[PATCH] Use an 'l' prefix when an 'L' prefix would produce an invalid MachO

[kledzik at apple.com]

  These test case changes look like you just fixed the cases that failed with your change.  Perhaps you should add a test case that explicitly tests the label generated by a few kinds of GVs.

  Thanks for looking into and fixing this!


================
Comment at: lib/CodeGen/TargetLoweringObjectFileImpl.cpp:539
@@ +538,3 @@
+    // FIXME: is this really the case? Can we check this without using the
+    // names?
+    if (SMO.getSegmentName() == "__DATA" &&
----------------
Yes, there are a bunch of sections that the linker recognizes by name.  __DATA/__cfstring is one of them.  


http://llvm-reviews.chandlerc.com/D2727