On 3 July 2016 at 09:53, Tristan Seligmann <mithrandi@mithrandi.net> wrote:

The tests directories can simply be excluded in coverage.py (or codecov), I don't think there's any need to do something more complicated than that.

That is, don't report coverage at all for the test code?

I would prefer to see the coverage reports for tests, even if we don't enforce 100% coverage. Is a quick way to check that the test is executed on at least one builder.

While I agree that 100% test coverage is an ideal worth aspiring to, I think getting there from the current state is going to be a large amount of work that yields very little benefit at this point in time; I would say that there are more important things to spend that effort on.

It might yield (arguably) little benefit for existing contributors, but I think that for the new contributors it sends a bad message.

For me, as a reviewing it helps a lot if a contributors can get a quick feedback about the quality of their code and I don't have to repeat in each review that a topfile is required of that the code needs tests... the topfile checker or coverage checker is red and this mean that a contributor needs to work more before requesting a review.

We are preaching the Ultimate Quality Development System (UQDS) (tm) but when new contributors are onboarded they found out that they should ignore the tons of errors produced by twistedchecker, that the 100% coverage only applied for implementation code and that test code is a 2nd class citizen ... that the api docs will produce error even if you have created a correct cross reference to zope (similar to what existing code does)... etc

But maybe is just my fault for taking the UQDS too seriously and instead I should read it as the Almost Ultimate Quality Development System.

pyflakes run is not clean on trunk... and I hope that pydoctor will soon run without errors on trunk ... after that we can hit twistedchecker :)

Adi