Ben can you write up something sane and we can review the proposal in tomorrows IRC meeting? Either a Jira as you propose; or a policy change for the developers guide if you want all supported modules to behave in the same way.
Cory Horner started looking at this issue before he left Refractions; perhaps you can pick up where he left off? My understanding is that the first connection failure should log the exception? Jody Ben Caradoc-Davies wrote: > I am writing some unit tests that use OnlineTestCase, the online test > fixture infrastructure. OnlineTestCase provides two services: > (1) configuration of online test fixtures using files in the user home > directory, and > (2) silencing of test fixtures whose connect() method fails. > > Service (1) is good. I am happy with (1). Service (2) is, in my view, a > bug, not a feature. > > The current behaviour of OnlineTestCase is to call connect(), and if an > exception is thrown, swallow it, print a stack trace, and disable the > test. This causes the test to pass when connect() fails. This is a > dangerous practice because it causes unit test failures to be silent > (stack traces do not count). > > The default maven configuration disables online tests. If online tests > are being run, it typically means that the developer has explicitly > turned them on using both a command-line flag and by providing a fixture > properties file. It was surprising to me when my online tests silently > passed when my connect() was failing. > > Even more frightening was the realisation that, if I broke my code, > OnlineTestCase would "helpfully" disable the unit test to prevent me > from being troubled by a pesky unit test failure. In my view, this > defeats the purpose of a unit test. If I turn on an optional test and go > to the trouble of providing a fixture properties file, and something is > broken, I want to know about it. > > I propose to change the behaviour of OnlineTestCase so that exceptions > during connect() or disconnect() are not swallowed. Yes, this will > inconvenience some people who expect tests to pass when their network > resources go away, but ignoring unit test failures is very dangerous > practice. I know I do not have to use connect(), and could just override > setUp(), but this seems to be a misuse of the class. > > If there is any support for my proposed change, I will make a patch and > put it in Jira. > > ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ Geotools-devel mailing list Geotools-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/geotools-devel
