Benjamin Shine wrote:
>
> On Aug 28, 2006, at 9:03 AM, Jim Grandy wrote:
>
>>
>> On Aug 27, 2006, at 11:03 PM, Mark Davis wrote:
>>
>>> Benjamin Shine wrote:
>>>
>>>>
>>>> ...which brings up the idea of only doing most of the build on linux,
>>>> and pushing java build products onto the windows and mac builder,
>>>> which I should have done a long time ago, but didn't. So if I'm going
>>>> to rsync the cache, I might as well rsync the jar and docs too.
>>>
>>>
>>> So... One of the benefits of the prefetch on all platforms is that we
>>> know that it works.
>>>
>>> We've gone back and forth on this a number of times and we keep coming
>>> back to the build being our "canary in the mine". I'll argue for
>>> separate builds, but doing it on faster systems.
>>>
>>
>> I agree with this. Our nightly builds are primarily a QA check.
>> Timely delivery of nightlies is important, but not nearly as
>> important as test coverage. Doing builds on all the platforms where
>> we support building is critical.
>
>
> So moving forward, it sounds like the consensus plan for 3.4 is...
> run complete builds and complete prefetches on all supported
> platforms. (Except for doc, which only builds on linux and mac, and
> which we will need to push to windows-builder for inclusion in the
> installer.)
Huh? Does the doc build fail on Windows *by design*?
>
> The ongoing work with planning the next-gen test system includes a
> few levels of testing, from fast/developer/minimal to slow/thorough/
> automated. I don't think we necessarily need faster machines (can you
> believe ben shine just said that?) if we plan our testing system
> well. Even with our split-coast development, nightly builds between
> midnight and 3 am pacific will be done before developers wake up and
> need to see nightly results. Then individual bugs that broke should
> be testable in isolation during the day.
>
> The exception to the "3 hours in the middle of the night is enough"
> rule is developing the build/test systems themselves... which seems
> to be my job, and well, I wouldn't mind if the complete tests ran in
> cup-of-coffee time rather than go-to-the-grocery-store-then-cook-
> dinner time. As far as that goes, the current situation is acceptable.
There should be a subset of the tests that you can use to test the build
system.
~md
--
---
Mark Davis
Sr. Manager, QA & IT -- Laszlo Systems
${this.title} -- OpenLaszlo.org
_______________________________________________
Laszlo-dev mailing list
[email protected]
http://www.openlaszlo.org/mailman/listinfo/laszlo-dev
