many but not all clang fixes

benchmarks/graph/benchmark_straight_line.cpp

@@ -55,7 +55,7 @@ class StraightLineBenchmark : public benchmarks::BenchmarkBase
 public:
     //! We need to create @c Kokkos objects in the @ref SetUp, not using the constructor or in-class default member initializers.
     void SetUp(const ::benchmark::State&) override {
-        this->exec = Kokkos::Experimental::partition_space(execution_space{}, 1)[0];
+        this->exec = Kokkos::Experimental::partition_space(execution_space{}, 1).at(0);
         this->data = view_t(Kokkos::view_alloc(*exec));
     }
 

tests/exec/adaptors/test_split.cpp

@@ -31,7 +31,7 @@ namespace tests::stdexec::adaptors {
 
 //! @test Simple test for @c exec::split.
 TEST(stdexec, split) {
-    exec::static_thread_pool pool{2};
+    exec::static_thread_pool pool{2}; // NOLINT(misc-const-correctness)
 
     ::stdexec::scheduler auto scheduler = pool.get_scheduler();
 

tests/graph/complex_dag/test_stdexec.cpp

@@ -43,7 +43,7 @@ TEST(graph, complex_dag_stdexec)
     using view_t = Kokkos::View<int[size], memory_space>;
 
     //! Get some execution context.
-    ::exec::static_thread_pool pool{1};
+    ::exec::static_thread_pool pool{1}; // NOLINT(misc-const-correctness)
 
     //! Initialize the data.
     const view_t data(Kokkos::view_alloc("data"));

tests/graph/runtime/test_stdexec.cpp

@@ -74,7 +74,7 @@ TEST_P(GraphTest, runtime_stdexec)
     const auto add_C = this->GetParam().at(1);
 
     //! Get some execution context.
-    ::exec::static_thread_pool pool{1};
+    ::exec::static_thread_pool pool{1}; // NOLINT(misc-const-correctness)
 
     //! Initialize the data.
     const view_t data(Kokkos::view_alloc("data"));

tests/graph/while/test_while.cpp

@@ -81,7 +81,7 @@ class WhileTest : public ::testing::Test
 public:
     void SetUp() override
     {
-        this->exec = Kokkos::Experimental::partition_space(Kokkos::Cuda{}, 1)[0];
+        this->exec = Kokkos::Experimental::partition_space(Kokkos::Cuda{}, 1).at(0);
         this->data = data_t(Kokkos::view_alloc(*exec, "data"), size);
     }
 

tests/kokkos_ext/execution_space/test_fork_join.cpp

@@ -51,7 +51,7 @@ class ForkJoinTest
 TEST_F(ForkJoinTest, diamond) {
     const view_s_t data(Kokkos::view_alloc("data - shared space"));
 
-    ::exec::static_thread_pool pool{4};
+    ::exec::static_thread_pool pool{4}; // NOLINT(misc-const-correctness)
     const context_t esc{exec};
 
     auto chain =

tests/kokkos_ext/execution_space/test_inter_op.cpp

@@ -149,7 +149,7 @@ TEST_F(InterOpTest, transition_to_static_thread_pool) {
 
     const context_t esc{exec};
 
-    ::exec::static_thread_pool pool{1};
+    ::exec::static_thread_pool pool{1}; // NOLINT(misc-const-correctness)
 
     SHOW_EXEC_SPACE_ID(exec)
 
@@ -182,7 +182,7 @@ TEST_F(InterOpTest, transition_from_static_thread_pool) {
 
     const context_t esc{exec};
 
-    ::exec::static_thread_pool pool{1};
+    ::exec::static_thread_pool pool{1}; // NOLINT(misc-const-correctness)
 
     SHOW_EXEC_SPACE_ID(exec)
 
@@ -215,7 +215,7 @@ TEST_F(InterOpTest, transition_from_static_thread_pool_and_back) {
 
     const context_t esc{exec};
 
-    ::exec::static_thread_pool pool{1};
+    ::exec::static_thread_pool pool{1}; // NOLINT(misc-const-correctness)
 
     SHOW_EXEC_SPACE_ID(exec)
 
@@ -251,7 +251,7 @@ TEST_F(InterOpTest, transition_to_static_thread_pool_and_back) {
 
     const context_t esc{exec};
 
-    ::exec::static_thread_pool pool{1};
+    ::exec::static_thread_pool pool{1}; // NOLINT(misc-const-correctness)
 
     SHOW_EXEC_SPACE_ID(exec)
 

tests/kokkos_ext/execution_space/test_split.cpp

@@ -62,7 +62,7 @@ TEST_F(SplitTest, split_and_sync_wait) {
 TEST_F(SplitTest, within) {
     const view_s_t data(Kokkos::view_alloc(exec, "data - shared space"));
 
-    ::exec::static_thread_pool pool{4};
+    ::exec::static_thread_pool pool{4}; // NOLINT(misc-const-correctness)
     const context_t esc{exec};
 
     ::stdexec::sender auto fork = ::stdexec::schedule(pool.get_scheduler()) | ::exec::split();

tests/kokkos_ext/graph/test_inter_op.cpp

@@ -153,7 +153,7 @@ TEST_F(InterOpTest, transition_to_static_thread_pool) {
     const context_t esc{exec};
     SHOW_EXEC_SPACE_ID(exec)
 
-    ::exec::static_thread_pool pool{1};
+    ::exec::static_thread_pool pool{1}; // NOLINT(misc-const-correctness)
 
     auto chain = ::stdexec::schedule(esc.get_scheduler())
                | ::stdexec::then(
@@ -187,7 +187,7 @@ TEST_F(InterOpTest, transition_from_static_thread_pool) {
     const context_t esc{exec};
     SHOW_EXEC_SPACE_ID(exec)
 
