I meant to say that projects that use pharo should rethink the tests and rules they run in order to ensure that they do not break the Pharo constraints.

For example, for Moose, we should have, like Stephan suggested, an integration job that runs all tests in the image (at least). Ideally, it should also run other rules that ensure that the system does not break due to semantic mismatches.

If we do it this way, we push the signals upstream and discover the problems sooner.

Cheers,

Doru