diff --git a/base/include/hidl/HidlSupport.h b/base/include/hidl/HidlSupport.h
index 9343f0f..0a0beeb 100644
--- a/base/include/hidl/HidlSupport.h
+++ b/base/include/hidl/HidlSupport.h
@@ -412,7 +412,7 @@
     }
 
     T *releaseData() {
-        if (!mOwnsBuffer && mSize > 0) {
+        if (!mOwnsBuffer && mBuffer != nullptr) {
             resize(mSize);
         }
         mOwnsBuffer = false;
diff --git a/test_main.cpp b/test_main.cpp
index e4cdd29..0a1e97b 100644
--- a/test_main.cpp
+++ b/test_main.cpp
@@ -222,6 +222,24 @@
     EXPECT_ARRAYEQ(v3, array, v3.size());
 }
 
+TEST_F(LibHidlTest, VecReleaseTest) {
+    // this test indicates an inconsistency of behaviors which is undesirable.
+    // Perhaps hidl-vec should always allocate an empty vector whenever it
+    // exposes its data. Alternatively, perhaps it should always free/reject
+    // empty vectors and always return nullptr for this state. While this second
+    // alternative is faster, it makes client code harder to write, and it would
+    // break existing client code.
+    using android::hardware::hidl_vec;
+
+    hidl_vec<int32_t> empty;
+    EXPECT_EQ(nullptr, empty.releaseData());
+
+    empty.resize(0);
+    int32_t* data = empty.releaseData();
+    EXPECT_NE(nullptr, data);
+    delete data;
+}
+
 TEST_F(LibHidlTest, VecIterTest) {
     int32_t array[] = {5, 6, 7};
     android::hardware::hidl_vec<int32_t> hv1 = std::vector<int32_t>(array, array + 3);
diff --git a/transport/base/1.0/vts/functional/Android.bp b/transport/base/1.0/vts/functional/Android.bp
index 6f91b04..3f1c239 100644
--- a/transport/base/1.0/vts/functional/Android.bp
+++ b/transport/base/1.0/vts/functional/Android.bp
@@ -34,7 +34,7 @@
         "libinit_test_utils",
     ],
     test_suites: [
-        "vts-core",
+        "vts",
     ],
     require_root: true,
     auto_gen_config: true,