Dragan Simic <dsimic@xxxxxxxxxxx> writes: > Well, I actually wasn't lazy, but instead I wanted to give this patch > better chances to be accepted, by introducing as a small amount of > changes to the code as possible. This was a lesson not to take that > route, but to do "the right thing" instead. The lesson you should be taking is a bit different here, though. As we discussed, doing the separator correctly is harder and doing the terminator correctly is much easier and less error prone. So if we chose to do terminator semantics and punt on doing separator semantics, the right thing could have been to adjust the existing tests to make sure that the new output with unnecessary trailing blank line is expected. Either that, or we should fix the code not to break the current expectation. Doing neither of these is not quite acceptable--- that is the lesson from this episode, I would think. Thanks.
