libstdc++: Fix parallel std::exclusive_scan [PR108236]

The standard says that std::exclusive_scan can be used to work in
place, i.e. where the output range is the same as the input range. This
means that the first sum cannot be written to the output until after
reading the first input value, otherwise we'll already have overwritten
the first input value.

While writing a new testcase I also realised that the serial version of
std::exclusive_scan uses copy construction for the accumulator variable,
but the standard only requires Cpp17MoveConstructible. We also require
move assignable, which is missing from the standard's requirements, but
we should at least use move construction not copy construction.

A similar problem exists for some other new C++17 numeric algos, but
I'll fix the others in a subsequent commit.

libstdc++-v3/ChangeLog:

	PR libstdc++/108236
	* include/pstl/glue_numeric_impl.h (exclusive_scan): Pass __init
	as rvalue.
	* include/pstl/numeric_impl.h (__brick_transform_scan): Do not
	write through __result until after reading through __first. Move
	__init into return value.
	(__pattern_transform_scan): Pass __init as rvalue.
	* include/std/numeric (exclusive_scan): Move construct instead
	of copy constructing.
	* testsuite/26_numerics/exclusive_scan/2.cc: New test.
	* testsuite/26_numerics/pstl/numeric_ops/108236.cc: New test.

3 files changed, 8 insertions, 7 deletions

diff --git a/libstdc++-v3/include/pstl/glue_numeric_impl.h b/libstdc++-v3/include/pstl/glue_numeric_impl.h
index 490175c..10d4912 100644
--- a/libstdc++-v3/include/pstl/glue_numeric_impl.h
+++ b/libstdc++-v3/include/pstl/glue_numeric_impl.h
@@ -108,7 +108,7 @@ exclusive_scan(_ExecutionPolicy&& __exec, _ForwardIterator1 __first, _ForwardIte
 
     using namespace __pstl;
     return __internal::__pattern_transform_scan(__dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last,
-                                                __result, __pstl::__internal::__no_op(), __init, __binary_op,
+                                                __result, __pstl::__internal::__no_op(), std::move(__init), __binary_op,
                                                 /*inclusive=*/std::false_type());
 }
 
diff --git a/libstdc++-v3/include/pstl/numeric_impl.h b/libstdc++-v3/include/pstl/numeric_impl.h
index 7ba83ee..e1ebec1 100644
--- a/libstdc++-v3/include/pstl/numeric_impl.h
+++ b/libstdc++-v3/include/pstl/numeric_impl.h
@@ -160,11 +160,12 @@ __brick_transform_scan(_ForwardIterator __first, _ForwardIterator __last, _Outpu
 {
     for (; __first != __last; ++__first, ++__result)
     {
-        *__result = __init;
+	_Tp __v = std::move(__init);
         _PSTL_PRAGMA_FORCEINLINE
-        __init = __binary_op(__init, __unary_op(*__first));
+        __init = __binary_op(__v, __unary_op(*__first));
+        *__result = std::move(__v);
     }
-    return std::make_pair(__result, __init);
+    return std::make_pair(__result, std::move(__init));
 }
 
 // Inclusive form
@@ -225,7 +226,7 @@ __pattern_transform_scan(_Tag, _ExecutionPolicy&&, _ForwardIterator __first, _Fo
                          _OutputIterator __result, _UnaryOperation __unary_op, _Tp __init, _BinaryOperation __binary_op,
                          _Inclusive) noexcept
 {
-    return __internal::__brick_transform_scan(__first, __last, __result, __unary_op, __init, __binary_op, _Inclusive(),
+    return __internal::__brick_transform_scan(__first, __last, __result, __unary_op, std::move(__init), __binary_op, _Inclusive(),
                                               typename _Tag::__is_vector{})
         .first;
 }
diff --git a/libstdc++-v3/include/std/numeric b/libstdc++-v3/include/std/numeric
index dd98f40..37579c5 100644
--- a/libstdc++-v3/include/std/numeric
+++ b/libstdc++-v3/include/std/numeric
@@ -491,8 +491,8 @@ namespace __detail
     {
       while (__first != __last)
 	{
-	  auto __v = __init;
-	  __init = __binary_op(__init, *__first);
+	  _Tp __v = std::move(__init);
+	  __init = __binary_op(__v, *__first);
 	  ++__first;
 	  *__result++ = std::move(__v);
 	}