Thanks Vladimir, that's interesting (particularly about Rubinius).

Question:

Is this the same issue as reported here?
http://rubyforge.org/tracker/index.php?func=detail&aid=15425&group_id=426&atid=1698
and discussed here:
http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-core/15325 (see
also http://svn.ruby-lang.org/cgi-bin/viewvc.cgi?view=rev&revision=10689
)

I found that undoing those changes to parse.y fixes the leak.  But
that change was made to fix a problem which I would now re-introduce I
presume.  Unfortunately I don't know what that original problem was
apart from it was fixed by preventing local_tbl from being garbage
collected as per the cvs comment.

Can anyone shed any light?

- Stephen


On Sun, Mar 23, 2008 at 9:33 PM, Vladimir Sizikov <vsizikov / gmail.com> wrote:
> Hi Stephen,
>
>  Out of curiosity, I tried the test case, and here are my results:
>
>  MRI 1.8.6 pl 111, memory grows with 80Mb/sec rate (!!!)
>  Latest MRI 1.8 branch: memory grows with 80Mb/sec rate too
>  MRI 1.9 - memory doesn't grow
>  JRuby 1.1RC2 - memory doesn't grow
>  Rubinius - memory grows at 500Kb/sec
>
>  Thanks,
>   --Vladimir
>
>