home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.unix.wizards
- Path: sparky!uunet!decwrl!deccrl!news.crl.dec.com!pa.dec.com!decprl!decprl!boyd
- From: boyd@prl.dec.com (Boyd Roberts)
- Subject: Re: Queueing of signals
- Message-ID: <1992Aug14.112212.1265@prl.dec.com>
- Keywords: SIGCHLD, signals, queueing, sigvec
- Sender: news@prl.dec.com (USENET News System)
- Nntp-Posting-Host: prl313.prl.dec.com
- Organization: Digital Equipment Corporation - Paris Research Laboratory
- References: <1992Aug13.160824.11077@bwdls61.bnr.ca> <1992Aug14.072625.26349@fcom.cc.utah.edu>
- Date: Fri, 14 Aug 1992 11:22:12 GMT
- Lines: 23
-
- I wasn't going to followup to this thread because the whole `signal
- on child death' is a quagmire, but:
-
- In article <1992Aug14.072625.26349@fcom.cc.utah.edu>, terry@cs.weber.edu (A Wizard of Earth C) writes:
- > You know, if I had any sense, I'd be going to bed right this instant...
- >
-
- Hmm...
-
- > 1) Signals aren't queued
-
- Well it depends. On System V `death of child' persists until all your
- zombies have been waited for. And it's near indistinguishable from a
- queue when your signal handler is coded correctly; you get one signal
- per child death. You must ensure that you wait _before_ you reset
- the SIGCLD handler because signal(2) will search the process table
- looking for any outstanding zombies, which in turn generates another
- signal. Get this wrong and your stack explodes.
-
-
- Boyd Roberts boyd@prl.dec.com
-
- ``When the going gets wierd, the weird turn pro...''
-