[MAYBE SOLVED] was Can't resolv "sh: can't access tty; job control turned of

[Denis Vlasenko]

On Thursday 18 January 2007 13:22, Luciano Miguel Ferreira Rocha wrote:
> On Thu, Jan 18, 2007 at 02:28:24AM +0100, Denis Vlasenko wrote:
> > On Wednesday 17 January 2007 23:27, Laurent B wrote:
> > > >
> > > >On Wed, Jan 17, 2007 at 09:26:10PM +0100, Laurent B wrote:
> > > > > rm /dev/tty
> > > > > ln -s /dev/pts/0 /dev/tty
> > > >
> > > >That's wrong. I hope you don't use more than one login at the same time,
> > > >as those will get pts/1, pts/2, etc., but /dev/tty will point to the
> > > >wrong one.
> > 
> > Here I have current svn's telnetd running in standalone mode
> > _without_ controlling tty (i.e. it is not run from command line,
> > it runs, like in your case, detached from any terminal):
> 
> telnetd doesn't need a controlling tty, but it needs to create a
> controlling pseudo-tty for the children it spawns.

Of course it doesn't need a ctty. I didn't say it needs one.
I just demonstrated that on my machine backgrounded telnetd works.
Laurent B may want to compare my strace output to his one.

> <strace deleted>
> I didn't find any call to setsid and ioctl(TIOCSCTTY), but there was
> some strace output missing from the start of the child.

In fact, I didn't use strace -f, so ALL child output is missing.

> The code for 
> 1.2.2.1 seems to be correct.
> Anyway, I'd be more interested in the strace output of the cases where
> it is known to fail.

Yes.
--
vda