I've updated the patch to apply to the current tip. (This patch was an opportunity for me to update to an Hg workflow.)

Alexander, I disagree with you about the tests. The unittests use the exact same pattern/model that testIteration uses. I find your complaint that a unittest, which does test the feature under question, is not good enough, despite the prevailing unittests being designed in the same manner, to be absurd. Practicality beats purity -- a test that works is better than no test at all.

By rejecting unittests on the merits of its coding style, you are creating a double-standard for people like me (outside of the core committers), which eventually wears out my interest in helping you get this improvement into your project. I have been chased around this obstacle course before. For example, issue5949, when I was asked to provide unittests for a module that had *none*, for a 3-line patch that nobody disagreed it being correct. I had a vested interest in jumping through the obstacles of getting that patch in, but here I am again being blocked from making a 3-line patch, except this time, purely for stylistic reasons.