[cfe-commits] [llvm-branch-commits] [cfe-branch] r168830 - in /cfe/branches/release_32: ./ lib/Sema/TreeTransform.h test/SemaTemplate/instantiate-overload-candidates.cpp

[Pawel Wodnicki]

On 11/28/2012 11:09 PM, Douglas Gregor wrote:
> 
> On Nov 28, 2012, at 9:09 PM, Pawel Wodnicki <pawel at 32bitmicro.com> wrote:
> 
>> On 11/28/2012 10:41 PM, Douglas Gregor wrote:
>>>
>>> On Nov 28, 2012, at 8:08 PM, Pawel Wodnicki <pawel at 32bitmicro.com> wrote:
>>>
>>>> Takumi,
>>>>
>>>>>> Modified: cfe/branches/release_32/test/SemaTemplate/instantiate-overload-candidates.cpp
>>>>>> URL: http://llvm.org/viewvc/llvm-project/cfe/branches/release_32/test/SemaTemplate/instantiate-overload-candidates.cpp?rev=168830&r1=168829&r2=168830&view=diff
>>>>>> ==============================================================================
>>>>>> --- cfe/branches/release_32/test/SemaTemplate/instantiate-overload-candidates.cpp (original)
>>>>>> +++ cfe/branches/release_32/test/SemaTemplate/instantiate-overload-candidates.cpp Wed Nov 28 17:44:46 2012
>>>>>> @@ -19,3 +19,34 @@
>>>>>> void test(int x) {
>>>>>>  f(&x, 0);
>>>>>> }
>>>>>> +
>>>>>> +// Ensure that we instantiate an overloaded function if it's selected by
>>>>>> +// overload resolution when initializing a function pointer.
>>>>>> +template<typename T> struct X {
>>>>>> +  static T f() { T::error; } // expected-error {{has no members}}
>>>>>> +  static T f(bool);
>>>>>> +};
>>>>>> +void (*p)() = &X<void>().f; // expected-note {{instantiation of}}
>>>>>
>>>>> It has been introduced in r167918 and causes failure in release_32.
>>>>>
>>>>
>>>> Yeah, I am was just looking at this.
>>>>
>>>>> http://bb.pgr.jp/builders/clang-3stage-x86_64-linux/builds/74
>>>>> --
>>>>> error: 'error' diagnostics expected but not seen:
>>>>> Line 26: has no members
>>>>> error: 'note' diagnostics expected but not seen:
>>>>> Line 29: instantiation of
>>>>> 2 errors generated.
>>>>> --
>>>>>
>>>>> Pawel, I suggest you a couple of options;
>>>>>
>>>>> 1) Remove the extra test.
>>>>>
>>>>> 2) Apply Richard's r167918, too. Doug and Richard, how do you think?
>>>>
>>>> I think applying r167918 is the best way and I'll test it but let's wait
>>>> till Doug and Richard had a chance to look at this.
>>>
>>> r167514 is small, looks good, and fixes a regression. Let's take it.
>>>
>>> 	- Doug
>>>
>>>
>> Doug,
>> Did you have r167918 in mind rather then r167514 ?
>> Pawel
> 
> Yes, sorry. Pasted the wrong revision number (but reviewed the right one!).

r167918 - Committed revision 168888.
And we have diagnostics back where they are
expected.
> 
> 	- Doug
> 
> 
> 

Pawel