https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81336
--- Comment #3 from bburgerm at googlemail dot com --- Yes, I'm using this option because the gfortran code should be included in a larger project with a mixture of C, C++, gfortran and partially intel fortran code with several libraries only available in binary form. Due to the lack of gfortran function attributes to define the symbol explicitly, using -fno-underscoring was the most successfull solution. Using bind(C) did not work for all functions because gfortran is quite picky regarding variable types (like it is not possible to pass a fortran user defined type with an allocatable variable to a C function which only passes it to a fortran function without reading or touching it. For example a dlopen-wrapper). As mentioned above I have read the warnings about -fno-underscoring and only filed a bug report because the bug is already visible for the small attached program only using libraries that are also part of gcc. If it would have happened for our large project only, then I would accept it as the described incompatibilities. Btw: The same problem appears for omp_set_num_threads using the pointer value as number of threads if -fno-underscoring is used from fortran which usually leads to a crash.
