[cfe-commits] r119341 - /cfe/trunk/lib/Checker/GRCXXExprEngine.cpp

[Zhongxing Xu]

2010/11/17 Marcin Świderski <marcin.sfider at gmail.com>

> W dniu 17 listopada 2010 12:24 użytkownik Zhongxing Xu <
> xuzhongxing at gmail.com> napisał:
>
>
>>
>> 2010/11/17 Marcin Świderski <marcin.sfider at gmail.com>
>>
>>> W dniu 17 listopada 2010 12:08 użytkownik Zhongxing Xu <
>>> xuzhongxing at gmail.com> napisał:
>>>
>>>>
>>>>
>>>> 2010/11/17 Marcin Świderski <marcin.sfider at gmail.com>
>>>>
>>>> 2010/11/17 Zhongxing Xu <xuzhongxing at gmail.com>
>>>>>
>>>>>> On Tue, Nov 16, 2010 at 4:50 PM, Marcin Swiderski <
>>>>>> marcin.sfider at gmail.com> wrote:
>>>>>>
>>>>>>> Author: sfider
>>>>>>> Date: Tue Nov 16 02:50:38 2010
>>>>>>> New Revision: 119341
>>>>>>>
>>>>>>> URL: http://llvm.org/viewvc/llvm-project?rev=119341&view=rev
>>>>>>> Log:
>>>>>>> Refactored GRExprEngine::getCXXThisRegion to use
>>>>>>> CXXMethodDecl::getThisType instead of calculating it by hand.
>>>>>>>
>>>>>>
>>>>>> Please don't do this for now. CXXMethodDecl::getThisType() takes
>>>>>> qualifiers into account. But in the analyzer, we sometimes analyze methods
>>>>>> in isolation. So we always use unqualified type.
>>>>>>
>>>>>
>>>>>>
>>>>> What would this change impact? Maybe there should be a comment
>>>>> regarding this, because it's counter-intuitive.
>>>>>
>>>>
>>>> To be more precise, we cannot do this. We do not always have
>>>> CXXMethodDecl. We sometimes only have CXXRecordDecl when visiting
>>>> constructors.
>>>>
>>>>
>>> But what did you meant by analyzing methods in isolation? What would be
>>> impacted by using type of 'this' with const qualifier for methods that are
>>> const qualified?
>>>
>>
>> Using qualified this is not a problem itself.
>>
>> But in GRExprEngine::ProcessInitializer(), we only have CXXRecordDecl to
>> create the 'this' region.  So to uniform interface, we use CXXRecordDecl to
>> create 'this' region.
>>
>> So if a checker will try to get CXXThisRegion based on
> CXXMethodDecl::getThisType() returned type it will get different region then
> the one used in state?


I think so.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20101117/57e0c3b3/attachment.html>