[gclist] Finalizer flame wars.

Nick Barnes nickb@harlequin.co.uk
Mon, 13 May 1996 09:27:07 +0100


> > The first time I read this I assumed it was a joke, but I suppose with
> > weak pointers it could make sense. Weak pointers are, of course, a
> > very useful feature which render GCs non-transparent.
> 
> Let us suppose a garbage collected system is leaving a monster
> storage footprint to the point where the program is useless. Hardly
> an unknown phenomia. Asking why some objects are being kept alive
> by the collector is smart. You may be able to zero some pointers etc.
> 
> Information on why all objects are being kept alive is a good way
> to reduce monster footprints. Face it people put in conservative
> collectors to reduce storage leaks and often get the opposite. 

I agree with what you say; my observation was this: the fact of the
mutator asking the question "why is this object alive?" is enough to
keep the object alive. Unless there are weak pointers (or, say, the
question is being asked by a debugging process).

Nick Barnes