Re: popen2 vs pclose

[Rob Mahurin]

On Aug 25, 2009, at 3:22 PM, John W. Eaton wrote:
> Maybe it is just best to expect users of the popen2 function to do
> something like
>
>   [in, out, pid] = popen2 (...);
>   ...
>   fclose (in);
>   fclose (out);
>   waitpid (pid);
>
> and not try to do this automatically for them.  If so, then I think we
> just need to document this behavior and make sure that this is done
> properly for any code in Octave that uses popen2.

This seems to be the approach taken by Perl's IPC::Open2:

> > open2() does not wait for and reap the child process after it  
> > exits.  Except for short programs where it's acceptable to let the  
> > operating system take care of this, you need to do this yourself.  
> > This is normally as simple as calling "waitpid $pid, 0" when  
> > you're done with the process. Failing to do this can result in an  
> > accumulation of defunct or "zombie" processes. See "waitpid" in  
> > perlfunc for more information.

Python (http://pydoc.org/2.4.1/popen2.html) seems to have a class  
with poll() and wait() methods, closer perhaps to your proposed patch.

The reason to call wait() is to collect the status of the child, and  
the reason that zombies stay in the process table is to allow this  
status to be collected.  If you write auto-reaping code where waitpid 
() is called at least twice for each pipe, then the whichever call  
happens second is guaranteed to fail and there's no reliable access  
to the exit status unless you invent a user-space-accessible object.

I think the documentation solution is the simplest and the least  
likely to cause mysteries later on.

Rob

--
Rob Mahurin
University of Manitoba, Dept. of Physics & Astronomy
at:     Oak Ridge National Laboratory   865 207 2594
        Oak Ridge, Tennessee            address@hidden