snapshot 20050114 race (on list)

Pierre A. Humblet Pierre.Humblet@ieee.org
Sat Jan 15 03:20:00 GMT 2005


At 09:38 PM 1/14/2005 -0500, Christopher Faylor wrote:
>On Fri, Jan 14, 2005 at 08:17:30PM -0500, Pierre A. Humblet wrote:
>>The more I think about it, the less I see why you don't get rid of it.
>>PID_EXITED is a transient state into or out of PID_ZOMBIE.  From a user
>>interface point of view, it doesn't matter if PID_EXITED is shown as
>>zombie, the result might well be correct one ms earlier or later.
>>Internally Cygwin doesn't care either, except in one spot in sigproc.cc
>>(see original mail in thread).  But even there merging the states won't
>>hurt.
>
>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?
But even then, in Cygwin the process still goes briefly through the zombie
state.
And when the child has truly been waited on, the pinfo isn't set to EXITED
again, it is just closed.
 
>It's funny but one of the things that kept me putting this back was
>arguing this point with you in my head.  I guess my Pierre Humblet
>simulator isn't too good because I thought you'd be arguing that zombie
>should maintain the meaning that it has had previously in cygwin and in
>UNIX.

Perhaps I should be, but I seem to be missing the differences!

>I guess I'll think about this some more.  If not for the transient state
>when a process has a zero exitcode, we could almost use exitcode for
>everything.  I've been down that road, too.

There is something to be said to keep a simple external interface, with
process status flags, instead of relying on context.

Pierre
 
>cgf
>



More information about the Cygwin-developers mailing list