This is the mail archive of the cygwin mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: close on exec atomics


On Jul 21 12:05, Christopher Faylor wrote:
> On Tue, Jul 21, 2009 at 04:40:03PM +0200, Corinna Vinschen wrote:
> >On Jul 21 10:17, Christopher Faylor wrote:
> >>I agree about waiting, but I think this is one for my TODO list.  The
> >>close_on_exec code is all mine, AFAIK.  I reworked it all back in 1998
> >>or so, so it is pretty fresh in my mind.
> >
> >I don't think there's a lot to do in close_on_exec code itself related
> >to this.  There is no need to call set_close_on_exec at all, rather the
> >fhandler should create the handle not inheritable in the first place.
> >That goes for open as well as fcntl.  Time to add Linux' dup3 call, I
> >guess :)
> 
> There is a lot to deal with here because:
> 
>   % grep ::set_close_on_exec *.cc
>   fhandler.cc:fhandler_base::set_close_on_exec (bool val)
>   fhandler_console.cc:fhandler_console::set_close_on_exec (bool val)
>   fhandler_console.cc:  fhandler_base::set_close_on_exec (val);
>   fhandler_socket.cc:fhandler_socket::set_close_on_exec (bool val)
>   fhandler_socket.cc:  fhandler_base::set_close_on_exec (val);
>   fhandler_tape.cc:fhandler_dev_tape::set_close_on_exec (bool val)
>   fhandler_tape.cc:  fhandler_dev_raw::set_close_on_exec (val);
>   fhandler_tty.cc:fhandler_tty_common::set_close_on_exec (bool val)
>   fhandler_windows.cc:fhandler_windows::set_close_on_exec (bool val)
>   fhandler_windows.cc:    fhandler_base::set_close_on_exec (val);
> 
> and even more importantly:
> 
>   % grep '^fhandler[a-z_0-9]*::dup' *.cc
>   fhandler.cc:fhandler_base::dup (fhandler_base *child)
>   fhandler_clipboard.cc:fhandler_dev_clipboard::dup (fhandler_base * child)
>   fhandler_console.cc:fhandler_console::dup (fhandler_base *child)
>   fhandler_dsp.cc:fhandler_dev_dsp::dup (fhandler_base * child)
>   fhandler_fifo.cc:fhandler_fifo::dup (fhandler_base *child)
>   fhandler_floppy.cc:fhandler_dev_floppy::dup (fhandler_base *child)
>   fhandler_mem.cc:fhandler_dev_mem::dup (fhandler_base *child)
>   fhandler_random.cc:fhandler_dev_random::dup (fhandler_base *child)
>   fhandler_raw.cc:fhandler_dev_raw::dup (fhandler_base *child)
>   fhandler_registry.cc:fhandler_registry::dup (fhandler_base *child)
>   fhandler_serial.cc:fhandler_serial::dup (fhandler_base *child)
>   fhandler_socket.cc:fhandler_socket::dup (fhandler_base *child)
>   fhandler_tape.cc:fhandler_dev_tape::dup (fhandler_base *child)
>   fhandler_tty.cc:fhandler_tty_slave::dup (fhandler_base *child)
>   fhandler_tty.cc:fhandler_pty_master::dup (fhandler_base *child)
>   fhandler_virtual.cc:fhandler_virtual::dup (fhandler_base * child)
>   pipe.cc:fhandler_pipe::dup (fhandler_base *child)
> 
> I was thinking about a way of collapsing most of those into one generic
> call.  But not for 1.7.1, obviously.

Yes, but that's a different problem.  Your work on set_close_on_exec
does not touch open(O_CLOEXEC).  The O_CLOEXEC stuff doesn't need to
call set_close_on_exec at all.  Granted, using it would be a simple way
to implement O_CLOEXEC, but creating the handles inheritable vs. not
inheritable in the first place is the more correct solution.

Talking about dup3/fcntl(F_DUPFD_CLOEXEC), that's of course another
story.  I guess we should just split it between these two items.


Corinna

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

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]