new panics

[Thierry Herbelot]

Le Friday 24 June 2005 22:56, Mike Silbersack a écrit :
>
> That makes some sense, because we may not be getting around to the
> modified memory until we hit some heavy memory usage.  Also, one backtrace
> showed the panic happening when uma_reclaim was called from the vm pageout
> daemon.  Sounds like I should throw something in to (optionally) call
> uma_reclaim on a regular basis so that we might catch this more quickly.

some light coming into the debate !

my test machine is somewhat low on memory (128Megs) and is used "headless", so 
the memory could only become rare somewhat late in the buildworld process.

	cheers,

	TfH