diff --git a/mlir/lib/IR/Operation.cpp b/mlir/lib/IR/Operation.cpp
index b5a7a20bdf7e..bc8947c7ae73 100644
--- a/mlir/lib/IR/Operation.cpp
+++ b/mlir/lib/IR/Operation.cpp
@@ -776,7 +776,7 @@ static LogicalResult verifyShapeMatch(Type type1, Type type2) {
 
   // Check unranked tensor cases
   if (type1.isa<UnrankedTensorType>() || type2.isa<UnrankedTensorType>())
-    return failure();
+    return success();
 
   // Check normal vector/tensor cases
   if (auto sType1 = type1.dyn_cast<ShapedType>()) {
diff --git a/mlir/unittests/IR/OpDefinitionTest.cpp b/mlir/unittests/IR/OpDefinitionTest.cpp
index 40740097eca9..dcc83ac6bf9e 100644
--- a/mlir/unittests/IR/OpDefinitionTest.cpp
+++ b/mlir/unittests/IR/OpDefinitionTest.cpp
@@ -25,13 +25,14 @@ using namespace mlir::OpTrait::impl;
 
 namespace {
 
-// TODO: Replace with regular test once this trait is used by operation in core.
-TEST(OpDefinitionTest, SameOperandAndResultElementType) {
-  MLIRContext context;
 #define FILE_LOC                                                               \
   FileLineColLoc::get(UniquedFilename::get(__FILE__, &context), __LINE__, 0,   \
                       &context)
 
+// TODO: Replace with regular test once this trait is used by operation in core.
+// TODO(b/132891206): Replace with dialect test.
+TEST(OpDefinitionTest, SameOperandAndResultElementType) {
+  MLIRContext context;
   Builder b(&context);
   auto *operandtF32x10x10 = Operation::create(
       FILE_LOC, OperationName("some_const", &context), /*operands=*/{},
@@ -84,8 +85,47 @@ TEST(OpDefinitionTest, SameOperandAndResultElementType) {
                     b.getTensorType({5}, b.getF32Type())));
   EXPECT_FALSE(valid(FILE_LOC, operandtI32x1, operandvF32x1,
                      b.getTensorType({5}, b.getF32Type())));
-
-#undef FILE_LOC
 }
 
+TEST(OpDefinitionTest, SameOperandAndResultShape) {
+  MLIRContext context;
+  Builder b(&context);
+  auto *operandtF32x10x10 = Operation::create(
+      FILE_LOC, OperationName("some_const", &context), /*operands=*/{},
+      /*resultTypes=*/{b.getTensorType({10, 10}, b.getF32Type())},
+      /*attributes=*/llvm::None, /*successors=*/{}, /*numRegions=*/0,
+      /*resizableOperandList=*/false, &context);
+  auto *operandtF32x1 = Operation::create(
+      FILE_LOC, OperationName("some_const", &context), /*operands=*/{},
+      /*resultTypes=*/{b.getTensorType({1}, b.getF32Type())},
+      /*attributes=*/llvm::None, /*successors=*/{}, /*numRegions=*/0,
+      /*resizableOperandList=*/false, &context);
+  auto *operandtF32xunranked = Operation::create(
+      FILE_LOC, OperationName("some_const", &context), /*operands=*/{},
+      /*resultTypes=*/{b.getTensorType(b.getF32Type())},
+      /*attributes=*/llvm::None, /*successors=*/{}, /*numRegions=*/0,
+      /*resizableOperandList=*/false, &context);
+
+  // SameOperandAndResultShape trait.
+  auto valid = [&](Location loc, Operation *x, Operation *y, Type resultType) {
+    auto op = Operation::create(loc, OperationName("some_op", &context),
+                                /*operands=*/{x->getResult(0), y->getResult(0)},
+                                /*resultTypes=*/{resultType},
+                                /*attributes=*/llvm::None, /*successors=*/{},
+                                /*numRegions=*/0,
+                                /*resizableOperandList=*/false, &context);
+    return succeeded(verifySameOperandsAndResultShape(op));
+  };
+
+  EXPECT_TRUE(valid(FILE_LOC, operandtF32x1, operandtF32x1,
+                    b.getTensorType({1}, b.getF32Type())));
+  EXPECT_FALSE(valid(FILE_LOC, operandtF32x1, operandtF32x1,
+                     b.getTensorType({12}, b.getF32Type())));
+  EXPECT_FALSE(valid(FILE_LOC, operandtF32x1, operandtF32x10x10,
+                     b.getTensorType({1}, b.getF32Type())));
+  EXPECT_TRUE(valid(FILE_LOC, operandtF32x1, operandtF32xunranked,
+                    b.getTensorType({1}, b.getF32Type())));
+}
+
+#undef FILE_LOC
 } // end namespace