On Sun, Jan 05, 2020 at 08:34:59AM -0800, Guenter Roeck wrote:
I think we are going into the absolutely wrong direction. Expecting that
everyone would use mainline is absolutely unrealistic, and backporting
more and more patches to stable branches can only result in destabilizing
stable branches, which in turn will drive people away from it (and _not_
to use mainline). The only reason it wasn't a disaster yet is that we have
better testing now. But we offset that better testing with backporting
more patches, which has the opposite effect. One stabilizes, the other
destabilizes. The end result is the same. Actually, it is worse - the
indiscriminate backporting not only causes unnecessary regressions,
it (again) gives people an argument against merging stable releases.
And, this time I have to admit they are right.
Just to get an idea of how the AUTOSEL commits affect the kernel tree I
tried the following:
We have 10648 commits on top of 4.19 in the 4.19 LTS tree:
$ git log --oneline v4.19..stable/linux-4.19.y | wc -l
10468
I've tried to identify how many of them have a "Fixes:" tag pointing to
them, and how many were reverted (using it to identify buggy commits in
the stable tree), ending up with:
$ wc -l fixes
637 fixes
So about 6% of the commits that go in the stable tree have a follow up
fix or revert. Now, let's see where commits in the 4.19 LTS tree come
from:
$ git log --oneline --grep "Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>" v4.19..stable/linux-4.19.y | wc -l
5475
$ git log --invert-grep --oneline --grep "Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>" v4.19..stable/linux-4.19.y | wc -l
4993
In general, Greg is the one who picks up commits that are tagged for
stable, security issues, and patches requested by folks on the mailing
list. I'm mostly doing AUTOSEL, other distro trees, and mailing list
(but Greg still does the most of the mailing list work).
Anyway, looks like mostly an equal split between stable tagged commits
and AUTOSEL ones.
Now, looking at the buggy commits again, I check whether the commit came
via Greg or myself (just using it as a way to differentiate between
stable tagged commits/commits requested by users/etc and commits that
came in using AUTOSEL), and I get:
$ for i in $(cat fixes | awk {'print $2'}); do if [ $(git show $i | grep 'Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>' | wc -l) -gt 0 ]; then echo "Sasha"; else echo "Greg"; fi; done | sort | uniq -c
367 Greg
270 Sasha
Which seems to show that AUTOSEL commits are actually less buggy than
stable tagged commits. Sure, this analysis isn't perfect, but if we
treat it purely as ballpark figures I think that it's enough to show
that picking up more fixes doesn't contribute to "destabilizing" the
stable trees.
--
Thanks,
Sasha
