Release RTL bodies after compilation (sometimes)

[Daniel Jacobowitz]

On Wed, Sep 15, 2004 at 11:36:45AM -0400, David Edelsohn wrote:
> >>>>> Daniel Jacobowitz writes:
> 
> Daniel> I don't think your "definition" is correct, or that marking objects as
> Daniel> explicitly free conflicts with the design of using a garbage collector.
> 
> 	Diego's definition is correct.  If the memory is not getting
> collected at the next GC phase, figure out why the GC still sees a
> reference.
> 
> 	If you want to tell the GC to collect the memory, then explicitly
> NULL what you believe are the references.  Add assertions that there
> should be no references to objects of type X at some appropriate barrier
> and fix any dangling references if objects of type X still exist.  What
> you are saying should be done with explicit ggc_free(), I (and I believe
> Jeff and Diego) are saying should be done by removing any references.  If
> you want to explicitly GC right after the reference is removed, fine.
> Just don't explicitly tell the garbage collector to free a particular
> object.

[Continued this discussion offline.]

I think that what David is suggesting, whether it's more ideally
correct or not, is infeasible today.  We have a certain obligation to
do what we can in the short term since no one has expressed interest in
massive overhauls.

But it's clear that everyone else with anything to say in this thread
disagrees with my position on ggc_free, so I will shut up now.

-- 
Daniel Jacobowitz