From: Hugh Sasse <hgs@...>
Date: 2007-10-03T01:39:35+09:00
Subject: Re: gc.c -- possible logic error?

On Wed, 3 Oct 2007, Joel VanderWerf wrote:

> Hugh Sasse wrote:
> ...
> > So a whole load of components of data get marked by the GC, but don't
> > get freed in the free function, because only the frame component
        [...]
> 
> Mark and free are not symmetric. Mark is for objects referred to by self
> (which might also be referenced elsewhere); free is for internal data that
> can't be directly referenced by other ruby objects. For example, an array's
> mark func must mark the objects that are elements of the array (and any
> instance vars of the array). The free func releases the storage of the array
> itself (the C array of VALUEs), but can't touch the objects because someone
> else might still be referring to them.

Thanks, that confirms the results of my experiment which failed, and ties
the low level aspects into the high level mental model I had.

With the tools I have here, I am now at a loss to explore this
non-freeing of objects further.  I could do basic experiments with
1.9 if I could find it :-) but I've not found the svn/tarball of it
from wandering round ruby-lang.org

        Thank you,
        Hugh