NetNews Usenet Archive 1992 #18/NN_1992_18.iso/spool/comp/unix/programm/4286

Internet Message Format | 1992-08-13 | 3.0 KB

Path: sparky!uunet!olivea!apple!bionet!uwm.edu!cs.utexas.edu!ut-emx!ibmchs!auschs!awdprime.austin.ibm.com!ravikm
From: [e-mail] (Ravi K Mandava)
Newsgroups: comp.unix.programmer
Subject: Re: Children and waitpid()'s
Message-ID: <[e-mail]>
Date: 12 Aug 92 16:56:56 GMT
References: <[e-mail]>
Sender: [e-mail] (USENET News)
Reply-To: piobe![e-mail]
Organization: IBM Austin
Lines: 56
Originator: [e-mail]


In article <[e-mail]>, [e-mail] (Kevin Wang aka The Scarecrow) writes:
 > exit value of the child.  My problem however is this:  the child exits
 > before the parent can do a waitpid() on it, and I cannot get the exit value.
 >
 > So I guess my quesions are thus:
 > 1) under what circumstances does the child wait for the parent to ackowledge
 > it's death (via waitpid())?

    Child process becomes a zombie (defunct) process, and remains like that
    until either the parent does a wait()/waitpid() for it, or until the
    parent exits.  That is, even if the child process exits before the
    parent process calls waitpid(), the child process remains in defunct
    state until the parent reads its exit status.  This is *only* true,
    when the parent does not ignore the SIGCHLD (or SIGCLD) signal.
    However, if the parent ignores the SIGCHLD signal (see man pages on
    sigaction()), the child exits (and its process table entry is
    immediately cleaned up) without passing its exit status to the parent.
    That is, a wait() call by a parent would block until the child exits,
    but does not fetch child's exit status.  Instead, errno is set to
    ECHILD (no child processes).

 > 2) will a waitpid() on one process allow another to exit?

    No, unless the process id that is passed to it is not a positive
    integer.  Man pages on waitpid() discuss all the possible scenarios
    in this case.

 > 3) any better way to send messages to another program on another machine?
 > (no filesystem access, only sockets)

    Sockets or TLI are the best bet, I guess.

 > 4) is it possible for signal(SIGCHLD...) to affect it?

    I presume you mean, if signal(SIGCHLD...) affects the behaviour of
    a child being able to pass its exit status to the parent process.
    As I discussed before, only if sigaction()/signal() is used to ignore
    SIGCHLD, a child's exit status is not returned to its parent.

    However, one good suggestion is always to set up a signal handler
    for SIGCHLD, in which a wait() can be performed to fetch the exit
    status of a child and let it be purged from the system process table
    immediately.

 > I am running on (ick) AIX 1.2.1

    The above comments should generally apply to any standard Unix/AIX
    flavours.  But I could be wrong.  Would someone please correct me, if
    so?

Hope this helps,
--
*******************************************************************************
Ravi K Mandava        email:    piobe![e-mail]
*******************************************************************************