A strange thing in parallel::util::detail::handle_local_exceptions.

[taeguk]

hpx/hpx/parallel/util/detail/handle_local_exceptions.hpp

Lines 29 to 40 in 76f7c3b

	HPX_ATTRIBUTE_NORETURN static void call(std::exception_ptr const& e)
	{
	try {
	std::rethrow_exception(e);
	}
	catch (std::bad_alloc const& ba) {
	throw ba;
	}
	catch (...) {
	throw;
	}
	}

Does 'Line 38' make sense?
That function seems to just rethrow exception_ptr.
As I think, 'Line 38' should be changed to

throw exception_list(std::current_exception());

[hkaiser]

@taeguk please construct a test which exposes the problem you believe we have.

[taeguk]

@hkaiser This is not bug which can be tested. This is the problem about intention of that function. And I want to just question about it.
I want to know what is the meaning of that function.
Is it intended that function just rethrows std::exception_ptr?

hpx/hpx/parallel/util/detail/handle_local_exceptions.hpp

Lines 56 to 68 in 76f7c3b

	template <typename T>
	static void call(std::vector<hpx::future<T> > const& workitems,
	std::list<std::exception_ptr>& errors)
	{
	for (hpx::future<T> const& f: workitems)
	{
	if (f.has_exception())
	call(f.get_exception_ptr(), errors);
	}
	if (!errors.empty())
	throw exception_list(std::move(errors));
	}

Like above, other overloaded functions of that function are related to exception_list.
So, I guess that 'Line 38' is an mistake and changing to throw exception_list(std::current_exception()); may be reasonable.
I can't determine whether my guessing is correct because that code is not written by me.

[hkaiser]

@taeguk you're right, this is a nice catch. I agree this has to be changed as you suggest. But a test would be nice anyways, mostly to ensure that those functions throw an exception_list under any circumstances.

[taeguk]

@hkaiser I want to change 'Line 38' to what I said.

[taeguk]

@hkaiser Do you agree that I change that?
With changing, handling exception for parallel algorithms can be abstracted as a function than explicitly coded. (https://github.com/taeguk/hpx/blob/f2d9cf60f692b50e81a69e78a93c54d0596a3294/hpx/parallel/algorithms/merge.hpp#L296-L304)

[hkaiser]

@taeguk, yes, please go ahead.