[aspectc-user] reset tjp->that()?

[Olaf Spinczyk]

Hi,

yes, the AspectC++ join point API exposes the address of function
arguments (even if the actual argument type is unknown, because the
advice code is generic). It also exposes the address of the object on
which a method is invoked. However, you can't get the address of the
address, because it is not guaranteed that the method was invoked by
using a pointer.

Olaf

Michael Gong schrieb:
> "pass-by-value" is fine, because you might pass the value of a
> pointer, and deference it inside the advice body.
>
> Just for you information, AspectC ( not AspectC++) has a mechansim to
> expose the address of a function parameter to the advice body, like:
>
> before(int * p): args(*p) && call(void foo(int)) {
>                        <-- p will have the address of the parameter
> passed to foo
>        *p = 3;    <-- this will actually change the value of the
> parameter passed to foo
> }
>
> before(int p): args(p) && call(void foo(int)) {
>                        <-- p will have the value of the parameter
> passed to foo
>        p = 3;       <-- this will NOT change the value of the
> parameter passed to foo
>
> }
>
> I am not sure whether AspectC++ could do that or not.
>
> Anyway, it is an interesting problem.
>
> Regards,
>
> Mike
>
> ----- Original Message ----- From: "Yan Mao" <maoyan at gmx.net>
> To: "Michael Gong" <mwgong at cs.toronto.edu>
> Cc: <aspectc-user at aspectc.org>
> Sent: Saturday, December 09, 2006 4:40 PM
> Subject: Re: Re: Re: [aspectc-user] reset tjp->that()?
>
>
>> Hello Mike,
>>
>> just as Daniel said, AC++ use pass-by-value in advice body, then it
>> is hard to change the object pointer. Maybe it's a inveterate
>> problem? Although I would like to use AC++ to do something with those
>> pointers.
>>
>> Regards,
>>
>> Yan
>>
>>
>> -------- Original-Nachricht --------
>> Datum:  Sat, 9 Dec 2006 13:24:08 -0500
>> Von: "Michael Gong" <mwgong at cs.utoronto.ca>
>> An: "Yan Mao" <maoyan at gmx.net>, "Daniel Lohmann"
>> <daniel.lohmann at informatik.uni-erlangen.de>
>> Betreff:  Re: Re: [aspectc-user] reset tjp->that()?
>>
>>> Hi, Daniel & Yan,
>>>
>>> It brings an interesting question.
>>>
>>> The template works fine for this case.
>>>
>>> > I am afraid what you are trying to achieve is not possible.
>>> > // Template, works for any class that offers a Release() method
>>> > template< class T > void Release( T*& _this ) {
>>> > _this->Release(); // destroy
>>> > _this = 0; // invalidate
>>> > }
>>> >
>>>
>>> But how about I want to do the "invalidation" for method "Release2" ,
>>> "Release3", or etc ? Using template, you might have :
>>>
>>> template< class T > void Release2( T*& _this ) {
>>>  _this->Release2(); // destroy
>>>  _this = 0; // invalidate
>>>  }
>>>
>>> template< class T > void Release3( T*& _this ) {
>>>  _this->Release3(); // destroy
>>>  _this = 0; // invalidate
>>>  }
>>>
>>> ...
>>>
>>>
>>> Looks familiar ? :-)
>>>
>>> It is typical crosscutting concern, I think.
>>>
>>> Since AOP is targeted to modularize crosscutting concern, can it be
>>> used
>>> for this case ? More specifically, can AspectC++ capture it ?
>>>
>>> Regards,
>>>
>>> Mike
>>>
>>>
>>>
>>> ----- Original Message ----- From: "Yan Mao" <maoyan at gmx.net>
>>> To: "Daniel Lohmann" <daniel.lohmann at informatik.uni-erlangen.de>
>>> Cc: <aspectc-user at aspectc.org>
>>> Sent: Saturday, December 09, 2006 11:18 AM
>>> Subject: Re: Re: [aspectc-user] reset tjp->that()?
>>>
>>>
>>> Hello Daniel,
>>>
>>> thank for the explation, it's very helpful.
>>>
>>> regards,
>>>
>>> Yan Mao
>>>
>>>
>>>
>>> -------- Original-Nachricht --------
>>> Datum:  Fri, 08 Dec 2006 17:06:02 +0100
>>> Von: Daniel Lohmann <daniel.lohmann at informatik.uni-erlangen.de>
>>> An: aspectc-user at aspectc.org
>>> Betreff:  Re: [aspectc-user] reset tjp->that()?
>>>
>>> > Hi Yan,
>>> >
>>> > I am afraid what you are trying to achieve is not possible.
>>> >
>>> > > //the C++ code
>>> > > ClassA *p = new ClassA();
>>> > > p->DoSomething();
>>> > > p->Release();
>>> > > //now i want to set p=NULL with AOP
>>> >
>>> > In C++ the object instance pointer (this) is always passed >
>>> by-value.
>>> > Technically, it is passed as a "hidden" first parameter:
>>> >
>>> > class A {
>>> >    void Release();
>>> > };
>>> >
>>> > becomes (technically) something such as
>>> >
>>> > void ClassA::Release( ClassA* this ) {
>>> > // this is by-value parameter
>>> > }
>>> >
>>> > The object reference inside some method (this) is a copy of the
>>> > original
>>> > parameter. The same holds for the value returned by tjp->that() > in
>>> > advice code, it is another value copy of the originally passed
>>> > instance
>>> > pointer. It is not possible to affect the "original" from advice >
>>> code.
>>> >
>>> > Side node:
>>> > The common C++ idiom to invalidate the reference when destroying an
>>> > instance is to use a non-member function:
>>> >
>>> > // Template, works for any class that offers a Release() method
>>> > template< class T > void Release( T*& _this ) {
>>> > _this->Release(); // destroy
>>> > _this = 0; // invalidate
>>> > }
>>> >
>>> > //the C++ code
>>> > ClassA *p = new ClassA();
>>> > p->DoSomething();
>>> > Release(p);
>>> > // now p == 0
>>> >
>>> >
>>> > Daniel
>>>
>>> -- 
>>> "Ein Herz für Kinder" - Ihre Spende hilft! Aktion:
>>> www.deutschlandsegelt.de
>>> Unser Dankeschön: Ihr Name auf dem Segel der 1. deutschen America's
>>> Cup-Yacht!
>>> _______________________________________________
>>> aspectc-user mailing list
>>> aspectc-user at aspectc.org
>>> http://www.aspectc.org/mailman/listinfo/aspectc-user
>>
>> -- 
>> "Ein Herz für Kinder" - Ihre Spende hilft! Aktion:
>> www.deutschlandsegelt.de
>> Unser Dankeschön: Ihr Name auf dem Segel der 1. deutschen America's
>> Cup-Yacht!
>>
>
> _______________________________________________
> aspectc-user mailing list
> aspectc-user at aspectc.org
> http://www.aspectc.org/mailman/listinfo/aspectc-user