Matt Benson wrote:
The AntUnit documentation says:

"Each test target is run in a fresh Ant project; i.e.
each test target has a fresh set of properties and
references."  But if I have test targets:

  <target name="testFoo">
    <property name="foo" value="foo" />
    <au:assertTrue message="$${foo}=${foo}">
      <equals arg1="${foo}" arg2="foo" />
    </au:assertTrue>
  </target>

  <target name="testBar">
    <property name="foo" value="bar" />
    <au:assertTrue message="$${foo}=${foo}">
      <equals arg1="${foo}" arg2="bar" />
    </au:assertTrue>
  </target>

one will fail, as I would expect given that the
AntUnit code uses a single Project instance to run all
targets.

The fun part: if we change the behavior to suit the
documentation, the utility of beforeTests/afterTests
is pretty much reduced to filesystem artifacts.  I
would choose to correct the documentation to describe
the actual behavior, and see if we can't get scoped
properties working (minus the memory leaks) for Ant
1.7.1-1.8 , to alleviate the discomfort of having to
choose new properties for every test (see the apply
testcase in core).

Thoughts?

Every test run should be in its own project.

if you want setup before a test, you can depend on setup targets, and we already do a teardown at the end of each run

If you want before/after targets, why not modify <au:antunit> to let you declare the name of the before/after targets to run?



---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to