On 1/26/11 3:14 AM, Ryan Hill wrote: > On Tue, 25 Jan 2011 12:38:03 +0100 Tomáš Chvátal > <scarab...@gentoo.org> wrote: > > Won't this just pile on more work on already stressed to the max arch > teams? As in, now they have to stabilize more packages to get back to > where they were in the first place?
This seems to be a self-balancing system to me. If the arch team is so stressed that it can't stabilize something within 90 days, and can't even state a reason for that, just move the package back to testing. After some time, the stable set for that arch should be small enough to let the arch team handle it on time. > And as I understand it, the reason maintainers are complaining is > because they want to drop old versions. I'm not sure why maintainers are complaining, but generally managing bugs that sit there for a long time is harder. > Meaning stable users of these archs can suddenly lose large parts of > the tree if this happens. From their point of view, we've just > broken perfectly working systems. That's pretty much the opposite of > what stable is supposed to promise. That's an important point. I think that a message should be sent somewhere (gentoo-dev-announce?) that something like that is going to happen, and wait some 60 days for someone to save the package. > And I've never been an arch tester, but I bet after the first few > times I tested a package only to have it dropped to ~arch because no > developer was around to commit the keyword change, I wouldn't feel > much like doing it anymore. Good point. > How about the opposite? If everyone but $arch has stabilized the > package, and you can't get a response from them in a reasonable time, > then use your discretion as maintainer and mark it stable yourself. Very dangerous, especially for exotic arches. I think we should not go that way, or at least _require_ the maintainer to test on that arch. We have some development machines for various arches so it should be technically possible. But it generally seems to me that maintainers miss more problems than arch testers/developers. > Arch testers would remain useful by giving the maintainer some > measure of assurance that they won't accidently break anything for > that arch. Good point, again provided the maintainer at least compile-tests the package on that arch. Paweł
signature.asc
Description: OpenPGP digital signature
