On Mon, 30 Jul 2012 10:32:50 +0200 Dirkjan Ochtman wrote: > On Mon, Jul 30, 2012 at 10:23 AM, Richard Yao wrote: > >> I've always thought renaming python-3 to python3 is > >> faux-namespacing, and the thing SLOTs are supposed to help out > >> with. Why aren't SLOTs helping us with this? > > > > Portage will attempt to upgrade software to a newer SLOT if it will > > satisfy a dependency. This works when you cannot select versions via > > eselect, but it causes problems when you can. There is no way to > > tell it to prefer the selected version upgrades in other slots > > unless the selected version cannot satisfy it. > > Your last sentence fails to parse for me, perhaps expand one of the > "it"s? > > > I think that having to switch back would cause far less pain than > > the current situation would, assuming that we ever do. If the python > > developers refuse to make python 2.8, it is likely that someone > > else will. > > Please don't hope for a 2.8, it's simply not going to happen. > > >> I agree that installing both is probably overkill for most users. I > >> think the solution is somewhere outside the dev-lang/python > >> package, though, in having the system set or portage or whatever > >> the hell it is that first pulls in python prefer python-2. > > > > This would require amending the package manager specification. > > Well, maybe we should explore that option. It would seem to solve a > real problem that doesn't just apply to python. For example, the SLOT > value could be prefixed with something to indicate that it should not > be selected for upgrades automatically (i.e. other slots should be > preferred). [facepalm] Or maybe we should explore the option of fixing python.eclass to not depend on random python versions implicitly? -- Best regards, Michał Górny