MSVC2015 does not support a mix of struct and initializer list

initialization so this patch works around the issue by allowing
the initialization of DerivedScale with an initializer list of
float initializer lists. While this enables compilation of TFLite
with MSVC2015 it also has the unfortunate side effect of moving
incorrect initialization of DerivedScale
(e.g `DerivedScale foo = {1, 2};`) from a compile time to a runtime
error.

PiperOrigin-RevId: 329385442
Change-Id: Ia609f2d06cb573c8d791b600501d97aa3efed1f3

tensorflow/lite/tools/optimize/operator_property.h

@@ -26,6 +26,27 @@ namespace operator_property {
 // the scales. For example, for bias in conv, derived_scale = {{0, 1}, {}, {}}
 // and for lstm gate bias, the derived scale is {{}, {0}, {2^-10}}
 struct DerivedScale {
+  // MSVC2015 version 14.0 and below doesn't support struct initialization with
+  // initializer lists so emulate the behavior using a float initializer list.
+#if _MSC_VER <= 1900
+  DerivedScale() {}
+  // Construct this object with a list of initializer lists. All list elements
+  // are cast to float values to avoid ambiguous construction of a union-style
+  // object that could take either std::initializer_list<float> or
+  // std::initializer_list<int>.
+  DerivedScale(std::initializer_list<std::initializer_list<float>> values) {
+    assert(values.size() == 3);
+    std::vector<std::initializer_list<float>> items(values);
+    for (auto& it : items[0]) {
+      input_tensors.push_back(static_cast<int>(it));
+    }
+    for (auto& it : items[1]) {
+      intermediate_tensors.push_back(static_cast<int>(it));
+    }
+    factors.assign(items[2]);
+  }
+#endif  // _MSC_VER <= 1900
+
   std::vector<int> input_tensors = {};
   std::vector<int> intermediate_tensors = {};
   // This is a list of extra factors that are not associated with any other