Hi,

while I haven't noticed RAM getting eaten I did notice crashes, yes. :-$
My current 'pet candidate' would be the pmap code btw, as documented by
pr's 29019 - 29021.

My Amigas' kernel is currently at 30M (on a 96M 060), I'll check tomorrow 
whether it has grown.

One 'fun' thing I noticed once (with no immediate crash following) was
that the system had swapped out most processes, was doing essentially
nothing, had 'real' mem free and wasn't swapping in anything for quite a
while (not even processes that got the order to go kill themselves).
This resolved itself after a while, as if the swapper had finally woken up
and decided to come back to work. :-} During all this the machine pinged
fine, so it's possibly related to the hangs Florian Stoehr saw.

Things that might help nail this down:
- put DDB, DIAGNOSTIC, DEBUG in the kernel. If you have another computer 
  with a free serial interface (and a nullmodem cable) consider SERCONSOLE. 
  (conserver from pkgsrc is nice to look after such a port, but cu or tip 
  will also work well, just remember to keep it running so you grab the 
  panic message)
- run vmstat -s every once in a while to watch the development (save the
  output)
- when the machine has crashed, get a trace and a show uvmexp

any more suggestions? amiga doesn't have a show pmap has it?

regards,
	spz
-- 
spz@serpens.de (S.P.Zeidler)