[PATCH] IR: Move the complex address expression field out of DIVariable and into an extra argument of the dbg.declare/dbg.value intrinsics.

[David Blaikie]

Permissions change on test/BugPoint/compile-custom.ll, test/MC/COFF/linker-options.ll, test/Verifier/inalloca-vararg.ll - what's that about?

These tests would probably be easier to review if we made a separate change to add a no-op 4th operand to the intrinsic, and committed that along with the test case updates necessary - then did the change to catually move stuff. But I'm not sure if that's worthwhile - just a thought, not a request.

================
Comment at: lib/CodeGen/AsmPrinter/DwarfUnit.cpp:618
@@ +617,3 @@
+    } else if (DV.getVariable().isIndirect())
+      addRegisterOffset(*Loc, Location.getReg(), 0);
+    else
----------------
How come this turned into an explicit special case? I assume in the old code this was handled by the code on 618 (in the before source)?

================
Comment at: test/DebugInfo/X86/autoupgrade-intrinsics.s:1
@@ +1,2 @@
+	.section	__TEXT,__text,regular,pure_instructions
+	.macosx_version_min 14, 0
----------------
What's this file for? Is it referenced from some test?

http://reviews.llvm.org/D4919