------- Comment #2 from fxcoudert at gcc dot gnu dot org 2005-12-10 17:52 ------- (In reply to comment #1) > Confirmed. I guess this is because rec=n is transferred to the library as a > gfc_integer_4, but it should be gfc_offset.
Yes, I was beginning to think about that, but it has strongly implications on the general front-end/library configury, I guess. See my last mails in ml thread "Generalizing the library to arbitrary floating point modes" (the topic deviated!). -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25289
