Hi Martin, you're right that RCVBUF is in control in this context. The problem with that, though, is that it prevents setting the buffer size if you try to set it on socket that doesn't support RCVBUF, such as a push socket for example. Maybe for this case the code should check for both SNDBUF and RCVBUF and if either or both are set, choose the smaller value of the two? --steve On Wed, Oct 8, 2014 at 9:57 AM, Martin Sustrik <sustrik@xxxxxxxxxx> wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Hi Steve, > > IIRC POSIX specifies that one of the options (SNDBUF or RCVBUF) is > ignored when applied to IPC socket. There's only one buffer under the > covers anyway. > > Anyway, feel free to have a look at it. I may be wrong. > > Martin > > On 08/10/14 15:42, Steve Vinoski wrote: > > I'm wondering why the inproc transport doesn't support NN_SNDBUF? > > Seems like all the machinery to check the queue size and return > > EAGAIN if there's no room is already in place, and so just wiring > > the setting of the NN_SNDBUF option value into the already-existing > > variable, overwriting the default 128k, would make it work. > > > > I'm happy to go create a pull request for this, but wanted to > > check first to make sure there's not some design decision I'm > > overlooking. > > > > thanks, --steve > > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1.4.11 (GNU/Linux) > > iQEcBAEBAgAGBQJUNULVAAoJENTpVjxCNN9Y1+EH+wUetjx1FMF3dIxfJB17yMZ9 > ssWbOQ+bbrbTzSfCrQSBa6yCd82nq+gY6Btq6UCqNvTl99b8VcG0eFoYsgrS4dcL > HwgJlXaXLBdlcFRjVt2vXCDH9UQavCHfyTO3FnC/Fh97RlBfld90dlstspoXb2TW > cHIfr7P3nQ9Wc1tg5B7tjaOj816bjcq5XUAfA+ne6YmsjN5WHn8R4ci/y5sYPU8p > +Sblh3cxYo/ioeZhcQFIMYyMhAPRNEluUg0du09trr6E46qplBah5q+YTAtSJq7I > CPqH67NwCNd7bC5nvp4tPYdztyAJ45vIv6HT9FjtxcLMc32jK6Tr6XSjqioS55M= > =Bqd9 > -----END PGP SIGNATURE----- > >