[racket-dev] A standard for testing Racket's libraries?

On Wed, Feb 27, 2013 at 7:01 AM, Jay McCarthy <jay.mccarthy at gmail.com> wrote:
> The tests are run by running "main.rkt", which has no rerrors when you
> run it. It is a bit strange that those errors happen when run that
> way, but I think it just means that "eval" should control the context
> of the sub-programs more.


Ok.

Is there anything that can be done to regularize how all the
collections in Racket are to be tested?  It doesn't have to be as
strict as forcing use of rackunit or eli-tester: just having a
convention like "Run the test-all.rkt at the toplevel of a collection
to run all its tests" would be a major improvement.

At the moment, each collection follows its own rules for testing it.
The non-uniform structure of the testing infrastructure makes it
difficult to develop and maintain Racket's libraries.   I can make a
bug fix, but without a fixed point of entry to test the changes I've
made, I get nervous: maybe I'm running all the tests for this
collection, but maybe I've missed something.

I had assumed that all of Racket's automated test suite would be
driven by collects/tests/run-automated-tests.rkt, but it only runs
some of the core Racket tests: it doesn't hit the entire collects
tree.  In absence of that, I thought raco test was the way to go, but
that doesn't currently work either.