https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101660
Tobias Burnus <burnus at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |sandra at gcc dot gnu.org --- Comment #2 from Tobias Burnus <burnus at gcc dot gnu.org> --- The problem is the patch set for PR 101305, which moved from a static libgfortran/ISO_Fortran_binding.h to a generated $build/$arch/libgfortran/ISO_Fortran_binding.h And in the include file: -#include "../../../libgfortran/ISO_Fortran_binding.h" +#include "ISO_Fortran_binding.h" I am not quite sure whether that should be "..." or now "<...>". Notes: * When doing incremental builds, ISO_Fortran_binding.h was not build; I had to delete $build/{*,*/*}/libgfortran/ (or $build to do a proper bootstrap) to get it generated. * Here, it does work in that case with in-build-tree testing, except that with testing -m64 and -m32 at the same time, the wrong header is used, cf. https://gcc.gnu.org/pipermail/gcc-patches/2021-July/576194.html It would be useful to know: * whether you did an incremental build or not (cf. above) * whether your are testing in build tree or with the installed compiler. → If the latter, does changing "ISO_Fortran_binding.h" to <ISO_Fortran_binding.h> help? * Do you use RUNTESTFLAGS - in particular --target_board= to run multiple archs at the the same time (→ email above)