All Zn tests that run local servers work like that, using a random port from a range of 32. With fixed ports, we would get many more conflicts because tests and/or CI jobs are running concurrently on slaves. We could increase the range. We could pass some small unique ID to each image and use that to compute a unique port. Testing if a port is available is a total mess because the different socket plugins behave slightly different. Maybe starting servers and letting the OS give out a next available port could work, I should try that. In any case, unless this is a persistent problem I would not do it now.
On 05 Mar 2015, at 10:29, Nicolai Hess <nicolaihess@web.de> wrote:
Anyone has an idea why this test fails so often?
In ZnImageExampleDelegateTests>>withServerDo: block we use ZnServer on: 1700 + 32 atRandom.
I don't know if this is related to the validation failure, but the randomly generated portnumber makes it hard get reliable results.
nicolai