-    ::exec::static_thread_pool pool{1};
+    ::exec::static_thread_pool pool{1}; // NOLINT(misc-const-correctness)
 
     auto chain = ::stdexec::schedule(pool.get_scheduler())
                | ::stdexec::then(
@@ -221,7 +221,7 @@ TEST_F(InterOpTest, transition_from_static_thread_pool_and_back) {
     const context_t esc{exec};
     SHOW_EXEC_SPACE_ID(exec)
 
-    ::exec::static_thread_pool pool{1};
+    ::exec::static_thread_pool pool{1}; // NOLINT(misc-const-correctness)
 
     auto chain = ::stdexec::schedule(pool.get_scheduler())
                | ::stdexec::then(

tests/native/test_conditional_if.cpp

@@ -77,6 +77,7 @@ TEST_P(cuda, conditional_if)
     graph_exec.submit(stream);
 
     const auto mirror = data.get_host_copy(stream);
+    stream.fence();
 
     ASSERT_EQ(mirror.at(0), this->GetParam() ? 1 : 0);
 

tests/native/test_conditional_while.cpp

@@ -85,6 +85,7 @@ TEST(cuda, conditional_while)
     graph_exec.submit(stream);
 
     const auto mirror = data.get_host_copy(stream);
+    stream.fence();
 
     EXPECT_EQ(mirror.at(0), 42);
 

tests/native/test_diamond.cpp

@@ -55,6 +55,7 @@ TEST(cuda, diamond_graph)
     graph_exec.submit(stream);
 
     const auto mirror = data.get_host_copy(stream);
+    stream.fence();
 
     ASSERT_THAT(mirror, ::testing::ElementsAre(4));
 

tests/native/test_memcpy_node.cpp

@@ -74,6 +74,7 @@ TEST(cuda, memcpy_node)
     //! Check the values in @c dst and @c src are identical (and as expected).
     const auto src_h = src.get_host_copy(stream);
     const auto dst_h = dst.get_host_copy(stream);
+    stream.fence();
 
     std::vector<scalar_t> expected(size);
     std::iota(expected.begin(), expected.end(), 0);

tests/native/test_subgraph.cpp

@@ -74,7 +74,10 @@ TEST(cuda, subgraph)
     graph_exec.submit(stream);
 
     //! Check results.
-    ASSERT_EQ(data.get_host_copy(stream), (std::vector<value_t>{5, 5, 4}));
+    const auto mirror = data.get_host_copy(stream);
+    stream.fence();
+
+    ASSERT_EQ(mirror, (std::vector<value_t>{5, 5, 4}));
 }
 
 } // namespace tests::cuda

tests/stdexec/test_check_scheduler.cpp

@@ -36,7 +36,7 @@ TEST(check_scheduler, default) {
 
 //! @test @c exec::static_thread_pool scheduler.
 TEST(check_scheduler, static_thread_pool) {
-    ::exec::static_thread_pool pool{1};
+    ::exec::static_thread_pool pool{1}; // NOLINT(misc-const-correctness)
 
     auto chain = ::stdexec::schedule(pool.get_scheduler()) | check_scheduler<static_thread_pool_scheduler_t>()
                | THEN_SHOW_ID;
@@ -50,7 +50,7 @@ TEST(check_scheduler, static_thread_pool) {
  * See also https://github.com/NVIDIA/stdexec/issues/1736#issuecomment-3720622409.
  */
 TEST(check_scheduler, split_when_all_no_forward) {
-    ::exec::static_thread_pool pool{1};
+    ::exec::static_thread_pool pool{1}; // NOLINT(misc-const-correctness)
 
     auto fork = ::stdexec::schedule(pool.get_scheduler()) | check_scheduler<static_thread_pool_scheduler_t>()
               | THEN_SHOW_ID | ::exec::split();
@@ -68,7 +68,7 @@ TEST(check_scheduler, split_when_all_no_forward) {
  * See also https://github.com/NVIDIA/stdexec/issues/1736#issuecomment-3720622409.
  */
 TEST(check_scheduler, split_transfer_when_all_no_forward) {
-    ::exec::static_thread_pool pool{1};
+    ::exec::static_thread_pool pool{1}; // NOLINT(misc-const-correctness)
 
     auto fork = ::stdexec::schedule(pool.get_scheduler()) | check_scheduler<static_thread_pool_scheduler_t>()
               | THEN_SHOW_ID | ::exec::split();
@@ -83,7 +83,7 @@ TEST(check_scheduler, split_transfer_when_all_no_forward) {
 
 //! @test Multiple splits.
 TEST(check_scheduler, multiple_splits) {
-    ::exec::static_thread_pool pool{1};
+    ::exec::static_thread_pool pool{1}; // NOLINT(misc-const-correctness)
 
     auto fork_A = ::stdexec::schedule(pool.get_scheduler()) | check_scheduler<static_thread_pool_scheduler_t>()
                 | THEN_SHOW_ID | ::exec::split();

tests/stdexec/test_value_receiver.cpp

@@ -27,7 +27,7 @@ namespace tests::stdexec {
 TEST(ValueReceiver, int) {
     ::stdexec::sender auto sndr = ::stdexec::just(int{42}) | ::stdexec::continues_on(::stdexec::inline_scheduler{});
 
-    int placeholder = 0;
+    int placeholder = 0; // NOLINT(misc-const-correctness)
 
     auto opstate = ::stdexec::connect(sndr, ValueReceiver<int>{.value = std::addressof(placeholder)});