Rubinius' GC...oddity

[rogerdpack]

I was following the thread at

http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-core/27498

And I thought to try Rubinius' GC on it, thinking that it would
perform well on the following code:

(1..7).each{|n| puts n; a = ['a']*(10**n); a.inspect;}

However, on my virtualbox Ubuntu it seems to eat up ~ 500MB of RSS and
start swapping (MRI finishes in 135s with about 80MB RSS, Jruby
finishes in 8s with about 215MB RSS).

Is this an edge case or something more?
Much thanks.
-r