Kathey Marsden wrote:
On 11/12/2010 6:13 AM, Rick Hillegas wrote:
Just a gentle reminder to please get your 10.7 changes in by 5:00 am Monday morning San Francisco time. Soon after that time I will cut the 10.7 branch, regenerate the release notes to sweep up last minute bug fixes, and then build the release candidate.

If you need more time to finish up your 10.7 work, please let me know before then.

Hi Rick,

The schedule says Buddy testing is: 2010-11-01 - 2010-11-15 and then release candidate 2010-11-15 Although I think it is important to test features before the release candidate is cut, doesn't buddy testing normally happen against the actual bits and documentation we plan to release?
Hi Kathey,

Right, we have done it that way in the past. My thinking on this topic has evolved a bit, though. Nowadays I think of buddy-testing as a sort of integration test which we should perform before producing the release candidate. I lump it together with the sanity-checking of the release notes. Here's how I map the Apache process onto traditional closed-source terms these days:

Buddy-testing = Integration test

Release candidate vetting = Alpha test

First release from a new branch = Beta test

Second release from a branch = Production-quality release

Five years ago when I first joined the Derby community, I thought of the release vetting as a Beta test. But I was wrong. The Apache process requires us to share the release candidate only with people on the developer list, not with our broader user community. We don't get real Beta testing by real users until we actually publish the release on the download page. The first brave users who end up running a Derby feature release end up being our beta-testers. Working backward from that fact makes the candidate vetting an Alpha test at best.

The value which we get out of buddy-testing seems to me to be what people expect from an integration test in closed-source companies. Up until buddy testing, it seems to me that we merely unit-test fixes and features. We don't perform any systems or integration-testing before then. So buddy-testing is the closest fit which I can find to the traditional concept of an integration test. In the closed-source environments where I have worked, the integration test precedes the alpha test, and the alpha test happens either on the release candidate or even before that.
I plan to do some testing on DERBY-728 today but will continue next week with the release candidate.
All testing appreciated whenever it happens.

Thanks,
-Rick

Thanks

Kathey




Reply via email to