TODO

TODO:

    - Add --start and --end switches to runtests.
    - Clear all known T::H bugs, if possible.
    - T:H to separate distro when we're ready to have do the switch.
    - Make Harness capable of running multiple tests in parallel.
    - Investigate compatibility with Test::TAP::Model (which uses Test::Harness::Straps)
    - Make sure we play nicely with Test::Warn
    - Work out how to test PERL_TEST_HARNESS_DUMP_YAML without jumping on someone who's
      already using it. Maybe if PERL_TAPX_PREFIX is defined prepend the value of that
      to the env variables it reads.
    - YAML dump of test results to file. PERL_TEST_HARNESS_DUMP_YAML=/home/me/somefile.yaml.
    - Implement TAP stream capture to a file. PERL_TEST_HARNESS_DUMP_TAP=/home/me/somedir.
    - Try it with Test::Smoke and see what's needed to run tests in parallel.
      This is a key feature that p5p needs.
    - Question:  what to do if tests hang?  Jan Dubois and Steven Hay are
      looking into this for another project.  They're looking at Win32::Job to
      manage test sessions on Windows.
    - Make TAP::Parser thread-safe!
    - Convert tprove_gtk into TAP::Harness::GTK.
    - Have the aggregator return an iterator.
    - Make sure we get the right harness output with no tests run.
    - Make the 'colored' test output easier to override.
    - More tests, particularly IPC::Open3
    - Run it against Perl's test suite (first prove, then tprove).
    - The 'result' objects are currently blessing hashrefs.  This is bad.
      Instead, they should have proper getter/setters to provide future
      support for subclassing.
    - Horrible idea:  have Mac's voice synthesis tell you when tests fail :)
    - Or better yet use Growl :)
    - YAML test summary output for Test::Reporter et al.