On 05/24/2012 04:26 AM, Koen Kooi wrote:
Op 24 mei 2012, om 00:00 heeft Scott Garman het volgende geschreven:
Hello all,
As most of you know, I've been pulling commits into a pair of
sgarman/denzil-next branches which are intended to eventually
become the next Denzil point-release, 1.2.1.
oe-core based branch:
http://git.openembedded.org/openembedded-core-contrib/log/?h=sgarman/denzil-next
Would it be possible to do that in the main repo? That saves people
from needing to add and fetch yet another repo.
I can't start with the main repo, because I don't have commit access to
it, and until I have done some significant testing with my commits, I
want to preserve the ability to rebase my branch.
However, now that I have some autobuilder resources to do more extensive
testing, what I'm proposing is to get my *tested* commits moved from
sgarman/denzil-next into the denzil branch after review from the
community and a merge that Richard will do.
That means the denzil branch will lag my sgarman/denzil-next branch by a
certain amount of time, which I intend to keep down to less than a week.
So now to address issue #2. The goal is to incorporate community
feedback, so I'm looking to get ACKs or NAKs for these commits
before they go into the main denzil branch.
My proposal is to send denzil pull requests to the appropriate
mailing lists, and Richard can merge them into the main repo denzil
branches once they've received review by the community.
My goal is to send these pull requests about once per week, once
I've managed to get a green build out of the Yocto autobuilder for
my contrib branch.
What do folks think about this? Now's your opportunity to offer
feedback and influence this process.
You're missing a process for people to propose commits for the denzil
branch. How about this:
If you want to propose a commit from master cherry-pick it into a
local denzil branch and then send if for review to the oe-core list
with the [for-denzil] subject prefix. Scott will take care of putting
it in the pull requests based on the feedback the patch gets.
I can work with that. But keep in mind that if the request is not
associated with a current bug marked for 1.2.1 in the Yocto bugzilla, I
will need to see strong community support for including it (aka,
multiple ACKs). This is needed to avoid scope creep in the point-release.
What would be nice to have is a (soft) requirement to add the testing
data in the commit e.g. "Runtime tested on beaglebone/angstrom". Not
sure how to automate it, but pull-requests should include a
buidhistory diff when possible.
The more testing that's done with a pull request for 1.2.1, the more it
can be proven that the change is low risk, and the higher chances it
will have of making it to the final denzil branch, for certain.
Scott
--
Scott Garman
Embedded Linux Engineer - Yocto Project
Intel Open Source Technology Center
_______________________________________________
Openembedded-core mailing list
Openembedded-core@lists.openembedded.org
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core
