On Sun, Aug 21, 2016 at 10:23 AM, Peter Humphrey <pe...@prh.myzen.co.uk> wrote: > On Sunday 21 Aug 2016 07:28:17 Rich Freeman wrote: > >> ... there is nothing wrong with having some internal QA on kernel >> releases. 4.1 had a nasty memory leak a release or two ago that was >> killing my system after only an hour or two uptime. They took over a >> week to stabilize the fix as well (though a patch was out fairly quickly). >> So, I'm not in nearly the rush to update kernels as I used to be > > I've formed the impression that a good many kernel updates are mainly just > to incorporate code for new devices, so I don't rush into it normally > either. However, this box does have some hardware that's not yet a year old, > so I do keep this one up to date.
The 3rd decimal almost never has code for new devices. It is intended to be 100% bugfixes. So, you generally don't want to be too far behind on that. > >> (granted, unless you read all the lists it is easy to miss this sort of >> thing). > > Do you recommend any in particular for this? Gentoo-dev, perhaps? Nope. That memory leak was on lkml I think. Only reason I spotted it was that I searched for it after getting bitten by it. I doubt I'd have even noticed the thread but for looking for it. I do tend to search the btrfs lists before switching between series, because that is the thing I figure is most likely to break. Honestly, kernel QA could be better in some ways. When some crippling bug comes along they don't always rush to release fixes, and they don't have any way to communicate with end-users. They just assume that distros are paying attention to that sort of thing. And most probably are (probably including Gentoo, but I'm not running a Gentoo kernel since the Gentoo kernels aren't really going to be optimized for btrfs stability). > >> I ended up bailing on gentoo-sources all the same. Not that there was >> really anything wrong with it, but since I'm running btrfs and they've >> had a history of nasty regressions that tend to show up MONTHS later >> I've been a lot more picky about my kernel updates. I'm currently >> tracking 4.1. I might think about moving to 4.4 in a little while. > > Well, according to eix, there's only 4.4.19 between 4.1.30 and 4.7.2. Those are just the versions packaged for Gentoo. kernel.org has 4.4.19 as the only non-EOL version in-between, and it is longterm (I think 4.7 is too, but it isn't marked as such yet). > > Sound policy, I'm sure. How does an ordinary mortal know which versions are > here for the long term? > kernel.org The Gentoo team will not let down ordinary users. If you're using semi-experimental features then you're best off keeping a close eye on upstream no matter what distro you use. -- Rich
