Still, we've got all but a couple of the tests working again, usually by changing the tests to reflect changed functionality, but often by fixing the bugs that the tests should have been making plain.
Now I'd like to see what our test coverage is; I'm sure that it's far below 100%, and I'd like to start moving it in the right direction. Has anyone used JXCL and UCovered? Any good? Or is there something else I should be looking at?
Aside from "run your bloody unit tests", there's one other lesson learned; some of the code breakage resulted from changes to our database definitions, which for some reason were never kept under source control, so it's not possible to establish who made a change, when, and most importantly, why. The database change might have been made for a good reason, so you just can't remove it - especially if your unit tests are spotty.
I gather that keeping your database definitions under source control isn't trivial when you are using SQL Server, but it must be possible. We'll do it next time; if I can't manage to convince the team to move away from SQL Server, we'll just have to work it out.Posted to Java by Simon Brunning at July 12, 2004 01:42 PM