[PATCHES] messaging iov / recvfrom

[Simo]

On Fri, 2014-05-30 at 10:28 +0200, Michael Adam wrote:
> On 2014-05-30 at 09:35 +0200, Andreas Schneider wrote:
> > On Thursday 29 May 2014 14:31:55 Jeremy Allison wrote:
> > > On Mon, May 26, 2014 at 05:07:44PM +0200, Michael Adam wrote:
> > > > Oops, there was a bug in the first patch.
> > > > Sorry for posting prematurely...
> > > > Attaching a fixed version.
> > > > 
> > > > Thanks - Michael
> > > > 
> > > > On 2014-05-26 at 15:52 +0200, Michael Adam wrote:
> > > > > Hi,
> > > > > 
> > > > > as a result of my work towards adding support for fd-passing
> > > > > to our messaging, find attached two first preparatory patches
> > > > > that might already be useful.
> > > > > 
> > > > > The first changes the send_fn to use struct  iovec
> > > > > instead of data blob. (Volker has already looked
> > > > > over this one.)
> > > > > 
> > > > > The second one lets unix_dgram_recv_handler()
> > > > > use recvmsg() instead of recv().
> > > > > 
> > > > > Review/push/comments appreciated.
> > > 
> > > LGTM ! Pushed to autobuild.
> > 
> > 
> > 
> > 
> > I gave a NAK! Why do you push it now?
> 
> I guess there was a misunderstanding.
> I will post a follow-up patch to add the
> missing checks for msg_control.

FWIW I was confused as well about the issue, what Andreas wanted to
point out is that the member 'msg_control' is not always there on all
platforms, so initializing it unconditionally will simply cause build
issues.

However given we are initializing it to NULL, the solution is simply to
omit the initialization, a memset is not necessary.

Simo.