4e4o

i have similar issue, maybe it is related https://github.com/chriskohlhoff/asio/issues/1027

I think this problem is related not only to parallel_group, but to cancellation mechanism at whole. More details: when some_sync_operation completes it will call awaitable_handler and will clear cancellation_slot in...

Well, if no one knows solution it seems i'v found one. I made wrapper for co_spawn that redirects cancellation signal to correct executor. https://github.com/4e4o/boost_base/blob/master/src/Coroutine/CancellableSpawn.hpp. Any comments will be appreciated.

bug disappears if i change args capture to by reference: ``` co_return co_await spawn([&io, args...]() -> awaitable { ``` to this: ``` co_return co_await spawn([&io, &args...]() -> awaitable { ```

gcc 11.3.1 still not fixed. I will close this issue.

> Furthermore, I do not know if, or how, to make it so that everything that happens within a co_spawn is bound to a specific strand. You can create executor...

I v'got the same issue. Here is extra logs from tensorrt [11.txt](https://github.com/cyrusbehr/YOLOv8-TensorRT-CPP/files/13471658/11.txt)