Re: [PATCH 3.18 00/36] 3.18.94-stable review

[Guenter Roeck]

On Wed, Feb 07, 2018 at 06:37:06AM -0800, Greg Kroah-Hartman wrote:
> On Tue, Feb 06, 2018 at 06:48:17AM -0800, Guenter Roeck wrote:
> > On 02/06/2018 05:14 AM, Greg Kroah-Hartman wrote:
> > > Thanks for letting me know, that's great to hear as I just had a
> > > question from some companies who are worried that taking stable patches
> > > will cause tons of merge issues.  It hasn't in my experience, and seeing
> > > reports of this from others is great news.
> > > 
> > 
> > From merging v4.4 and v4.14 into the respective ChromeOS branches, I would conclude
> > that merging is for the most part easy. Major source of conflicts, if they happen,
> > is that we may already have picked up additional commits from upstream.
> > 
> > This is only true, though, if merges are done on a release-by-release basis
> > and if the merge happens shortly after a stable release is available. Otherwise,
> > as time goes by, merges become more and more difficult (almost exponentially
> > over time). Wait for more than 2-3 months between merges and it becomes almost
> > impossible.
> > 
> > For reference, the top of tree for both branches is
> > 	v4.14.16-3823-g597d36f1d331
> > 	v4.4.114-12977-ga207b53fe939
> 
> Ugh, 12k patches is not trivial, gotta love those graphic drivers :(
> 
That, plus a gazillion backports, plus vendor specific out-of-tree wireless
drivers, plus all the Android stuff. No, there is no "love" involved, only
annoyance :-(.

> > meaning there are _lots_ of patches on top of the mainline stable releases.
> > 
> > Also, overall, the rate of regressions is quite low (if I recall correctly,
> > somewhere between 3 and 5 in chromeos-4.4, and one so far in chromeos-4.14,
> > ie below 0.1%).
> 
> Thanks a lot for the feedback, this helps out a lot when talking to
> companies that are "afraid" of doing merges with stable.
> 
> Based on this, one phone vendor just forward merged their 4.4.y based
> tree to the latest release in a few hours with no reported problems at
> all, so it has already helped out with tangable results.
> 
Excellent. More test coverage!

Guenter