[PATCH] To fix an assertion crash within a member template of a non-template class in ms-compatibility & delayed template parsing

[Richard Smith]

Thanks, I slightly simplified the test and inlined it, and committed as
r180701.


On Thu, Apr 25, 2013 at 4:00 AM, Faisal Vali <faisalv at gmail.com> wrote:

> Ok, adopted your suggestions.  Hope this is what you had in mind.
> Thanks for your patience!
>
>
> Faisal Vali
>
>
>
> On Thu, Apr 25, 2013 at 4:04 AM, Richard Smith <richard at metafoo.co.uk>wrote:
>
>> You can use CXXRecordDecl::hasAnyDependentBases instead of writing it out
>> for yourself. The large example in the comment seems unnecessary; the
>> existing comment and unit test are enough.
>>  On 24 Apr 2013 17:49, "Faisal Vali" <faisalv at gmail.com> wrote:
>>
>>> Of course, that approach seems to work!
>>>
>>> Patch is attached - thanks!
>>>
>>>
>>> Faisal Vali
>>>
>>>
>>>
>>> On Wed, Apr 24, 2013 at 9:58 AM, Richard Smith <richard at metafoo.co.uk>wrote:
>>>
>>>> Hmm, OK, I see. I think we should just not be calling
>>>> ActOnDependentIdExpression here in the first place. That is trying to cope
>>>> with access to a member of 'this' in a dependent base class, and there are
>>>> no dependent base classes. In ActOnIdExpression, we should be looking at
>>>> whether the class is a dependent context (or, better, whether it has
>>>> dependent bases), not whether the method is.
>>>>
>>>>
>>>> On Wed, Apr 24, 2013 at 7:44 AM, Faisal Vali <faisalv at gmail.com> wrote:
>>>>
>>>>> But that doesn't catch the case where the qualifier is the containing
>>>>> class itself (A::XX) i.e which also causes an assertion violation - for
>>>>> e.g. this does not work:
>>>>>
>>>>>  const bool IsInMicrosoftModeWithoutAQualifierLoc =
>>>>>                   getSema().getLangOpts().MicrosoftMode &&
>>>>>                   !QualifierLoc.getNestedNameSpecifier();
>>>>>
>>>>> Any thoughts?
>>>>>
>>>>> thanks!
>>>>>
>>>>> Faisal Vali
>>>>>
>>>>>
>>>>>
>>>>> On Wed, Apr 24, 2013 at 6:57 AM, Richard Smith <richard at metafoo.co.uk>wrote:
>>>>>
>>>>>> Hi!
>>>>>>
>>>>>> E->isTypeDependent() will always be true for a
>>>>>> CXXDependentScopeMemberExpr, so the patch would just turn off the
>>>>>> optimization in MicrosoftMode. I think the right thing to check here is
>>>>>> that we have a QualifierLoc.
>>>>>>
>>>>>>
>>>>>> On Wed, Apr 24, 2013 at 4:29 AM, Faisal Vali <faisalv at gmail.com>wrote:
>>>>>>
>>>>>>> Hi Richard,
>>>>>>>     was working on the TemplateParameterDepth patch and while
>>>>>>> crafting some tests, I stumbled upon an assertion violation out of CodeGen
>>>>>>> for the following code compiled with -fms-compatibility and
>>>>>>> -fdelayed-template-parsing
>>>>>>>
>>>>>>> Once again, this might be a very narrow fix, and there might be a
>>>>>>> more general way to address this issue - so welcome the feedback.
>>>>>>>
>>>>>>> Thanks!
>>>>>>> Faisal Vali
>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> cfe-commits mailing list
>>>>>>> cfe-commits at cs.uiuc.edu
>>>>>>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20130429/ef117111/attachment.html>