Daniel, > ... but the value read with g in your first test (read_float_2.f03) > does not match the expected one, although no error is issued. ... Jerry found a regression in trunk fixed at r143103 (see pr38735). This may be what you see. Dominique