snapshot 20050114 race (on list)

Pierre A. Humblet Pierre.Humblet@ieee.org
Sat Jan 15 16:35:00 GMT 2005


At 10:34 PM 1/14/2005 -0500, you wrote:
>On Fri, Jan 14, 2005 at 10:20:01PM -0500, Pierre A. Humblet wrote:
>>At 09:38 PM 1/14/2005 -0500, Christopher Faylor wrote:
>>>Zombie isn't a cygwin invention, though.  Something could be displayed
>>>as a zombie that would never actually be a zombie.
>>
>>Like what?  The case where SIGCHLD is ignored?
>
>If a parent starts a child and then the parent exits while the child is
>still running.

OK, it's needed there. But in that case the pinfo is released immediately
so it's extremely unlikely to be observable.

It's observable if another process already had the pinfo open (e.g. race
between
exit and ps or kill). In such a case the pinfo can survive after the end of
the
process. To cover such an event, the test in pinfo init 
!created && createit && (procinfo->process_state & PID_EXITED)
should also include PID_ZOMBIE.

Pierre



More information about the Cygwin-developers mailing list