"Fishy" Misinformation Was Re: Concerns about GCI

Jonathan Leto jaleto at gmail.com
Mon Jan 3 16:24:53 UTC 2011


On Sun, Jan 2, 2011 at 9:44 PM, Jason Garrett-Glaser
<darkshikari at gmail.com> wrote:
> Now if you want something fishy, go look at the parrot tasks.
> According to my students, not only are their "test coverage" tasks
> trivial, but they don't even require code review!

Sorry to inform you, but "your students" are incorrect, as are you.

> rfw has been using
> these to spam dozens of points a day to get a ridiculous lead, despite
> the fact that the tasks require no maintenance (ours do) and require
> no code review (ours do).

What now? You are wrong on both accounts. I have no clue what you mean
by "no maintenance" and you are of course very wrong when you say they
require no code review.

I am not sure how your open source project deals with code review, but
Parrot Virtual Machine does a strict code review before merging code.

I code review almost every task, and if I don't, another core Parrot
developer does.

If you go to


and click on the "Closed" button, you will see almost 20 closed pull
requests from rfw. This means that 20 times, a parrot core developer
has gone to the pull request, pulled it into a branch for further
testing or merged directly to master, after looking at the code and
running the tests, then closed the pull request.

> And people complain about *our* tasks being too easy?  This is
> extremely silly.  Currently a dedicated student can get 50 points in a
> day without even having his code looked at by another human.

You are taking rfw as your single datapoint and extrapolating from
there? As any first-year statistics student knows, linear
interpolation from a single datapoint is just lies.

rfw is an exceptional student who plows through difficult tasks. He
sits in the Parrot IRC channel all day, asking as many questions to as
many mentors as possible. He has this contest "figured out", where
most students are still trying to figure out the rules.

Every code coverage task requires reading code coverage output, which
is in C, then writing code in either PIR (a language students have to
learn first) or Perl 5. Each code coverage task requires installing
Perl 5 modules such as Devel::Cover, which then requires students
understand CPAN or local::lib and other intricacies. The written tests
then need to be run and pass, and then the student must generate their
own local coverage report so see if they raised the test coverage
enough to complete the task.

Jason, please refrain from spreading misinformation.


PS: I have cc'ed this to parrot-dev in case any Parrot mentors would
like to comment.

Jonathan "Duke" Leto
jonathan at leto.net

More information about the parrot-dev mailing list