Stephan, I think you refer to the tests that fail when run in a headless image? This is often because they test behavior that assumes an interactive UI Manager (i.e. which prompts the user). The solution lies in making the tests work with the NonInteractiveUIManager. Either adapting the test, or adapting the UI Manager. On 20 Apr 2011, at 07:58, Stephan Eggermont wrote:

Is the Hudson setup used to generate the Pharo images published? Last friday there were some issues that were not reproducible with the standard test-runner.

Stephan