[PATCH] Fix / slightly simplify smbd/scavenger

[Jeremy Allison]

On Fri, Oct 30, 2015 at 04:57:39PM +0100, Volker Lendecke wrote:
> On Fri, Oct 30, 2015 at 08:34:45AM -0700, Jeremy Allison wrote:
> > On Fri, Oct 30, 2015 at 07:46:39AM +0100, Ralph Boehme wrote:
> > > On Thu, Oct 29, 2015 at 03:26:53PM -0700, Jeremy Allison wrote:
> > > > On Thu, Oct 29, 2015 at 08:23:26PM +0200, Uri Simchoni wrote:
> > > > > 
> > > > > 
> > > > > On 10/29/2015 06:02 PM, Volker Lendecke wrote:
> > > > > >+	ret = read_data(fd, child, sizeof(child));
> > > > > >+	if (ret == -1) {
> > > > > Doesn't seem to take EOF into account, but neither does the original
> > > > > code. How come?
> > > > 
> > > > fd's are created by socketpair() and not set non-blocking,
> > > > so shouldn't get a zero return.
> > > 
> > > But wouldn't one side get 0 when the other does a close?
> > 
> > Yeah, but I thought these were long-lived sockets (only
> > closed on shutdown). I'll look again.
> 
> Ok, this goes too far. Attached find a minimal patch that
> just fixes the very obvious bug.
> 
> Maybe that's a bit less controversial.

It's not controversy, it's people trying to
remember how code works :-).

> Review&push appreciated!

Nah, first patch is better. Pushed that one :-).