ResumableTestFailure

Hi all! :-)

I was just working a bit with Squeak by Example (the book) and found out that resumable assertions as described in the book (see chapter SUnit, Advanced, Continuing after a failure) do not any longer work in a trunk image. If you write a test case with some resumable assertions, turn on logging, and run the test, only the first failure will be logged, not the rest, too.

(Methods to reproduce:

isLogging

    ^  true "You will need to open a Transcript"

testResumable

    #(1 2 3 4 5) do: [:i |

        self assert: i even description: ['{1} is not even' format: {i}] resumable: true]

)

A second issue I experienced frequently when I felt the desire to work with resumable assertions was that they are basically only available for two selectors: #assert:description:resumable: and #deny:description:identical:. I often would like to insert a resumable into any test without being that verbose, and nearly redoubling the number of assertion selectors by adding a resumable argument to each of them did not feel right, either.

Please find the attached changeset, which fixes both issues.

• In TestResult >> #runCase:, resumable assertions are identified and - finally - resumed, after marking the test as failed. Unfortunately, other test runners (yes, I'm thinking of smalltalkCI) will need to reimplement this feature unless they reuse the TestResult implementation.
• In TestCase, I refactored some methods that raise assertion failures and added the most basic #signalFailure:resumable: as the single source of signalization. Instead of setting the resumable bit manually for each assertion, you now can also make the whole test execution resumable by:
  • overriding #isResumable,
  • setting a resumable flag using #resumable:,
    
  • setting a resumable flag temporarily using #beResumable:during:, or
    
  • adding a <resumable> or <resumable: true> pragma to the test method.
    
• In addition, I tested the added and fixed behavior in ResumableTestFailureTestCase.

Looking forward to your review! :-)

Best,

Christoph

PS: I ran all tests in a fresh trunk image before and after loading the changeset, there were no regressions.