https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77261
--- Comment #2 from Janne Blomqvist <jb at gcc dot gnu.org> --- (In reply to Dominique d'Humieres from comment #1) > AFAIU this occurs for platforms without real(10) and is due to the fact that > the sequence for real(16) is different from the sequence for other reals. Yes, I agree with this analysis. The test case already checks for the presence of fortran_large_real, but we need to distinguish whether selected_real_kind (precision (0.0_8) + 1) is 10 or 16.