how to reuse TcpStream code from src/sys/windows/tcp.rs ?

[Dushistov]

cc @alexcrichton @carllerche

I want to work with AF_BTH socket on windows OS,
to demonstrate what I am talking about here some code (blocking variant):

//create
let bt_socket: SOCKET = unsafe { socket(AF_BTH, SOCK_STREAM, BTHPROTO_RFCOMM) };
//skip error handling
let mut sock_addr = SOCKADDR_BTH {
      addressFamily: AF_BTH as USHORT,
      btAddr: bt_addr.into(),
      serviceClassId: RFCOMM_PROTOCOL_UUID,
      port: BT_PORT_ANY,
};
let sock_addr_ptr: *mut SOCKADDR_BTH = &mut sock_addr;
let err = unsafe { connect(socket, sock_addr_ptr as *mut SOCKADDR, mem::size_of_val(sock_addr) };

As I understand comments from lib.rs and windows mod.rs I basically need not only use windows::Binding to imlement mio::Eventd, but also starts I/O with overlapped structures on readiness request, so basicaly copy/paste code from windows/tcp.rs and modify for my need, am I right?

If so, do you accept PR that generalize windows/tcp.rs to any SOCK_STREAM and allow some kind of user defined connect and socket creation functions?

[alexcrichton]

I'd recommend taking a look at mio-named-pipes which are named pipes for mio, which you'll also likely want to mirror. In general though yes, you'll want to copy code from tcp.rs and adapt it as appropriate, it won't ever be publicly exposed from mio itself.

[Dushistov]

@alexcrichton I looked at mio-named-pipes, looks a lot of code duplication
with mio/sys/windows, do you have any plans for refactoring?

For example, why not just call in Binding::register_socket/register_handle
approapriate schedule functions?

[carllerche]

I don't think we are going to extract that as an official mio API. The main reason is that it creates additional API surface to support and resources are thin. However. Feel free to extract those bits to another crate.

…

On Mon, Mar 27, 2017 at 8:40 AM Evgeniy Dushistov ***@***.***> wrote: @alexcrichton <https://github.com/alexcrichton> I looked at mio-named-pipes, looks a lot of code duplication with mio/sys/windows, do you have plans refactoring? For example, why not just call in Binding::register_socket/register_handle approapriate schedule functions? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#578 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAAYJCU3OFF5bzmAlkc-kOHBaVohMYWuks5rp9hbgaJpZM4MpV-4> .

[alexcrichton]

Currently no I don't plan on deduplicating, I'd personally prefer the freedom to change TCP bindings in mio itself and then at my leisure update other bindings. Exposing an API as a public interface would require everything to change all at once (and would make it very difficult to change)