tmv doesn't build

[Athanaseus]

Using compiler: /usr/bin/g++
compiler version: 9.3.0
Determined that there are 2 cpus, so use this many jobs.
You can override this behavior with scons -jN
Debugging turned off
Checking for MKL... no
Checking for ACML... no
Checking for GotoBLAS... no
Checking for CBLAS... yes
Using CBLAS
Checking for OpenMP library... yes
TMV Version  0.75

scons: done reading SConscript files.
scons: Building targets ...
g++ -o src/TMV_Vector.os -c -O2 -fno-strict-aliasing -fopenmp -fPIC -DTMV_NDEBUG -DCBLAS -DNOLAP -DINST_INT -Iinclude src/TMV_Vector.cpp
g++ -o src/TMV_MultVV.os -c -O2 -fno-strict-aliasing -fopenmp -fPIC -DTMV_NDEBUG -DCBLAS -DNOLAP -DINST_INT -Iinclude src/TMV_MultVV.cpp
g++ -o src/TMV_AddVV.os -c -O2 -fno-strict-aliasing -fopenmp -fPIC -DTMV_NDEBUG -DCBLAS -DNOLAP -DINST_INT -Iinclude src/TMV_AddVV.cpp
g++ -o src/TMV_MultXV.os -c -O2 -fno-strict-aliasing -fopenmp -fPIC -DTMV_NDEBUG -DCBLAS -DNOLAP -DINST_INT -Iinclude src/TMV_MultXV.cpp
g++ -o src/TMV_BaseMatrix.os -c -O2 -fno-strict-aliasing -fopenmp -fPIC -DTMV_NDEBUG -DCBLAS -DNOLAP -DINST_INT -Iinclude src/TMV_BaseMatrix.cpp
g++ -o src/TMV_Matrix.os -c -O2 -fno-strict-aliasing -fopenmp -fPIC -DTMV_NDEBUG -DCBLAS -DNOLAP -DINST_INT -Iinclude src/TMV_Matrix.cpp
g++ -o src/TMV_MultXM.os -c -O2 -fno-strict-aliasing -fopenmp -fPIC -DTMV_NDEBUG -DCBLAS -DNOLAP -DINST_INT -Iinclude src/TMV_MultXM.cpp
g++ -o src/TMV_AddMM.os -c -O2 -fno-strict-aliasing -fopenmp -fPIC -DTMV_NDEBUG -DCBLAS -DNOLAP -DINST_INT -Iinclude src/TMV_AddMM.cpp
g++ -o src/TMV_MultMV.os -c -O2 -fno-strict-aliasing -fopenmp -fPIC -DTMV_NDEBUG -DCBLAS -DNOLAP -DINST_INT -Iinclude src/TMV_MultMV.cpp
g++ -o src/TMV_Rank1_VVM.os -c -O2 -fno-strict-aliasing -fopenmp -fPIC -DTMV_NDEBUG -DCBLAS -DNOLAP -DINST_INT -Iinclude src/TMV_Rank1_VVM.cpp
g++ -o src/TMV_MultMM_CCC.os -c -O2 -fno-strict-aliasing -fopenmp -fPIC -DTMV_NDEBUG -DCBLAS -DNOLAP -DINST_INT -Iinclude src/TMV_MultMM_CCC.cpp
g++ -o src/TMV_MultMM_CRC.os -c -O2 -fno-strict-aliasing -fopenmp -fPIC -DTMV_NDEBUG -DCBLAS -DNOLAP -DINST_INT -Iinclude src/TMV_MultMM_CRC.cpp
g++ -o src/TMV_MultMM_RCC.os -c -O2 -fno-strict-aliasing -fopenmp -fPIC -DTMV_NDEBUG -DCBLAS -DNOLAP -DINST_INT -Iinclude src/TMV_MultMM_RCC.cpp
g++ -o src/TMV_MultMM_Block.os -c -O2 -fno-strict-aliasing -fopenmp -fPIC -DTMV_NDEBUG -DCBLAS -DNOLAP -DINST_INT -Iinclude src/TMV_MultMM_Block.cpp
g++ -o src/TMV_IntegerDet.os -c -O2 -fno-strict-aliasing -fopenmp -fPIC -DTMV_NDEBUG -DCBLAS -DNOLAP -DINST_INT -Iinclude src/TMV_IntegerDet.cpp
In file included from include/tmv/TMV_Base.h:99,
                 from include/tmv/TMV_BaseMatrix.h:39,
                 from include/tmv/TMV_Matrix.h:492,
                 from src/TMV_IntegerDet.cpp:26:
/usr/include/c++/9/complex: In instantiation of ‘std::complex<long double>& std::complex<long double>::operator=(const std::complex<_Tp>&) [with _Tp = int]’:
include/tmv/TMV_SimpleMatrix.h:108:26:   required from ‘tmv::SimpleMatrix<T>::SimpleMatrix(const M2&) [with M2 = tmv::GenMatrix<std::complex<int> >; T = std::complex<long double>]’
src/TMV_IntegerDet.cpp:177:26:   required from ‘T tmv::Bareiss_1x1_IntegerDet(const tmv::GenMatrix<T>&) [with T = std::complex<int>]’
src/TMV_IntegerDet.cpp:254:35:   required from ‘T tmv::IntegerDet(const tmv::GenMatrix<T>&) [with T = std::complex<int>]’
src/TMV_IntegerDet.inst:11:1:   required from here
/usr/include/c++/9/complex:1470:13: error: cannot convert ‘std::complex<int>’ to ‘std::complex<long double>::_ComplexT’ {aka ‘__complex__ long double’} in assignment
 1470 |    _M_value = __z.__rep();
scons: *** [src/TMV_IntegerDet.os] Error 1
scons: building terminated because of errors.

Should I open an issue upstream perhaps, even though developments have taken place about 3 years ago?

[gijzelaerr]

i think TMV is a library that was used by a package we don't include any more, what depends on this?

[Athanaseus]

GalSim uses this.

From docs it says: GalSim can use either Eigen or TMV for its linear algebra routines.
https://github.com/GalSim-developers/GalSim

[gijzelaerr]

galsim is not that important, so nice to have but otherwise we can drop it. you could try to get it to work with eigen.

[gijzelaerr]

lets just drop this package, we don't need it anymore.