> From: Stefan Bodewig [mailto:[EMAIL PROTECTED] > > I'm not sure whether I've run the tests on my Mac after > testGetExtensionPoints has been added and it just fails now after I > upgraded Java or whether it has always been failing. > > The test depends on the implementation of reflection - in particular > on the order returned by getMethods. > > We had three valid adder methods: > > add(Number) > addConfigured(Map) > addConfigured(Hashtable) > > and the main point the test wants to assert is that the Hashtable arg > version comes before the Map arg version in the list returned by > getExtensionPoints. This works fine. But the test also assumes that > add(Number) comes first, while the relative order in reality is > undefined. There are three valid orders getExtensionPoints may > produce: [...] > Instead of trying to find which of the three results we are looking > at, I simply removed add(Number) completely from the test - feel free > to improve it 8-)
Sorry I broke MacOS X Stefan ;-) I'll see about find a testing logic independent of the order. BTW, is there some explicit resolution rules whether add() has higher priority than addConfigured(), or vice versa? What if I have add(HashMap) and addConfigured(Map)? Who's suppose to be first (win), the first one or the second? Intuitively I'd put an addConfigured() ahead of a add(), but a HashMap ahead of a Map, so who yields to whom in this case? This also brings up the fact that there are no nightly builds of Ant on MasOS X I guess... Is Ant supposed to rely exclusively on the few Gump instances to sample the various OS's it's supposed to run on? --DD --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
