I started a discussion on the forum http://forums.gentoo.org/viewtopic.php?t=168779&highlight= but I think it belongs to bugs.gentoo.org. Anyway.... here a samll abstract of the topic: What is wrong with the version numering in Gentoo? Why can't I install package-[major].[minor].[micro]-r[ebuild-revision] and be sure that if I remerge the SAME version in 1 month, everything will be the same? Why do the Gentoo developers change a ebuild and then they don't increase the revision? Just today I removed PostgreSQL from one of my servers and then remerged net-mail/postfix net-mail/courier-imap net-ftp/pure-ftpd dev-php/mod_php dev-php/php. And what do I see? Changed rc.init scripts all over the place! But I did not update or upgrade or installed a fresh version of the packages. NO! SAME version as I had bevore! EXACTLY SAME VERSION WITH EXACTLY SAME REVISION! Or on my desktop I had kdesu not working correctly. And after some search in the forums, I see that the ebuild for kdelibs changed and has now a fix for that problem. But once again: The ebuild version did NOT change! Same ebuild version as before, but diffrend functionality! Or I remember the time when VMWare needed this any-any driver to work correctly with the 2.6.x series of kernel. And all the time the exactly same ebuild got changed and changed and changed to include an new version of the any-any driver, instead of adding a new revision of the ebuild and include the new any-any driver that way. I can understand that alot of people would freak out, if they would need to recompile a package, just because something changed wich is not a big issue. But Gentoo is in my eye not an hobby OS. But why does it behave sometime like this then? I miss consisteny! I miss it big time. Okay... now I feel relieved, but still not happy about the versioning mess. cheers SteveB Reproducible: Always Steps to Reproduce: 1. 2. 3.
Can we have major revisions and minor revisions, and an option in make.conf to ignore minor revisions? Personally, I think that would be a much more elegant solution. A major revision bump would cause all users to have to upgrade as it does now; a minor revision bump would cause only users who have the option set in make.conf to watch for minor revisions to upgrade. IMHO, this would be much better behaviour as, as has been said above, the current approach lacks consistency.
I'm not a fan of multiple types of revisions in packages. If QA or a bunch of developers in general make a push for it, we can consider it. I'd suggest better/clearer policy.
As a temporary (?) workaround: http://home.jesus.ox.ac.uk/~ecatmur/revcheck.sh
EDIT: Stupid Rob. http://home.jesus.ox.ac.uk/~ecatmur/my-bin/revcheck.sh
robert, your URL isn't working, can you just attach the script to this bug report please?
Created attachment 31924 [details] revcheck.sh script Hmmm... well, it works here. But sure, will do.
Spanky spells is out pretty well in #31074 but it's also pretty clear from policy that rev bumps are at the discretion of the developer/maintainer. In specific cases, the wrong decision may have been made. Please file bugs about specific packages. Thanks. *** This bug has been marked as a duplicate of 31074 ***
