Re: PTY dies when master in parent process is closed.

[Corinna Vinschen <corinna-cygwin at cygwin dot com>]

On Mar 11 17:48, Corinna Vinschen wrote:
> Hi Takashi,
> 
> On Mar  8 16:37, Takashi Yano wrote:
> > On Fri, 6 Mar 2015 20:07:10 +0900
> > Takashi Yano <takashi.yano@nifty.ne.jp> wrote:
> > 
> > > On Thu, 5 Mar 2015 14:58:39 +0100
> > > Corinna Vinschen <corinna-cygwin@cygwin.com> wrote:
> > > 
> > > > I applied a patch.  Please have a look.
> > > 
> > > I have tested the latest CVS version, and found
> > > a new problem.
> > > 
> > > With new CVS version, slave side can not detect
> > > closure of master.
> > > 
> > > Please use following Test Case 3. Test Case 3 is
> > > not terminated by itself with latest CVS.
> > > 
> > > It seems that the program is stopping at
> > > cygwait(input_available_event, time_to_wait)
> > > in fhandler_pty_slave::read().
> > > 
> > > I guess input_available_event should be set when
> > > the last valid master fd is closed.
> > 
> > For this problem, I have made a patch.
> > 
> > With this patch, it has been confirmed that the problems
> > in Test Case 1, 2 and 3 are fixed.
> > 
> > I am glad if this would be a help.
> > 
> > However, one matter to be concerned is irregular use of
> > PeekNamedPipe(). Maybe alternative means could be better
> > for detecting closure of all master fds.
> 
> I'm inclined to apply this patch.  I think using PeekNamedPipe this way
> is pretty nice, albeit it's not atomic.  It might be a good idea to add
> an acquire/release_output_mutex bracket, even if that doesn't help for
> cases in which a process holding a master fd crashes hard.

Scratch that.  Atomicity doesn't really matter in this case.


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Maintainer                 cygwin AT cygwin DOT com
Red Hat

Attachment: pgpWaaXU77jdu.pgp
Description: PGP signature