Re: GWeakNotify fired earlier than expected

[IdaRub <idarub gmail com>]

On Mon, Mar 16, 2009 at 5:51 PM, Havoc Pennington
<havoc pennington gmail com> wrote:
> Hi,
>
> On Mon, Mar 16, 2009 at 12:45 PM, IdaRub <idarub gmail com> wrote:
> > After reading the gobject code a bit more, it seems that weak
> > references are fired on dispose, not finalize:
>
> Right, that's correct.
>
> > All of the documentation I could find states that it happens during
> > finalization, but it seems it should all say that it happens during
> > dispose.
>
> Probably true. dispose is typically done during finalize, but can
> happen sooner also. dispose can happen multiple times.
>
> > In my example this detail is actually very important.
> > Doesn't look like there is a way to catch dispose without shimming the
> > instance handler.  Any ideas?
>
> You mean there's no way to catch finalize?
>
> Finalize deliberately can't be intercepted, because it raises a lot of
> thorny issues about reentrancy; what if the object is used or its
> refcount increased during finalization?

It would be nice if you could intercept it, the author being
responsible for understanding what finalize means, and not doing
anything inappropriate.  I was attempting to write some ownership
debugging code, basically something like leak and use-after-free
detection for a few specific object.  We'll catch this other ways
though, so I suppose I will drop this idea.

Thanks for your great reply.

> So things are split into two phases, dispose() which you can get
> notifications about (and which can happen multiple times); and
> finalize which actually frees the object.
>
> Havoc