----- Original Message ----- From: "Stefan Bodewig" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Monday, March 26, 2001 8:51 AM Subject: Re: [VOTE] abstract goals and guidelines
> >> * no commit of new features without testcases > > > > -1 unless someone tells me how to really provide a testcase for > > something like vssget > > I see this as a problem as well - especially in cases like vssget where > the task is just a layer on exec. > > One can test <tar> be comparing the result of a task with the one a > command line invocation creates, but if you are basically just doing a > command line invocation it is difficult to have an expected result. On could do that with something like vssget as well. The problem is that you need a repository to run the tests (big problem on Unix) and that MS translated VSS to different languages which results in different "results". I don't even have an idea how you would write a testcase for tar (doing more than check some arguments to the task). Where do you get the expected result from? Should it be delivered with ant (and put into CVS)? I will change my vote to +1 if we define for what kind of task a testcase must be provided. And I think we could enforce testcases for the core of Ant. Nico
