RE: memory leaks in fork(?)

["Dave Korn" <dk at artimi dot com>]

> -----Original Message-----
> From: cygwin-owner On Behalf Of Luc Hermitte

> * On Fri, Jan 30, 2004 at 04:25:20PM +0600, Mike Jastrebtsoff 
> wrote:
> > While fulfilling of any building procedure under cygwin via make, 
> > configure(utils, that widely use spawning of other 
> processes) appears 
> > memory leak, that leads to termination of building process: "fork:
> > permission denied" or sometimes to system reboot. [...]
> > 
> > I saw similar messages in maillist archive, but I haven't reveal 
> > concrete solutions for these items. What to do? Wait for 
> next release 
> > of cygwin?
> 
> If you are using Agnitum Outpost Firewall ... try uninstall 
> it a use another firewall.
> On this new machine, (where XP has been reinstalled from 
> scratch), I'm no longer using AOF, but Kerio. Since then, I 
> don't need to reboot every two days or every half-make-configure.
> So far, the hyppothesis of a ressource leak in AOF seems accurate.

  I can confirm that there are definitely memory management problems in
Agnitum outpost.

  I downloaded it a few months ago and tried to install it on a machine on
which I was also running the M$ driver verifier.  Within a couple of seconds
of completing the installation - before even rebooting for the first time -
driver verifier BSOD'd the machine because AOF was either writing past the
end of its allocated memory or rewriting memory after it had freed it, I
can't remember precisely which.

  My machine then went into an endless cycle of reboot, BSOD, reboot, BSOD,
as AOF messed up every time it got started up and so I couldn't get to the
desktop to disable driver verifier!  To repair my machine, I had to boot in
safe mode and disable the device driver that AOF had installed, then reboot
in normal mode and uninstall it.

  I can't understand how AOF could be released with such bugs in it.  I
don't know whether the people at Agnitum simply didn't test their software
under driver verifier, or whether it didn't show this behaviour when they
did, but I believe it speaks poorly to their quality control standards.



    cheers, 
      DaveK
-- 
Can't think of a witty .sigline today....
 


--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/