On Mon, 26 Jul 2010, Rajith Attapattu wrote:
My own view is that on checkin we should need only to be running
*unit* tests (which are in rather short supply to be fair). We should
have a CI environment where *system* / *integration* tests are being
run constantly with all possible profiles being tested. IMHO unit
tests for the client should not be requiring a broker of any kind -
broker functionality would be mocked up as part of the testing
suite.... however there'd need to be a fair amount of work to get to
that point.
The highest priority for me would be looking to see if we can get a
shared CI environment running on Apache which reported failures to the
list.
-- Rob
I totally agree with you here. It would be nice if we had a publicly
accessible CI environment.
As you are aware at Red Hat we do have a CI environment.
All though I am not sure if the env itself can be opened up to
outside, at least we maybe be able to email if a build breaks.
I can follow up with Gordon to see if we could email the results to
the dev-list. I think in the short term that should help.
Rajith
Hi, I maintain the internal CI system. Rajith is correct. We could
easily send mails from our system. At present, however, we don't have a
good way to make the full build and test output available. That's
something I hope to address in the coming months. In the mean time, we
can manually provide the details if the mail message isn't enough.
I view solving this problem as one part of three related tasks:
- Refactor the qpid build process so that we're all building and
testing the same set of code
- Spin up a standalone interop suite
- Automate extended testing (as Rob and Rajith suggest)
Justin
---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project: http://qpid.apache.org
Use/Interact: mailto:[email protected]
