Same as #197 on SPARC: atomic operations and the build system. For atomic operations, I'd go for the gcc-4.1 builtins. We don't have the manpower for everything else.

For the build system, we could simply ditch the default target. Don't know if this works:

Index: SConstruct
===================================================================
--- SConstruct  (revision 1518)
+++ SConstruct  (working copy)
@@ -412,7 +412,7 @@
     elif re.search("powerpc", config[config_cpu]) != None:
         env['DIST_TARGET'] = 'powerpc'
     else:
-        env['DIST_TARGET'] = 'i686'
+        env['DIST_TARGET'] = config[config_cpu]
     print "Detected DIST_TARGET = " + env['DIST_TARGET']
 
 if ((re.search ("i[0-9]86", config[config_cpu]) != None) or (re.search ("x86_64", config[config_cpu]) != None) or (re.search ("powerpc", config[config_cpu]) != None)):

That looks like a sane solution to me, but I am not very experienced with scons.

Would that change anything for users and packagers on x86 and x86_64?

Works fine for me on both, x86 and amd64. No objections to include this fix.

(In [1675]) Lets see how many setups this breaks. see #199

(In [1676]) Make sure that the 2.0 branch breaks too. see #199

Should be fixed both in trunk and the 2.0 branch.