🚀 go-pugleaf

In article <FOX.92Apr12181014@graphics.nyu.edu>, fox@cs.nyu.edu (David Fox) writes:
|> In article <1992Apr10.052219.5103@cubetech.com> andrew@cubetech.com (Andrew Loewenstern) writes:
|> 
|>    In article <FOX.92Apr8102030@graphics.nyu.edu> fox@cs.nyu.edu (David Fox)  
|>    writes:
|> 
|>    >  I set hiwat and it had absolutely no effect.  Even if hiwat did
|>    >  work, the tiny difference between NeXT and Sun is that the Sun
|>    >  will release the swap space once it is finished.  The NeXT just
|>    >  fills up the space available and then dies.
|>    >  
|>    >  -david
|> 
|> 
|>    No it doesn't.  Although NeXTstep doesn't physically make the swapfile smaller  
|>    it does reuse the space within it:
|> 
|>    Filesystem            kbytes    used   avail capacity  Mounted on
|>    /dev/sd1a              98442   97172       0   104%    /private/vm/swapdisk
|> 
|>    This is my swapdisk.  It has been like this for several days and I've been  
|>    using tons of vm since then (I can hear the drive grind away when I use vm).   
|>    If what you said was correct, my machine would have gone down in flames a long  
|>    time ago.  Although the swapfile itself is huge, there is lots of free space in  
|>    it.  Even if hiwat worked, what is the poor computer supposed to do when it  
|>    runs out of memory?  It could die a little bit more gracefully though.
|> 
|> Ok, the thing is, I need that 100 meg back.  I would have preferred to
|> be told I couldn't run a process becauses there is insufficient swap
|> space rather than have it run and then fill up both swap _and_ disk
|> _and_ die a slow horrible death.  I've often exhausted swap on SunOS
|> and been told so with no adverse effect apart from the mercifully quick
|> death of a (randomly selected?) process.  (I haven't tried setting
|> hiwat correctly yet, but I haven't heard anything good about the
|> result of doing _that_ either.)
|> 
|> -david

If you set hiwat correctly you will get just that -- the process will grind
to a halt without filling the harddisk. You will usually have to kill it
manually, though. This happens once in a while with Mathematica :)
To set the hiwat in /etc/swaptab, read the man page, AND notice that there
can be no spaces after the commas in the arg list...
-- 




			Magnus Nordborg
			Department of Biological Sciences
			Stanford University
			Stanford, CA 94301-5020
			USA
			(415) 723-4952
			magnus@Kimura.Stanford.EDU