math/eigen3 recent update

Fernando Apesteguía fernando.apesteguia at gmail.com
Tue Jun 27 20:50:31 UTC 2017


On Tue, Jun 27, 2017 at 12:22 PM, Tobias C. Berner <tcberner at freebsd.org> wrote:
> Hi
>
> It seems, that if I add the patch (on 12-current)
> #####
> --- cmake/code_eli.hpp.in.orig  2017-06-27 10:15:58 UTC
> +++ cmake/code_eli.hpp.in
> @@ -29,6 +29,7 @@
>  // This section is dedicated to disabling warnings from Eigen that
> still show up in various
>  // build configurations.
>  //
> +#include <Eigen/Core>
>  #include <Eigen/src/Core/util/Macros.h>
>  #define EIGEN_NUM_VERSION
> (EIGEN_WORLD_VERSION*10000+EIGEN_MAJOR_VERSION*100+EIGEN_MINOR_VERSION)
>  #if defined _MSC_VER
> ##### EOF
>
> and disable cpptest
> (CMAKE_ARGS=-DCMAKE_DISABLE_FIND_PACKAGE_CPPTest:BOOL=TRUE) it builds
> again.
> [without disabling ctests I get static assert failures -- so maybe
> something to look at too].

Hi Tobias,

Thanks for the patch. I'll try it out and look at the other issues.

Best regards

>
>
> mfg Tobias
>
> On 26 June 2017 at 17:32, Fernando Apesteguía
> <fernando.apesteguia at gmail.com> wrote:
>> Hi there,
>>
>> I'm the porter and maintainer of graphics/code-eli and cad/openvsp,
>> both depending on math/eigen3. After the last update to math/eigen3
>> both ports fail to compile:
>>
>> http://package19.nyi.freebsd.org/data/103i386-default-build-as-user/444250/logs/code-eli-0.3.6_3.log
>>
>> The ports compile fine in 11.x but started to fail in 10.x. According
>> to the log there's some combination of __CUDACC__ definitions that
>> make the EIGEN_DEVICE_FUNC "symbol" to be unavailable.
>>
>> I tried using gcc-c++11-lib in the Makefile. It tries to compile the
>> port with g++5 but fails again.
>>
>> I temporarily fixed cad/openvsp by using the code-eli tarball included
>> in the distribution bundle.
>>
>> Can you help me out here?
>>
>> Thanks in advance.


More information about the kde-freebsd mailing list