[PATCH] [analyzer] [proposal] Fix for PR20653

[Jordan Rose]

Hm, yes, this would make sense. It's still not //complete// because things can //remove// bindings from the Store, which does not make formerly invalid symbols valid again. But it still seems like a strict improvement over the base. Thanks!

Please add some test cases, following the examples in test/Analysis.

================
Comment at: lib/StaticAnalyzer/Core/SymbolManager.cpp:451-452
@@ -451,1 +450,4 @@
+    const MemRegion *MR = cast<SymbolRegionValue>(sym)->getRegion();
+    KnownLive = isLiveRegion(MR) && !reapedStore.getStoreManager()
+        .hasDirectBinding(reapedStore.getStore(), MR);
     break;
----------------
Please reformat to be two separate lines. We try not to break before `.` when we can help it.

http://reviews.llvm.org/D5104