I think maven should ALWAYS run the unit tests as it does now. Maven should allow other directories to be treated similarly to test but have a specific goal to run them.
cheers bill On Wednesday 22 January 2003 12:40 pm, [EMAIL PROTECTED] wrote: > You are right about some of my unit test's being more oriented to being > Integration tests. But regardless of that, some larger projects that have > a 100's of tests, it still adds up to an appreciable amount of time to run. > Now, from the tone of the emails, I think everyone agree's that they would > run all the tests, including the various slower integration tests before > producing a "deliverable". But often you are not testing a deliverable. > When i change something small in my web app that then has to be deployed as > part of a war, adding in the testing time becomes a pain, and (as Henning > points out) leads to having a larger and larger list of excluded test > cases! > > A tenet of Agile development is to use what works for you, and discard the > rest. While I appreciate that Maven faciliates, and even encourages good > development practices, I don't think it should enforce a set of specific > practices. > > It does sound like a lot of people would like to be able to control the > level of testing. Proposed was a very static "ON/OFF" property. However, > what would be ideal is to declare various set's of tests, and be able to > declare which set to run. That way you could declare an empty set, when > you don't want any tests to run, which is what I think you are proposing as > well. > > Having said that, I think the ON/OFF is a quick solution for the short term > until the more sophisticated solution is found. > > Eric > > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]] > Sent: Tuesday, January 21, 2003 10:31 PM > To: Turbine Maven Users List > Subject: Re: Mechanism to skip tests > > [EMAIL PROTECTED] writes: > > I posted an email a while ago about that as well.. I would love to be > > able > > > to call maven java:jar -ignore test:test etc.... For the same reasons > > you have specified. Purists have said that you should ALWAYS run tests, > > but when they slow you down too much, they are just ignored. I think > > being > > able > > > to selectively turn them on and off would be great. > > > > Eric > > I may be a *purist* since I do believe that before a deliverable is > built you should have all the tests passing. Why are your test so > slow? Maybe they are not unit tests? I propose that maven ( or a > plug-in ) handle another ( maybe even more than one ) level of testing > instead of a toggle switch for unit tests. > > The unit tests ( fast, independent ) would be hitched into java:jar as > they currently are. But some tests are more *integration* type tests > ( slower with external dependencies ) that may be *not* be required to > run on each build. Currently my criteria for these types of tests are > speed and dependencies. Tests that require the database, other projects > deliverables in a reactor type environment, or sometimes even tests that > use the file system for data or processing are some examples. > > > -----Original Message----- From: Colin Sampaleanu > > [mailto:[EMAIL PROTECTED]] Sent: Tuesday, January 21, 2003 2:31 PM > > To: Turbine Maven Users List Subject: Mechanism to skip tests > > > > > > Is there an easy mechanism which I have missed to allow tests to be > > skipped when performing a target (such as java:jar) which has a > > dependency on test:test? I've looked at the source for the test and java > > plugins and can't see anything. > > > > Generally you do want to run tests when building a jar (or some artifact > > farther down the dependency chain). But it is also a pretty common > > occurrence that you want to do that target without executing tests, e.g. > > when you have just changed a property file, etc., and you know it will > > not affect tests. I realize that 'maven.test.failure.ignore' is > > available, but that just skips failing if the tests don't run, I am > > looking for something which allows skipping the running of tests. > > > > Generally, I think you should be able to set whether tests are compiled > > by default, and allow that value to be toggled easily at runtime, and > > then a similar mechanism to set whether tests are executed by default, > > and allow that value to be toggled easily at runtime. This should also > > work for the reactor. > > > > Does anybody agree or have comments? -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
