*BSD News Article 29233


Return to BSD News archive

Path: sserve!newshost.anu.edu.au!munnari.oz.au!metro!sunb.ocs.mq.edu.au!macadam.mpce.mq.edu.au!acagney
From: acagney@macadam.mpce.mq.edu.au (Andrew Cagney - aka Noid)
Newsgroups: comp.os.386bsd.bugs
Subject: Race condition in IO code???
Followup-To: comp.os.386bsd.bugs
Date: 4 Apr 1994 16:30:37 GMT
Organization: School of MPCE, Macquarie University, Australia.
Lines: 28
Distribution: world
Message-ID: <2npfbd$6d9@sunb.ocs.mq.edu.au>
NNTP-Posting-Host: macadam.mpce.mq.edu.au
Summary: race condition in stdio?
Keywords: 386bsd NetBSD stdio
Originator: acagney@macadam.mpce.mq.edu.au


Hi,

I'm running a _very_ long (especially on a 386) compilation using GNU
make on a NetBSD-0.9 system vis:

	bash$ which make
	/usr/local/make-3.70/bin/make
	bash$ nice make dependInstall > log.dependInstall 2>&1 &
	bash$ tail -f log.dependInstall

from time to time, both `make' and `cc1' drop core (bus error or
sigsegv) eg:

	make[1]: Entering directory `/home/local/src/auis-6.2'
	make: *** [World] Bus error (core dumped)

Where/when this occures is totally random and can't be reproduced.
The only clue I've found is that when I don't run the `tail' command
the problem appears to go away....

Is there a bug (somewhere :-) in the IO code that has been fixed in a
later version of *BSD that could be causing this problem?

			any help appreciated,

					Andrew