GNU bug report logs - #9488
23.2; `call-process' races with SIGCHLD handler

Previous Next

Full log

View this message in rfc822 format

From: Paul Eggert <eggert <at> cs.ucla.edu>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: 9488 <at> debbugs.gnu.org, hanche <at> math.ntnu.no, 12980 <at> debbugs.gnu.org
Subject: bug#9488: 24.3.50; Zombie process left after call-process
Date: Thu, 29 Nov 2012 12:28:24 -0800

On 11/29/12 10:04, Eli Zaretskii wrote:

>> +      /* CHILD must be a child process that has not been reaped, and
>> +         STATUS and OPTIONS must be valid.  */
>> +      eassert (errno == EINTR);
> 
> Are we sure that either CHILD will have exited at this point, or else
> OPTIONS won't include WNOHANG?

It's unlikely that CHILD will have exited at this point.  That can
happen only if CHILD exited after the immediately-preceding waitpid
call and before this eassert.  In such a case, CHILD has exited but
has not been reaped.

It's common for OPTIONS to not include WNOHANG (wait_for_termination
does this), and when that happens it's almost always the case that
CHILD has not exited at this point.

> Can this function be ever called if
> neither of these conditions is true?

By "this function" I assume you mean get_child_status.  Yes, it's
quite common.  For example, Fcall_process invokes wait_for_termination,
and it's common for wait_for_termination to invoke get_child_status
before CHILD has exited, and without WNOHANG in OPTIONS.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.