[webkit-dev] Criteria for becoming a "core" builder
aroben at apple.com
Wed Mar 2 11:30:33 PST 2011
On Mar 2, 2011, at 1:38 PM, Eric Seidel wrote:
> So if you believe the windows bots are regularly green, you should
> make them Core.
Well, they are regularly green except for the cases I mentioned. The biggest problem is people making changes that cause tests to fail on Windows and not noticing that this has happened, so I think making them core will help keep them quite a bit greener.
So it sounds to me like we should go ahead and do this. I'll wait a day or two to allow other opinions to be heard.
> Along the same logic, we should move bots like Leopard to non-core,
> since they haven't been green in weeks (even though I'm sure we wish
> it would be green).
The Leopard Debug bots are non-core now. The Release ones are still core for the moment.
> On Wed, Mar 2, 2011 at 9:51 AM, Adam Roben <aroben at apple.com> wrote:
>> Hi all-
>> I'd like to understand better what the criteria are for a builder to become a "core" builder.
>> We've been putting lot of effort into keeping the Windows testers green lately (especially the Windows 7 Release (Tests) builder), and it seems like it would be good for them to eventually join the "core" group so they can reap the benefits of sheriffbot, an increased number of people looking at them, etc. Most of the failures that happen on the Windows bots these days are either due to:
>> 1) a new test being added which needs results on Windows
>> 2) an existing test with Windows-specific results being modified without the Windows results being updated
>> 3) test flakiness
>> Being in the "core" group seems like it would help with (1) and (2) because the people who are making these changes would notice that they forgot Windows. It wouldn't help with (3), of course. (We're obviously trying to track down all causes of flakiness we can.)
>> So what does it take for a builder to be a "core" builder? Once we know this, we can hopefully focus our Windows efforts toward satisfying the criteria.
>> webkit-dev mailing list
>> webkit-dev at lists.webkit.org
More information about the webkit-dev