RE:trying to solve the pytango FTBFS with gcc5
> I have created a small test project (attached) that has a library exporting > "std::string Tango::ranges_type2const::str" symbol and a test program > using it. thanks a lot. now we can see if gcc and g++ generate different symbols. > Interestingly, GCC uses the symbol _ZN5Tango17ranges_type2constIjE3strE > (*without* the B5cxx11 part) for both library and client. S> o now I wonder why does libtango have B5cxx11 in it symbols and what is it > needed for. Because it was compiled with the new cxx11 ABI. so the question is why pytango is not compile with the new API :) Cheers Frederic
RE:trying to solve the pytango FTBFS with gcc5
> The actual error is about another symbol: > _PyTango.so: undefined symbol: _ZN5Tango17ranges_type2constIjE3strE > This symbol is old ABI, as opposed to > _ZN5Tango17ranges_type2constIjE3strB5cxx11E > (which *does* exist in libtango.so.8). > Can it be a bug in GCC? I don't think it should produce such a broken > _PyTango.so. Yes I have this feeling also. I do not understand why gcc produce something with the old ABI. Except thaht python build extensions with gcc and not g++. Maybe this is the problem. gcc and g++ do not behave the same when compiling c++ files ? Fred
RE:trying to solve the pytango FTBFS with gcc5
> I have created a small test project (attached) that has a library exporting > "std::string Tango::ranges_type2const::str" symbol and a test program > using it. > Interestingly, GCC uses the symbol _ZN5Tango17ranges_type2constIjE3strE > (*without* the B5cxx11 part) for both library and client. You are right the symbol dos not use the B5cxx11 part even with gcc5 ??? > So now I wonder why does libtango have B5cxx11 in it symbols and what is it > needed for. Yes why this B5cxx11 in tango and not in pytango ???