https://gcc.gnu.org/bugzilla/show_bug.cgi?id=50410
--- Comment #28 from Dominique d'Humieres <dominiq at lps dot ens.fr> --- I have revisited the draft patch in comment 8 and it fixes all the ICEs, but those in comment 23. The last two of them emit now in gfc_format_decoder, at fortran/error.c:934 libbacktrace could not find executable to open I have found a typo in the patch "initalized" should be replaced with "initialized". With the patch the following tests gfortran.dg/data_constraints_1.f90 gfortran.dg/data_initialized.f90 gfortran.dg/data_initialized_2.f90 gfortran.dg/data_invalid.f90 gfortran.dg/data_pointer_1.f90 regress. I'll attach the patch fixing these regressions, noting that there is probably something wrong for gfortran.dg/data_pointer_1.f90. The patch seems to break some gnu extensions (but AFAICT not all them). BTW what is the best expression in English "already is" or "is already", the later being easier to parse for French readers.
