For parallel_for_each iterator tag dispatch for move iterators + fix stack overflow on Windows

include/oneapi/tbb/parallel_for_each.h

@@ -409,12 +409,31 @@ using tag = typename std::iterator_traits<It>::iterator_category;
 
 #if __TBB_CPP20_PRESENT
 template <typename It>
-using iterator_tag_dispatch =
+struct move_iterator_dispatch_helper {
+    using type = It;
+};
+
+// Until C++23, std::move_iterator::iterator_concept always defines
+// to std::input_iterator_tag and hence std::forward_iterator concept
+// always evaluates to false, so std::move_iterator dispatch should be
+// made according to the base iterator type.
+template <typename It>
+struct move_iterator_dispatch_helper<std::move_iterator<It>> {
+    using type = It;
+};
+
+template <typename It>
+using iterator_tag_dispatch_impl =
     std::conditional_t<std::random_access_iterator<It>,
                        std::random_access_iterator_tag,
                        std::conditional_t<std::forward_iterator<It>,
                                           std::forward_iterator_tag,
                                           std::input_iterator_tag>>;
+
+template <typename It>
+using iterator_tag_dispatch =
+    iterator_tag_dispatch_impl<typename move_iterator_dispatch_helper<It>::type>;
+
 #else
 template<typename It>
 using iterator_tag_dispatch = typename

test/conformance/conformance_parallel_for_each.cpp

@@ -102,10 +102,8 @@ class ForEachInvokeItem {
     void do_action_and_feed(oneapi::tbb::feeder<ForEachInvokeItem>& feeder) const {
         CHECK_MESSAGE(change_vector.size() % 2 == 0, "incorrect test setup");
         std::size_t shift = change_vector.size() / 2;
-        std::cout << "Process " << real_value << std::endl;
         ++change_vector[real_value];
         if (real_value < shift) {
-            std::cout << "Add " << real_value + shift << std::endl;
             feeder.add(ForEachInvokeItem(real_value + shift, change_vector));
         }
     }

test/tbb/test_parallel_for_each.cpp

@@ -290,10 +290,10 @@ template <typename Category>
 void test_with_cpp20_iterator() {
     constexpr std::size_t n = 1'000'000;
 
-    no_copy_move elements[n];
+    std::vector<no_copy_move> elements(n);
 
-    cpp20_iterator<no_copy_move, Category> begin(elements);
-    cpp20_iterator<no_copy_move, Category> end(elements + n);
+    cpp20_iterator<no_copy_move, Category> begin(elements.data());
+    cpp20_iterator<no_copy_move, Category> end(elements.data() + n);
 
     oneapi::tbb::parallel_for_each(begin, end, [](no_copy_move& element) {
         element.item = 42;