> On June 8, 2018, 3:46 p.m., Andrew Schwartzmeyer wrote:
> > 3rdparty/libprocess/src/libwinio_io.cpp
> > Lines 78-101 (patched)
> > <https://reviews.apache.org/r/67390/diff/1/?file=2032286#file2032286line78>
> >
> > Wait... doesn't this belong in the `prepare_async` from earlier? Did
> > something get moved?
>
> Akash Gupta wrote:
> Hm what do you mean? There's two `prepare_asyncs`. One for POSIX
> (polling), one for Windows (IOCP). This is the first time for the Windows one.
Now I know that `libprocess/src/io.cpp` isn't POSIX only... and it has:
```
Try<Nothing> prepare_async(int_fd fd)
{
// TODO(akagup): Add windows iocp.
return os::nonblock(fd);
}
Try<bool> is_async(int_fd fd)
{
// TODO(akagup): Add windows iocp.
return os::isNonblock(fd);
}
```
added in https://reviews.apache.org/r/67386/
> On June 8, 2018, 3:46 p.m., Andrew Schwartzmeyer wrote:
> > 3rdparty/libprocess/src/libwinio_socket.cpp
> > Lines 105 (patched)
> > <https://reviews.apache.org/r/67390/diff/1/?file=2032287#file2032287line105>
> >
> > What is it, and why?
>
> Akash Gupta wrote:
> It bascially combines a lot of small packets into a single big packet, so
> it can improve throughout at the expense of latency. I only disable it
> because the POSIX `PollSocketImpl` disables it. I think the reason why it is
> disabled is because serving libprocess HTTP requets prefer latency over
> throughout.
Gotcha. Do they have anything more than this comment in their implementation?
Maybe we could check the blame and find an issue...
- Andrew
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/67390/#review204503
-----------------------------------------------------------
On May 30, 2018, 11:46 a.m., Akash Gupta wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/67390/
> -----------------------------------------------------------
>
> (Updated May 30, 2018, 11:46 a.m.)
>
>
> Review request for mesos, Andrew Schwartzmeyer, Benjamin Mahler, Eric Mumau,
> John Kordich, Joseph Wu, and Radhika Jandhyala.
>
>
> Bugs: MESOS-8668, MESOS-8671 and MESOS-8672
> https://issues.apache.org/jira/browse/MESOS-8668
> https://issues.apache.org/jira/browse/MESOS-8671
> https://issues.apache.org/jira/browse/MESOS-8672
>
>
> Repository: mesos
>
>
> Description
> -------
>
> The libprocess eventloops are used to implement the `EventLoop`,
> SocketImpl` and `io::read/write` interfaces. Thus, by providing those
> implementations using libwinio, libprocess now supports the libwinio
> backend.
>
>
> Diffs
> -----
>
> 3rdparty/libprocess/src/libwinio.hpp PRE-CREATION
> 3rdparty/libprocess/src/libwinio_eventloop.cpp PRE-CREATION
> 3rdparty/libprocess/src/libwinio_io.cpp PRE-CREATION
> 3rdparty/libprocess/src/libwinio_socket.cpp PRE-CREATION
>
>
> Diff: https://reviews.apache.org/r/67390/diff/1/
>
>
> Testing
> -------
>
>
> Thanks,
>
> Akash Gupta
>
>
