| Subject: | Tests may fail (on parallel testing?) |
http://matrix.cpantesters.org/?dist=Redis-NaiveBayes%200.0.4 shows a number of fail reports, mostly generated by Andreas Koenig's smoker. This smoker is known to test multiple perl versions in parallel. In fact, I could also reproduce a fail when running the test suite simultanously for two perl versions.
In t/11untrain.t I read the notice "# NOTE that this test uses the state set by t/10train.t". This probably also fails if the test suite is run with "prove -j" or "prove --shuffle". To fix these problems, maybe two things could be done:
- combine the 10train.t and 11untrain.t tests
- use a random namespace to avoid collisions with other running perl test instances