Yep, I think it can be useful but unless we have a bullet proof
implementation (which we don't feel is a hack and might break), it's
best not to do it. I'm coming in kind of late so I'll look into
whether this was resolved as I read my emails.
Andi
At 12:35 PM 10/31/2005, Dmitry Stogov wrote:
> -----Original Message-----
> From: Derick Rethans [mailto:derick@xxxxxxx]
> Sent: Monday, October 31, 2005 11:32 PM
> To: Dmitry Stogov
> Cc: 'Zeev Suraski'; 'Sebastian Bergmann';
> zend-engine-cvs@xxxxxxxxxxxxx; 'Andi Gutmans'; 'Marcus
> Boerger'; 'Ilia Alshanetsky'
> Subject: RE: [ZEND-ENGINE-CVS] cvs: ZendEngine2 /
> zend_builtin_functions.c php-src NEWS
>
>
> On Mon, 31 Oct 2005, Dmitry Stogov wrote:
>
> > The problem is conceptual.
> > Before this patch PHP looked for EG(exception) after calling
> > constructor and disabled destructor calling.
> >
> > if (EG(exception) && ctor_opline) {
> > if (RETURN_VALUE_USED(ctor_opline)) {
> > EG(This)->refcount--;
> > }
> > if (EG(This)->refcount == 1) {
> > zend_object_store_ctor_failed(EG(This) TSRMLS_CC);
> > }
> > }
> >
> > Now the object also stored in backtrace so refcount is more then 1.
> > Probably we can check refcount with 2, but I am not sure that is a
> > proper way.
>
> We can also revert it simply, as I don't see a huge point in
> having the
> object *itself* in the backtrace...
>
> Derick
Of course we can. :)
But probably $this can be really useful in debug backtrace.
Dmitry.
--
Zend Engine CVS Mailing List (http://cvs.php.net/)
To unsubscribe, visit: http://www.php.net/unsub.php
|
