Re: [eigen] Feeback to Rev #914777. (pre RC1)

[Benoit Jacob <jacob.benoit.1@xxxxxxxxx>]

2009/1/21 FMDSPAM <fmdspam@xxxxxxxxx>:
> Hi,
>
> here are some more of type b):
>
> Warning    224    warning C4244: 'argument' : conversion from 'double' to
> 'const float', possible loss of data   FILE c:\Develop\Eigen\test\lu.cpp
>  LINE 39
> FILE c:\Develop\Eigen\test\lu.cpp    LINE 45
>
> And at the same time at the same places:
> Warning    226    warning C4244: 'argument' : conversion from 'double' to
> 'const std::complex<float>::_Ty', possible loss of data   FILE
> c:\Develop\Eigen\test\lu.cpp    LINE 39
> FILE c:\Develop\Eigen\test\lu.cpp    LINE 45

These were valid complaints, fixed.

>
> Warning    231    warning C4244: 'argument' : conversion from 'int' to
> 'const std::complex<float>::_Ty', possible loss of data   FILE
> c:\Develop\Eigen\test\submatrices.cpp    LINE 143, 144, 146, 147
> FILE c:\Develop\Eigen\test\sum.cpp    LINE 37

This one is abusive, MSVC doesn't want me to construct a
complex<float> from an int even explicitly.
Still, other warnings of this type have been useful, so i made a workaround.

>
> Warning    237    warning C4700: uninitialized local variable 'dummy' used
> FILE c:\develop\eigen\eigen\src\core\arch\sse\packetmath.h    LINE 67

sounds valid -- except it has no importance in practice since it's
after an abortion.
i make a workaround.

>
> Warning    239    warning C4181: qualifier applied to reference type;
> ignored   FILE c:\Develop\Eigen\test\meta.cpp  LINE 48, 50

aha, this is a good catch. we wrote const Ref with Ref=T&, and we
meant const T& i.e. the const applied to T, but that's not what we
got.

I remove these 2 lines, I think they're just redundant.

> By the way: Many times  warning "C4003: not enough actual parameters for macro 'throw'" occurs.

OK, this was caused by a workaround for GCC so not needed for MSVC
anyway. Solved with a #ifdef.

All that's in r914899.
Wow, that's a nice amount of actual issues found thanks to your testing!

Cheers,
Benoit