On 6/5/06, Brett Porter <[EMAIL PROTECTED]> wrote:
On 06/06/06, Craig McClanahan <[EMAIL PROTECTED]> wrote: > Maven2 needs to support integration testing as a first class notion in the > architecture of what you envision a "project" to be. It may not have been clear enough, but that's exactly what I meant in my last email. I thought this is what the wiki page discussed - if you think something is missing there, let me know. Somehow I missed the correct issue I was thinking of, MNG-1381, so I've linked that now instead. I'm just trying to consolidate this. I can assure you it is the issue you are thinking of - sorry for the confusion, re-reading I see MNG-591 is a specific use case of integration testing and not very helpful. > It's not just webapps > .. you've got the same sort of issue with EJBs, or web services, or anything > that gets deployed in a container. Unit tests just don't give you the > confidence you need that the application will actually work. I've seen too > many cases where all the unit tests on a webapp all pass with flying colors, > but it throws an HTTP 500 on the welcome page because of a stupid coding > error in the JSP page that wasnt' tested with the unit tests. Yes, I'm well aware of that. > What's needed is a complete additional test environment, with its own > lifecycle, and its own classpath (i.e. dependencies tagged to this scope so > you only load things like HttpUnit or HtmlUnit here). If integration tests > exist, they should be part of the default "mvn install" processing, just > like unit tests are, unless it is explicitly disabled. Don't pretend that > there is only one kind of "test"!!! Again, what I was getting at. This was discussed at length on the Maven dev list and is summarised on the wiki page. > > Otherwise, you guys are not being serious about trying to encouraging best > practices in build environments :-(. > Of course we are.
Added some comments on the wiki page. - Brett Craig
