[Dune] [Dune-Commit] dune-grid r7145 - trunk/dune/grid/common

[Martin Nolte]

Hi Oli,

no offense taken. But discussing everything in the FlySpray for two weeks will 
result in such things never being fixed (like my little patch to the 
MCMGMapper ;-)). And, as I said, errors resulting from this "misuse" are very 
hard to track.

Best,

Martin

On 01/07/2011 02:34 PM, Oliver Sander wrote:
> Hi Robert, hi Martin!
> I agree with you that the public operators were bugs that needed fixing.
> But I disagree when you say that fixing a bug in the interface is not an
> interface change.
> This discussion is not academic, because we actually do use the assignment
> operator in our code. Is it anywhere to be seen that we were not supposed to?
> Hence your change broke our code, without a prior warning.
>
> Don't get me wrong: I am not complaining, just trying to inform you about
> the background of Carsten's mail.
>
> Best,
> Oliver
>
> Am 07.01.2011 14:25, schrieb Martin Nolte:
>> Hi Carsten,
>>
>> yes, it is very late. And, as Robert stated, I don't consider this an
>> interface change. The reason for changing is a "bug" in the dune-grid-howto
>> code:
>>
>> const Intersection i = *intersectionIterator;
>>
>> This failed to work with GeometryGrid. So I had to decide: Is this an error
>> in GeometryGrid or is this kind of use disallowed. My understanding of the
>> interface is the second and, since this "bug" was hard to find, I changed
>> the interface implementation accordingly.
>>
>> The final discussion should also depend on the outcome of the "Remove
>> EntityPointer" suggestion, which was postponed until the next developer
>> meeting. So, I suggust to postpone the issue until then.
>>
>> Finally, in my optionion, such discussions belong into the FlySpray to
>> simplify voting (and my vote here is obvious).
>>
>> Best,
>>
>> Martin
>>
>>
>> On 01/07/2011 11:12 AM, Carsten Gräser wrote:
>>> Hi,
>>> altough this comment is a little late:
>>>
>>> Was there any discussion on this interface change?
>>> Even if we agree to forbid the copy constructor
>>> and assignment they should at least remain public
>>> for the next release.
>>>
>>> In case we want to remove them: Is anyone aware of
>>> a mechanism that generates warnings only if it's
>>> used in a public context?
>>>
>>> Best,
>>> Carsten
>>>
>>>
>>> Am 27.11.2010 13:56, schrieb mnolte at dune-project.org:
>>>> Author: mnolte
>>>> Date: 2010-11-27 13:56:58 +0100 (Sat, 27 Nov 2010)
>>>> New Revision: 7145
>>>>
>>>> Modified:
>>>> trunk/dune/grid/common/intersection.hh
>>>> Log:
>>>> make copy constructor protected. Intersections, like entities, may not be
>>>> copied or assigned to
>>>>
>>>>
>>>> Modified: trunk/dune/grid/common/intersection.hh
>>>> ===================================================================
>>>> --- trunk/dune/grid/common/intersection.hh 2010-11-27 12:45:59 UTC (rev 7144)
>>>> +++ trunk/dune/grid/common/intersection.hh 2010-11-27 12:56:58 UTC (rev 7145)
>>>> @@ -407,10 +407,6 @@
>>>> /** Copy Constructor from IntersectionImp */
>>>> Intersection(const IntersectionImp<const GridImp> & i) :
>>>> real(i) {};
>>>> -
>>>> - /** Copy constructor */
>>>> - Intersection(const Intersection& i) :
>>>> - real(i.real) {}
>>>> //@}
>>>>
>>>> typedef typename remove_const<GridImp>::type mutableGridImp;
>>>> @@ -430,6 +426,17 @@
>>>> //! return reference to the real implementation
>>>> const ImplementationType& getRealImp() const { return real; }
>>>>
>>>> + /* hide copy constructor */
>>>> + Intersection ( const Intersection&i )
>>>> + : real( i.real )
>>>> + {}
>>>> +
>>>> + /* hide assignment operator */
>>>> + const Intersection&operator= ( const Intersection&i )
>>>> + {
>>>> + real = i.real;
>>>> + return *this;
>>>> + }
>>>> };
>>>
>>> _______________________________________________
>>> Dune mailing list
>>> Dune at dune-project.org
>>> http://lists.dune-project.org/mailman/listinfo/dune
>>
>
>
> _______________________________________________
> Dune mailing list
> Dune at dune-project.org
> http://lists.dune-project.org/mailman/listinfo/dune

-- 
Martin Nolte <nolte at mathematik.uni-freiburg.de>

Universität Freiburg                                   phone: +49-761-203-5630
Abteilung für angewandte Mathematik                    fax:   +49-761-203-5632
Hermann-Herder-Straße 10
79104 Freiburg, Germany