Compile error when any_io_executor based websocket stream with bind_executor strand about beast HOT 6 OPEN

Yeah, Chris usually doesn't respond to the issues, but he reads them. We should track the changes in the repository to see if it gets fixed or not.

from beast.

The issue is associated with creating a work_guard from a strand<any_io_executor>. It attempts to utilize on_work_started() and on_work_finished() functions, which are not defined for asio::any_io_executor. This results in a compile error in websocket::stream<tcp>::async_write because it attempts to create a work_guard from the associated executor.
It can be reproduced with the following code as well: https://godbolt.org/z/nvr9jjnrY

#include <boost/asio.hpp>

namespace asio = boost::asio;

int main()
{
    asio::io_context ioc;
    asio::any_io_executor exec = ioc.get_executor();
    auto strand = asio::make_strand(exec);
    auto wg = asio::make_work_guard(strand);
}

I will further investigate to determine whether we can address this in Beast or if it is a bug in Asio.

from beast.

I'm not 100% sure but passing strand.get_inner_executor() to make_work_guard could solve the issue ?

#include <boost/asio.hpp>

namespace asio = boost::asio;

int main()
{
    asio::io_context ioc;
    asio::any_io_executor exec = ioc.get_executor();
    auto strand = asio::make_strand(exec);
    auto wg = asio::make_work_guard(strand.get_inner_executor());
}

https://godbolt.org/z/jv1zEW7ET

from beast.

I'm not 100% sure but passing strand.get_inner_executor() to make_work_guard could solve the issue ?

This will generate a work guard for any_io_executor instead of strand<any_io_executor>. Calling get_executor on this work guard will return an instance of any_io_executor.
Although I believe it will function correctly in this case, as we don't retrieve the executor from the work guard, I am seeking a cleaner solution or addressing this in Asio.

from beast.

I've raised an issue in Asio for this; let's see what will be the response. chriskohlhoff/asio#1394

from beast.

chriskohlhoff/asio#1394 still no response. It seems that there are a lot of uncommented issues in the project.
I think that this issue is often appered using C++20 coroutine with strand in the real world.
See https://godbolt.org/z/7EnWa3Es1
If there were no update until the next release, I would appliciate if you implemented some workaround on the beast side .

from beast.