On Tue, Jan 10, 2017 at 8:14 PM, Eli Schwartz via arch-general wrote:
On 01/10/2017 08:53 AM, Carsten Mattner via arch-general wrote:
My criticism of the stable patch queue is that they mix fixes with actual feature patches, making it more risky and not upholding a important fixes only policy.
That would depend on whether you understand "stable" to be "LTS" or "let's not just pile on all the experimental stuff that may break everything".
Since drivers are bundled in the kernel tree, we regularly run into many driver regressions and that's my primary objection to the missing quality assurance there. The community is doing an outstanding amount of testing already but the ranger of supported hardware is not covered by the testers and constant churn of code because it's part of a moving amalgamation in linux.git causes more issues than we would have with drivers targering a kernel ABI. One thing it would help make abundantly clear is when a driver maintainer stops supporting an old driver version. Now it's russian roulette for hardware to break when updating from one stable to the next supported stable kernel. Like it happened with 4.2 in DRM or the 4.9 boot problems which seem to be UEFI-exclusive.
I am pretty sure there is already, in fact, an LTS kernel. You even mentioned it yourself.
There are multiple LTS branches with one LTS being Greg's tree.