From: Bob Hutchison <hutch@...>
Date: 2002-05-24T03:26:41+09:00
Subject: Re: Stymied by Ruby's garbage collector

On 5/23/02 2:09 PM, "Paul Brannan" <pbrannan@atdesk.com> wrote:

> On Fri, May 24, 2002 at 02:28:42AM +0900, Patrick May wrote:
>> A gc language prevents memory leaks, it makes no performance guarantees :-)
> 
> It helps to alleviate memory leaks, but it does not prevent them.  I can
> still get memory leaks by forgetting to remove a reference to an object
> from a hash, or by not being careful when registering finalizers.

Funny how meanings shift over time. Yes what you are saying is a usage of
the term that is occasionally used these days. However, once upon a time,
there were *real* 'memory leaks' and 'dangling pointers'. When allocated
memory had no reference to it that was leaked memory -- there was no finding
that memory again, so no deallocation of it. When you had a pointer that
pointed at deallocated memory you had a dangling pointer -- who know what
would happen but nothing good (the deallocated memory would be
re-allocated). Both exceedingly difficult to debug, even detect sometimes.
Both are completely prevented by GC. The meaning you use is not the same
thing at all.