Archive for October, 2010
In my daily maintenance routine I tend to throw an emerge -uav world against the sabayon trees and see what packages can be bumped. I also check http://www.gentoo-portage.com to see what is new. In this routine 90% off all things I bump for Entropy it is done manually writing each emerge -av command by hand.
Since I trust Gentoo developers for doing a good job within their own little expertise and interest, I kinda trust each package bump makes sense. If it is either some revision bump because there was some LD flags to respect, a fix for –as-needed or simple another minor thing I just bump them. Even though on the binary end this would not make any difference for the user experience I just do it.
However this workflow sometimes leads to some stupid breakages because somebody thought it was a great idea to bump some lib that in fact is there for a reason. Some things in the Portage tree simply are there because some, or maybe one, of the atoms in the tree need them. There are numerous examples I ran into lately but here is one off them:
In my daily routine, some while back I bumped net-libs/enet. Not a big deal really. However this seemed to break games-action/supertuxkart and my tools could not detect this breakage. Otherwise I would obviously reverted this update. I went looking into some history and I realized that only games-action/supertuxkart ,games-rpg/egoboo and games-puzzle/enigma actually need net-libs/enet within the Portage tree, so why was this package bumped anyway? Not because it was needed for one of its parent atoms , no, because some user on the gentoo bugzilla requested this atom (remember only needed so far by 3 atoms) to be updated because “there was a newer version”. http://bugs.gentoo.org/325809 (just an example, do not shoot anybody for this!)
Somebody requested without any motivation this package to be updated and without question this was done. This all leads to atoms currently in Portage to mall function, fail to compile, etc. Later this all was corrected by Gentoo QA team where they slotted the enet version.
So what happens next is that users notice this when they install one off the games and it doesn’t work? They file a bug to Sabayon, or a Gentoo user files a bug to the Gentoo bugzilla. A bug wrangler needs to check the bug is valid enough and assign it to a herd/maintainer. The maintainer then discovers the unwanted enet bump and got to do all the work to block the newer version. All that kinda work because somebody thought it would be nice to bump a library/piece off software.
If users request such a version bump do not be scared to ask for a motivation.
It is nice to be bleeding edge, but really uncool to fall off the cliff just to look cool.