[cfe-commits] [PATCH] GRExprEngine dummy this object region

[Zhongxing Xu]

2010/11/16 Marcin Świderski <marcin.sfider at gmail.com>

> W dniu 16 listopada 2010 09:04 użytkownik Marcin Świderski <
> marcin.sfider at gmail.com> napisał:
>
>
>>
>> 2010/11/16 Zhongxing Xu <xuzhongxing at gmail.com>
>>
>>
>>>
>>> 2010/11/16 Ted Kremenek <kremenek at apple.com>
>>>
>>>
>>>> On Nov 15, 2010, at 4:26 PM, Marcin Świderski wrote:
>>>>
>>>> W dniu 16 listopada 2010 01:14 użytkownik Ted Kremenek <
>>>> kremenek at apple.com> napisał:
>>>>
>>>>>
>>>>> On Nov 15, 2010, at 3:52 PM, Marcin Świderski wrote:
>>>>>
>>>>> W dniu 16 listopada 2010 00:44 użytkownik Ted Kremenek <
>>>>> kremenek at apple.com> napisał:
>>>>>
>>>>>> Hi Marcin,
>>>>>>
>>>>>> What is this memory region used for?  There are no comments above the
>>>>>> declaration of this class, so I'm having difficulty interpreting its
>>>>>> intended use.
>>>>>>
>>>>>> On Nov 15, 2010, at 3:42 PM, Marcin Świderski wrote:
>>>>>>
>>>>>> > Patch adds new MemRegion subclass CXXDummyThisObjectRegion, which is
>>>>>> used to initialize CXXThisRegion value for initial stack frame.
>>>>>> >
>>>>>> > Please aprove for commit.
>>>>>> >
>>>>>> > -Marcin
>>>>>> > <gr-eng-cxx-dummy-this-obj.patch>
>>>>>>
>>>>>> Hi Ted
>>>>>
>>>>> In general it's used for identifing object pointed with 'this' in
>>>>> analyzed method. I use it for checking in ctor/dtor if method is called on
>>>>> 'this'. I can write some comment for it before commiting.
>>>>>
>>>>>
>>>>> Why not use a symbolic region for the value of 'this'?  Why do we need
>>>>> a new type of memory region?  We can just have RegionStore::Retrieve() map
>>>>> from 'this' to a SymbolicRegion (or one wrapped with an ElementRegion that
>>>>> adds the typing information).
>>>>>
>>>>
>>>> I couldn't think of any other solution that wouldn't involve tricks.
>>>>
>>>> The current implementation of RegionStore::Retrieve() maps from 'this'
>>> to a SymbolicRegion. Is it sufficient?
>>>
>>> I didn't grasp the SymbolicRegion and SymbolRegionValue until now. Using
>> SymbolicRegion will be probably the best solution. Thanks Zhongxing.
>
>
> New patch. Added mapping from 'this' to it's SymbolicRegion in
> BasicStoreManager::getInitialStore.
>
> Please approve for commit.
>

Looks OK. I'm not sure if it makes sense to add partial support for C++ in
basic store. But if this is what you need, it seems not affect other parts
of the analyzer.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20101117/fbb5dd5c/attachment.html>