Re: problem with changeset fa917f1f0faf

[Carnë Draug]

On 11 August 2016 at 16:03, Ben Abbott <address@hidden> wrote:
>> On Aug 10, 2016, at 21:29, Carnë Draug <address@hidden> wrote:
>>
>> On 11 August 2016 at 01:51, Ben Abbott <address@hidden> wrote:
>>>> On Aug 10, 2016, at 8:37 PM, Carnë Draug <address@hidden> wrote:
>>>>
>>>>> On 11 August 2016 at 00:47, Ben Abbott <address@hidden> wrote:
>>>>> Carne,
>>>>>
>>>>> I think I’m seeing an error associated with the change below.
>>>>>
>>>>>       http://hg.savannah.gnu.org/hgweb/octave/rev/fa917f1f0faf
>>>>>
>>>>> [...]
>>>>>
>>>>> liboctave/array/MatrixType.cc:530:23: error: assigning to 'double' from 
>>>>> incompatible type 'complex<double>'
>>>>>                   d = std::conj (d);
>>>>>                     ^ ~~~~~~~~~~~~~
>>>>
>>>> How does your compiler handles this?
>>>>
>>>>   #include <iostream>
>>>>   #include <complex>
>>>>   #include <typeinfo>
>>>>
>>>>   int main()
>>>>   {
>>>>     std::cout << typeid (std::conj (double (5.0))).name() << std::endl;
>>>>   }
>>>>
>>>> If you get back a complex number, the issue is discussed on DR 1137 [1].
>>>> You should get back d for double.  I only found this DR number because
>>>> it's refered on the GCC implementation of conj.
>>>>
>>>> Carnë
>>>>
>>>> [1] https://lwg.github.io/issues/lwg-defects.html#1137
>>>
>>> [...]
>> [...]
>>
>> I will guess that replacing std::conj with octave::numeric::conj will also
>> trigger this issue.  Octave's conj() is now just imported from std [2].
>> If my guess is right, then I guess this means that octave::numeric::conj
>> is never used in Octave for real types.
>>
>> Carnë
>>
>> [2] 
>> http://hg.savannah.gnu.org/hgweb/octave/file/e43d83253e28/liboctave/numeric/lo-mappers.h#l64
>
> I assume this traced back to changeset 9c3a3d252e80?
>

If replacing std::conj with octave::numeric::conj on MatrixType does not
fix the issue, then yes.  The issue is at 9c3a3d252e80 and the fix should
be there.  Can you confirm that there's an issue there?