[webkit-dev] The Purpose of Core Builders

[Patrick Gansterer]

Am 09.11.2011 um 17:21 schrieb Adam Barth:

> On Wed, Nov 9, 2011 at 2:47 AM, Patrick Gansterer <paroga at paroga.com> wrote:
>> Some of my thought on this, since I have the same "problems" with my WinCE bot:
>> 
>> 1) Is there any real benefit in the CoreBuilders concept (since it's not used as intended as Eric already wrote)?
> 
> If it were up to me, I'd remove the concept because causes me more
> headaches than it's worth.  However, ggaren makes some good points
> about why it's useful.  Hopefully having that wiki page will help.

Sorry, but I don't see the benefit compared to the "is 90% green" solution.

>> 2) When we still want the CoreBuilder concept: Can we make the core/non-core transition without any SVN change? What about improving the buildbot-master, so that it dertiminates the core/non-core on the fly, by checking the last x builds? I don't like the idea of adding and removing the the bots to the "core list" all the time.
> 
> I'd be happy to review such a change.  Practically, however, if
> there's a configuration that a large number of developers use (e.g.,
> apple-mac), it's probably worth having on the front page even if it's
> red a lot.

An alternative might be to split the bots into categories like "apple", "chrome", "qt" and so on instead of "core"/"non-core". But this won't help in catching problems on the "other" bots. So not sure if that is a practicable solution.

>> 3) It's hard for non full time contributes to get informed about broken bots, if you don't watch the buildbot site the whole day.
>>   a) Maybe we can tell the buildbot to write a mail to the build slave maintainer, when the bot is broken. (It happened more than once that my bot was red for days, because I didn't get informed about it).
>>   b) I agree that keeping all bots green is not the task of the committer, but maybe someone should trigger the port maintainer to fix it. I have no idea about a general working rule, but as of my experience the committer simply "don't see" the problem. E.g. WinCE bot builds the interpreter version of JSC, which is a "core feature" and not port specific, but the commit author didn't see the break.
> 
> It's petty easy to write a webkit-patch command that monitors the bots
> and files bugs when they fail or whatever.  If I were you, I'd just
> write such a monitoring command rather than wait for other folks to
> solve this problem for me.

I don't want other folks to do "my work", I'm not happy, but ok with the current state. I wanted to add my view to this topic. :-) Maybe other folks have similar requirements, so that we find a more general solution, instead of many home brew ones.

- Patrick