https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70684
--- Comment #4 from Andy May <ajmay81 at googlemail dot com> --- Yes, I agree that demonstrates the bug - and I see it gives the desired output with 4.8.3 but not with 5.3.0. However, I would actually not mind if that modified testcase continued not to 'work' on Linux since this is equivalent to writing the file on Windows and reading on Linux, which although it's nice that it once did work, it's arguable it should not have (although if it doesn't breach the Fortran standard I guess there is no problem to have it). I think the only case that definitely should work, which currently does not on Windows, is writing and reading on the same system. In other cases one should be expected to run dos2unix/unix2dos on the file so that it's in the correct state before being read.
