diff --git a/lib/Transforms/Scalar/SimplifyLibCalls.cpp b/lib/Transforms/Scalar/SimplifyLibCalls.cpp
index b878e4b4790..c84c23340fc 100644
--- a/lib/Transforms/Scalar/SimplifyLibCalls.cpp
+++ b/lib/Transforms/Scalar/SimplifyLibCalls.cpp
@@ -514,11 +514,11 @@ struct VISIBILITY_HIDDEN StrCatOpt : public LibCallOptimization {
     // Now that we have the destination's length, we must index into the
     // destination's pointer to get the actual memcpy destination (end of
     // the string .. we're concatenating).
-    Dst = B.CreateGEP(Dst, DstLen, "endptr");
+    Value *CpyDst = B.CreateGEP(Dst, DstLen, "endptr");
     
     // We have enough information to now generate the memcpy call to do the
     // concatenation for us.  Make a memcpy to copy the nul byte with align = 1.
-    EmitMemCpy(Dst, Src, ConstantInt::get(TD->getIntPtrType(), Len+1), 1, B);
+    EmitMemCpy(CpyDst, Src, ConstantInt::get(TD->getIntPtrType(), Len+1), 1, B);
     return Dst;
   }
 };
diff --git a/test/Transforms/SimplifyLibCalls/StrCat.ll b/test/Transforms/SimplifyLibCalls/StrCat.ll
index a6ef6771b46..89ef4ea49a7 100644
--- a/test/Transforms/SimplifyLibCalls/StrCat.ll
+++ b/test/Transforms/SimplifyLibCalls/StrCat.ll
@@ -1,6 +1,9 @@
 ; Test that the StrCatOptimizer works correctly
+; PR3661
 ; RUN: llvm-as < %s | opt -simplify-libcalls | llvm-dis | \
 ; RUN:   not grep {call.*strcat}
+; RUN: llvm-as < %s | opt -simplify-libcalls | llvm-dis | \
+; RUN:   grep {puts.*%arg1}
 
 @hello = constant [6 x i8] c"hello\00"		; <[6 x i8]*> [#uses=1]
 @null = constant [1 x i8] zeroinitializer		; <[1 x i8]*> [#uses=1]