Run the regression tests before each commit |
 |
 |
 |
| Written by Graham Stoney | |||
|
You should aim to run your full regression test suite before every commit to the trunk, because you don't want to break the trunk. For this reason, your regression test suite needs to be efficient, and quick to run. In a large system, this may be too time-consuming, especially considering that you should be doing a commit for every single logical change, no matter how minor. If you are very confident that the change will have limited effect on other modules, you may get away with running a subset of the full suite; at the very least, you must run the unit test for the module in question. But be careful because often changes to one module have unexpected side-effects on others and catching this class of bug is part of the reason the regression tests exist. The best approach is to kick off a full regression test run, and start working on the next change that you want to make in a different workspace. Of course this only works if the next change isn't dependent on the one you're currently testing. Social Bookmark this page:         
|