I don't want to debate our architecture or processes since I have little to do with defining them, but its an N-tiered web application comprising from one to dozens of servers. Many of the tests are Selenium Integration tests on Windows clients hitting remote servers that are mostly Linux. We have dev and test instances that we hit during development. One of the key properties that changes is the baseURL used by selenium, as well as debugging levels.
The integration test code is not currently part of the source for the application. The tests never run on the app servers directly (I've never seen anyone do that anyway...our build servers don't run the web app.), particularly the integration tests as they are client-based. We are running the tests from dev workstations for now but will eventually expand to running a subset of integration tests via Hudson. We are leveraging as much of Maven through the entire process as possible in order to maximize reuse, portability and integration moving forward. Anything else I can clarify? On Jan 20, 2011 2:24 PM, "Ron Wheeler" <[email protected]> wrote:
