https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95980
--- Comment #14 from G. Steinmetz <gs...@t-online.de> --- > ... and real programmers wrote fine Fortran programs. Yeah, optimal world. That's probably the reason why some "real" programs don't need test cases - or at most three, because they cover everything ;-) Nobody likes "stupid" bugs. Aren't "smart" bugs from "smart" codes of "smart" people the better ones ?-) Some may proclaim 80% as the new standard threshold. But is 80% of 20% enough, and is there a VAT ?-) Most people underestimate the effort for thorough testing - not by a factor of 2, but by a factor of X or L or more. But enough with irony and sarcasm. --- Even the standards changed, too. F2018 has the audacity to demand chapter 4.2, item 2. I would like to see gfortran at a level where an average or casual programmer will not encounter an ICE with things like a forgotten/wrong keyword, attribute or syntactical element. It's easy enough to think logically wrong, to be in a hurry and mess things up. A compiler could be a helper and guide. No one said it wouldn't be hard and it won't take long. Thanks to all bold and brave people that work hard on it.
