While I agree that the precise behaviour was probably slightly more intuitive, I don't think coreutils is a place we want to diverge from upstream. While you might argue that people who relied on the precise behaviour in local scripts need to fix their scripts on upgrade to trusty (true), I could readily argue that there's an unknown quantity of generic Linux software that relies on the upstream behaviour, and would break on precise but work on trusty.
In reality, both those groups are probably tiny but, again, diverging from upstream isn't the answer.
What we should do, IMO, is fix this to parse /proc/self/mountinfo instead (should be trivial, it's just moving the column numbers around a bit to extract the same info), and propose that patch upstream, with a rationale that the results match old mtab behaviour more closely and are, in general more human-readably "pretty".
While I agree that the precise behaviour was probably slightly more intuitive, I don't think coreutils is a place we want to diverge from upstream. While you might argue that people who relied on the precise behaviour in local scripts need to fix their scripts on upgrade to trusty (true), I could readily argue that there's an unknown quantity of generic Linux software that relies on the upstream behaviour, and would break on precise but work on trusty.
In reality, both those groups are probably tiny but, again, diverging from upstream isn't the answer.
What we should do, IMO, is fix this to parse /proc/self/ mountinfo instead (should be trivial, it's just moving the column numbers around a bit to extract the same info), and propose that patch upstream, with a rationale that the results match old mtab behaviour more closely and are, in general more human-readably "pretty".