io_context replace io_service

[snahmad]

Should we been using io_context instead of io_service in latest boost.

[djarek]

io_service is deprecated. Yes, you should use io_context. Beware that the "old" API is deprecated as well (e.g. io_service.post(), you should use post(io_context, handler)).

[superlambchops]

Is there a guide to converting a program from using io_context to use io_service?

[djarek]

io_context -> io_service
io_context.post() -> io_context.get_executor().post()
io_context.dispatch() -> io_context.get_executor().dispatch()

io_context::strand -> strand<io_context::executor_type>

there were also changes to composed operation customization hooks - there are only 2 now - boost::asio::associated_allocator and boost::asio::associated_executor, which default to looking for get_allocator(), get_executor(), T::allocator_type, T::executor_type members of the composed operation function object.

This is not a complete list.

[GrumpyOldTroll]

Is there a doc outlining the reasons a non-backward-compatible change was required? This is an unpleasant surprise, and pretty disruptive to existing projects using boost...

[DinoStray]

I feel terrible for the changes

[TorstenRobitzki]

@djarek In your list above, there is one thing missing. In earlier versions of asio, we could obtain the io_context from get_io_service() in serial_port or other IO objects. Now, .get_executor().context() returns the base class to io_context and a downcast is required. Is there an other solution to get the io_context from an existing IO object?

[thejohnfreeman]

io_context -> io_service
io_context.post() -> io_context.get_executor().post()
io_context.dispatch() -> io_context.get_executor().dispatch()

io_context::strand -> strand<io_context::executor_type>

New user here. Is this supposed to mean "the left expression supersedes the right expression" or "change occurrences of the left expression to the one on the right"?

[ivan-v-kush]

Is there a doc outlining the reasons a non-backward-compatible change was required? This is an unpleasant surprise, and pretty disruptive to existing projects using boost...

I think due to Network TS. But now appears the question, why the WG have
choosen io_context

http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/p0112r0.html

[ghost]

This issue was moved by chriskohlhoff to chriskohlhoff/asio#615.