CUDA build errors with GCC 8

[Antoine Martin]

Similar to #1600, here's the new error:

found gcc version: 8.0.1
found CUDA compiler: /usr/local/cuda/bin/nvcc version 9.1.85
CUDA compiling ARGB_to_NV12     (no file)
 '/usr/local/cuda/bin/nvcc' '-fatbin' \
    '-c' 'xpra/codecs/cuda_common/ARGB_to_NV12.cu' \
    '-o' 'xpra/codecs/cuda_common/ARGB_to_NV12.fatbin' \
    '-std=c++11' \
    '-gencode=arch=compute_30,code=sm_30' \
    '-gencode=arch=compute_35,code=sm_35' \
    '-gencode=arch=compute_50,code=sm_50' \
    '-gencode=arch=compute_52,code=sm_52' \
    '-gencode=arch=compute_53,code=sm_53' \
    '-gencode=arch=compute_60,code=sm_60' \
    '-gencode=arch=compute_61,code=sm_61' \
    '-gencode=arch=compute_62,code=sm_62' \
    '-gencode=arch=compute_70,code=sm_70'
Error: failed to compile CUDA kernel ARGB_to_NV12
/usr/include/c++/8/type_traits(1061): error: type name is not allowed
/usr/include/c++/8/type_traits(1061): error: type name is not allowed
/usr/include/c++/8/type_traits(1061): error: identifier "__is_assignable" is undefined

3 errors detected in the compilation of "/tmp/tmpxft_00001aee_00000000-14_ARGB_to_NV12.compute_70.cpp1.ii".

error: Bad exit status from /var/tmp/rpm-tmp.F9Iw1o (%build)

[Antoine Martin]

Raising, also affects Fedora 28.

[Antoine Martin]

​Compiling CUDA with clang does not help, because as per ​CUDA Compilers compatibility matrix, clang 6 supports CUDA 9 but not 9.1

[Antoine Martin]

Build workaround in r19119 + r19120 (+fixups in r19121 + r19122): we can keep the compiled binary kernels in the source tree by excluding them from the build clean command, so Fedora 28 and later can re-use pre-compiled kernels generated using Fedora 27.

[Antoine Martin]

Workaround in r19685: switch to C++ 03 instead of C++ 11...
(found some hints there: ​https://negativo17.org/nvidia-driver/)

[migration script]

this ticket has been moved to: ​https://github.com/Xpra-org/xpra/issues/1800