[Issue 3463] Integrate Precise Heap Scanning Into the GC

d-bugmail at puremagic.com d-bugmail at puremagic.com
Thu Apr 14 13:04:17 PDT 2011


http://d.puremagic.com/issues/show_bug.cgi?id=3463



--- Comment #97 from Andrei Alexandrescu <andrei at metalanguage.com> 2011-04-14 13:00:36 PDT ---
(In reply to comment #96)
> (In reply to comment #95)
> > (In reply to comment #94)
> > > 
> > > I think that covers things, except for handling ambiguous pointers.
> > 
> > Can you explain why we care about ambiguous pointers?  That is, shouldn't we
> > just always consider that an ambiguous pointer is a pointer?  Why do we need a
> > separate designation?
> 
> To allow moving collectors (you can't update a pointer that's not really a
> pointer).
> 
> Anyway, this problem is taking care of in the implementation by nfxjfg, which
> Walter seem to keep ignoring, as he always did with this bug report and all the
> useful insight it have. I'm not saying the solution is the best, but it's
> completely ignored for no (publicly know reason).
> 
> I'll unsubscribe now from this bug report, as it really makes me sick =)
> 
> If you want to contact me, your know where to find me...

That's a fairly random thing to say, particularly now that this bug report is
again attracting interest. Walter has expressed a concern about the proposed
implementation, which implies he's definitely not ignoring it. If you wanted to
release some pent-up frustration accumulated from the past, fine. Otherwise,
the reaction is rather irrational. It would be in everybody's interest (and it
would definitely make you feel better) if you continued adding value to this
discussion and code.

Generally, after having accumulated more information on conservative vs.
precise collectors, I think D must aim at being as precise as possible - no ifs
and buts. So this report and patch must be taken to completion. I'll get back
with some more comments.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------


More information about the Digitalmars-d-bugs mailing list