I'll try to be as succinct yet informative as possible.
We've written some integration tests for our application that are asserting that the contents of a select > option drop down within an RJS response contain values that we expect. Each option > value is dynamically constructed (via links within our database), the only way to determine what the option > value should be is to generate it on the fly.
When running our test suite I'm seeing that 'rake test' results in the test passing as expected - the response rendered within the test case is what I'm expecting, something like:
(snippet)
In this case, we have an AR model representing a java class, SliceArrivalTime, whose calculated path is itinerary/slice/arrival@time. The path is *generated dynamically*, so it's as if our view file is, in one case, calling instance.path to render the response and calling instance.java_class.prettify (truncates com.foo.bar.Baz to Baz) in another.
This is the same test case using the same database being executed using a different task in the Rails rake task hierarchy. From everything I can poke at it looks as if the same controller/action/biz logic flow is being followed each time, the only difference is in what gets generated in the tests.
I know there probably isn't enough information here to get anyone started with debugging this, I'll be watching email and be in IRC (capndiesel) if anyone wants to chat about it.
