On Fri, Nov 13, 2015 at 4:37 AM, Karanbir Singh mail-lists@karan.org wrote:
The key to working out a plan for Uday's kernel is going to be driven by how he intends to support it and do maintenance work in that space.
Taking this opportunity, I'd like to share my recent experience that might show what it's like to maintain non-distro kernels.
For those who are not familiar, ELRepo (Alan Bartlett) maintains the latest mainline kernel for RHEL/CentOS/Scientific Linux as 'kernel-ml' [1]. When the first release candidate for kernel 4.2 (4.2-rc1) came out, Alan immediately noticed it did not build under RHEL 6. Long story short, a piece of kernel code in the netfilter section was not compatible with gcc 4.4.x (used in EL6) [2]. Apparently, kernel developers did not test-build against "older" versions of gcc.
Despite the fact the patch was submitted at the early stage of the 4.2 RC process, the fix did not make it into kernel 4.2. So, there was no kernel-ml 4.2. Fortunately, the patch was added to kernel 4.3-rc1 and ,therefore, kernel-ml 4.3 could be released.
With regard to kernel 4.2, the patch had to be backported. It was one of several hundreds of patches that were waiting to be added by GregKH. It eventually happened with kernel 4.2.4.
The whole thing involved email exchanges, testing proposed patches, etc. So, here we are talking about commitment ...
Akemi
[1] http://elrepo.org/tiki/kernel-ml [2] https://lkml.org/lkml/2015/7/5/74