Florian Weimer <f...@deneb.enyo.de> writes: > But until these versions are explicitly described as invalid, shouldn't > the comparison algorithm cover them, so that different implementations > behave in the same way?
Oh, definitely. And I believe the textual description does cover them; there just aren't any examples. So maybe we should add examples with a note that hopefully no one is using constructs like that. > Other corner cases are empty epochs and zero epochs. And while not > really being a corner case in the Policy, APT compares ~ and ~0 as equal > at the end of a version string: > >>> apt_pkg.version_compare('1~', '1~0') > 0 That's also a good thing to mention. There are some oddities with -0 as a Debian revision too, if I remember correctly. Could someone open a bug against debian-policy for this so that we don't lose track and can hash out some of these details? -- Russ Allbery (r...@debian.org) <https://www.eyrie.org/~eagle/>