The presence of this bug forces us to compile all of our libraries
twice: one time with GCC 5 and one time with Clang 4. Now I wonder how
high my chances are that someone implements a compile time switch for
Clang to emit GCC 5 compatible object files? And getting that change
into the stable branch of Clang 4?
Or could you envision a different band-aid like an linker script?