----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/47468/#review134655 -----------------------------------------------------------
3rdparty/libprocess/include/process/posix/pipe.hpp (line 25) <https://reviews.apache.org/r/47468/#comment199501> Rather than introducing a class to represent piped fds right now, can we follow the `os::pipe` function model we used for other system calls? This way we can think separately with the community about the abstractions we want to introduce around `FD` and `Pipe`, and consider things like life-time, RAII, etc. It seems like we can get away with something along these lines for now: ``` namespace process { namespace os { Try<Nothing> pipe(int fd[2]); } // namespace os { } // namespace process { ``` The support for `Handle`s on Windows is fine. - Joris Van Remoortere On May 17, 2016, 4:01 p.m., Alex Clemmer wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/47468/ > ----------------------------------------------------------- > > (Updated May 17, 2016, 4:01 p.m.) > > > Review request for mesos, Daniel Pravat, Artem Harutyunyan, Joris Van > Remoortere, and Michael Park. > > > Bugs: MESOS-4475 > https://issues.apache.org/jira/browse/MESOS-4475 > > > Repository: mesos > > > Description > ------- > > Windows: Implemented IPC pipe. > > > Diffs > ----- > > 3rdparty/libprocess/include/Makefile.am > a27e9f7aaa2b45139d4eb95f365d8b2041d7aad2 > 3rdparty/libprocess/include/process/pipe.hpp PRE-CREATION > 3rdparty/libprocess/include/process/posix/pipe.hpp PRE-CREATION > 3rdparty/libprocess/include/process/windows/pipe.hpp PRE-CREATION > > Diff: https://reviews.apache.org/r/47468/diff/ > > > Testing > ------- > > > Thanks, > > Alex Clemmer > >
