Hi Branden,

i think you are getting carried away quite a bit.

FWIW, i don't think the concept of "code coverage" has much value
in the first place.  Yes, i have been forced by management to
consider "code coverage reports" in previous jobs in the industry.
Yes, it is true that, if not a single test in a suite exercises
a certain line in the code, that test suite tells you nothing about
the presence (or later introduction) of bugs involving that
particular line.

But even if a test suite exercises a particular line multiple times,
that line may still contain bugs, and future changes may introduce
bugs into that line without the test suite noticing.

So coverage says pretty little about the quality of a test suite.

There are criteria that are way more important for the quality
of automatic testing than coverage - for example, simplicity and
clarity of the design of the test cases, readability of the test
code, logical organization (such that the test suite can be
maintained, reviewed, and extended by humans).  It is also way
more important to test complex situations that matter in practice,
and to logically structure the suite by features (as opposed to by
code lines) than hitting any particular line in the code.

All that said - not much harm done by wasting a bit of time on
colourful HTML generation if you want to please your inner child by
playing around a bit, or if management wants to produce some numbers
that can be put into reports that will impress even higher management
(as opposed to fostering actual improvement).

But don't deceive yourself, as if such traffic light coloration
would bring about enlightenment...

Yours,
  Ingo

  • Code coverage r... G. Branden Robinson
    • Re: Code c... Ingo Schwarze
    • Re: Code c... Simon Josefsson via GNU roff typesetting system discussion
      • Re: Co... G. Branden Robinson
        • Re... Bruno Haible via GNU roff typesetting system discussion
          • ... G. Branden Robinson

Reply via email to