Re: [eigen] compiler error C2516 for EIGEN_WORKAROUND_MSVC_STL_SUPPORT in details.h

[Radu Bogdan Rusu <rusu@xxxxxxxxxxxxxxxx>]

Hauke, thanks a _lot_ for fixing this so fast!

Are we getting closer to a releasable beta3? :)

Btw, while testing the latest snapshot, it seems that this code (provided by Gael some months ago) is no longer compiling:

https://code.ros.org/svn/ros-pkg/stacks/perception_pcl/trunk/pcl/include/pcl/common/eigen.h

ei_sqrt, ei_cos, ei_sin, etc... do not longer exist.

Have they been renamed to something else or are we doing something wrong?

[Edit: it looks like they might have been renamed to sqrt/cos/sin - is that correct? If so, we definitely need a beta3 
out soon :)]

Thanks,
Radu.


On 12/02/2010 03:43 AM, Hauke Heibel wrote:
> Hi Stefan,
>
> thanks for the test-program. I pushed a fix which hopefully works. ;)
>
> The issue was that for std::vector<T>  we tried at some point to derive
> from T which works only if T is a class whereas you used a "float". I
> agree that this should compile and work but initially the std::vector
> workaround was intended for Eigen-types only.
>
> A last comment on the fix which is primarily aimed at Benoit and Gael.
> Currently I am assuming that if  "!internal::is_arithmetic<T>::value"
> is exactly than true when "std::is_calss<T>::value" is true. I am not
> sure whether this is actually correct but I am also a bit at a loss of
> how to implement "is_class". The STL code here is so cryptic for that
> traits helper and contains so many defines that I quit looking after
> some time. Maybe you guys have an opinion whether we can live for the
> moment with my hack or whether we should do something else.
>
> - Hauke
>
>
> On Wed, Dec 1, 2010 at 2:32 AM, Stefan Holzer<holzers@xxxxxxxxxxxxxxxx>  wrote:
> > Hi,
> >
> > I'm trying to port some code (http://www.ros.org/wiki/pcl) from Linux to
> > Windows (Win7 x64 with Visual Studio 2010, compiling for 32-bit). This works
> > fine except for a compiler error I get in Eigen/src/StlSupport/details.h in
> > line 71:
> >
> > error C2516: 'T' is not a legal base class
> >
> > which occurs for the vector:
> >
> > std::vector<float, Eigen::aligned_allocator<float>  >  values;
> >
> > According to the comments in this file this piece of code is specially
> > written for Visual Studio. Since the code works fine in Linux there might be
> > something wrong with this VS-specific code. I also tested it with Visual
> > Studio 2008 and it gives the same error.
> >
> > I would be happy about any suggestions.
> >
> > Thanks,
> > Stefan