I just began to eat my own dog food, applying rcov to its own
regression tests. The coverage figure is still pathetic, but at least it's
going up: rcov had been essentially untested (or only tested manually, which
amounts to about the same) for way too long time. The main reason was bad
laziness, and the self-justification that it's not easy to test code that
relies on set_trace_func. Which was of course to a large extent false.
So here I am, adding regression tests semi-automatically with my
xmp filter:
I now feel much more confident hacking rcov, and I finally got around to
refactoring a bit. 0.4.0 is surely going to be quite cleaner and more
featureful that 0.3.0.