Has anyone ever experimented with or considered using self-Obsoletes as an alternative to Epoch bumps using plain rpm CLI and various depsolvers? Assuming there's a finite and known set of "supported" packages, listing all of them as an explicit, exact (note: always "=") Obsoletes: %{name} = $v-$r

On Tue, 2005-06-21 at 13:27 -0400, seth vidal wrote: > On Tue, 2005-06-21 at 19:06 +0300, Ville Skyttä wrote: > > Has anyone ever experimented with or considered using self-Obsoletes as > > an alternative to Epoch bumps using plain rpm CLI and various > > depsolvers? > > > > Assuming there's a finite and known set of "supported" packages, listing > > all of them as an explicit, exact (note: always "=") > > > > Obsoletes: %{name} = $v-$r > > well to make it catch older ones you'd need <= $v-$r > > and at that point wouldn't the package be obsoleting itself b/c of the > version confusion? Yes. But as I said above, always "=". Not "<=" or anything else.

On Tue, 2005-06-21 at 13:39 -0400, seth vidal wrote: > On Tue, 2005-06-21 at 20:38 +0300, Ville Skyttä wrote: > > On Tue, 2005-06-21 at 13:27 -0400, seth vidal wrote: > > > > > > and at that point wouldn't the package be obsoleting itself b/c of the > > > version confusion? > > > > Yes. But as I said above, always "=". Not "<=" or anything else. > > then you'll only catch that specific version - I can see that becoming > sticky once you have to limp them along for multiple releases. Multiple releases, yes, but not forever. If I understand "sticky" correctly in this context, that was also noted in my initial post. The question was whether the burden would be manageable and tolerable. To ease it a bit (but also limiting future options somewhat), dropping Release from the otherwise versioned Obsoletes would be possible in most cases.