[webkit-dev] Questions of a new committer

2010-06-25 Thread Tony Gentilcore
Is there any policy or guideline regarding when it is appropriate to use the
commit queue vs landing directly?

I feel like there is an unfortunate positive feedback loop right now:
1. Commit queue gets slightly backed up either due to a breakage or just
heavy volume.
2. Because the queue is backed up, it is more tempting to land directly.
3. Directly landed patches are more likely to break the tree and the new
breakage backs up the queue even more.
4. Goto #2.

Long queues punish developers not because of the wait for the patch to land,
but because those patches are more likely to have merge conflicts. So the
incentive seems to be to land without the queue. I'm just curious if this
has been brought up before and if folks more experienced than myself have
ideas about how we could improve this.

Tony
___
webkit-dev mailing list
webkit-dev@lists.webkit.org
http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev


Re: [webkit-dev] Questions of a new committer

2010-06-25 Thread Eric Seidel
Completely agreed!  That's why Adam and I are trying an experiment of
removing the wait for all bots to be green check from the
commit-queue:
http://trac.webkit.org/changeset/61831

So far so good.  It's probably too extreme to remove the check
entirely, but we'll know better after a few days.

The queue was over 50 entries last night, but it's down to 25 now and
should be 0 by tomorrow.  With the green check gone, time from cq+ to
landing should only be how long it takes to build and test your patch.

-eric

On Fri, Jun 25, 2010 at 11:49 AM, Tony Gentilcore to...@chromium.org wrote:
 Is there any policy or guideline regarding when it is appropriate to use the
 commit queue vs landing directly?
 I feel like there is an unfortunate positive feedback loop right now:
 1. Commit queue gets slightly backed up either due to a breakage or just
 heavy volume.
 2. Because the queue is backed up, it is more tempting to land directly.
 3. Directly landed patches are more likely to break the tree and the new
 breakage backs up the queue even more.
 4. Goto #2.
 Long queues punish developers not because of the wait for the patch to land,
 but because those patches are more likely to have merge conflicts. So the
 incentive seems to be to land without the queue. I'm just curious if this
 has been brought up before and if folks more experienced than myself have
 ideas about how we could improve this.
 Tony
 ___
 webkit-dev mailing list
 webkit-dev@lists.webkit.org
 http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev


___
webkit-dev mailing list
webkit-dev@lists.webkit.org
http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev