Permalink
Commits on Feb 28, 2011
  1. @schwern

    Version 2.00_07

    schwern committed Feb 28, 2011
Commits on Feb 23, 2011
  1. @schwern

    A skip() or todo_skip() would not start the stream

    If called before a plan is set, there would be no TAP header.
    
    Fixes #132
    schwern committed Feb 23, 2011
Commits on Feb 22, 2011
  1. @schwern
Commits on Feb 3, 2011
  1. @schwern
  2. @schwern

    Make TB1->diag and ->note issue log events and so use the formatter.

    Unfortunately this makes changing the todo_output make no sense.
    
    Closes #110
    schwern committed Feb 3, 2011
  3. @schwern

    Make the TAP formatter respect events.

    Re #110
    schwern committed Feb 3, 2011
  4. @schwern
  5. @schwern

    Implement log level comparison via between_levels()

    This is deliberately minimalist to avoid YAGNI and encourage range based level comparison.
    
    Also:
     Incompatible
     * The emergency and critical log levels have been removed.  They didn't have
       a use case.
     * TB2::Event::Log->levels now returns in order from lowest to highest.
    
     New Features
     * Added TB2::Event::Log->between_levels() to compare log levels.
    
     Documentation
     * The suggested meanings of event log levels are now documented.
    
    Closes #111
    schwern committed Feb 1, 2011
Commits on Feb 1, 2011
  1. @schwern

    Change the log event to use all symbolic levels.

    This lets us insert new levels later.
    
    Re #111
    schwern committed Jan 31, 2011
Commits on Jan 27, 2011
  1. @schwern

    Version 2.00_06

    schwern committed Jan 27, 2011
Commits on Jan 26, 2011
  1. @schwern
  2. @schwern

    TB2->ok() will now start the stream if there is not one already

    This is what you pretty much always want, it's compatible with the done_testing system.
    schwern committed Jan 26, 2011
  3. @schwern
  4. @schwern
  5. @schwern
  6. @schwern
  7. @schwern

    Remove TB1's Have_Plan tracking. Rely on History instead.

    This is in case someone else set the plan, or TB1 set it for a different EC.
    
    For #125
    schwern committed Jan 26, 2011
  8. @schwern
  9. @schwern

    Update MANIFEST

    schwern committed Jan 26, 2011
  10. @schwern

    Fix TB1 so it works without a formatter. This lets TB2::Tester work.

    There are two hacks here.  First, the Formatter is replaced with a BlackHole object.
    This avoids TB1 (and other builders) from having to check that the formatter is there so
    long as its functionality is optional.
    
    Second, changing the current test with TB1 relies on having a counter.  So we fake one
    using the history object.  TB1 will not remember the count has changed.  But that's ok,
    only the TAP formatter cares about that and there is no TAP formatter.
    
    Closes #81
    schwern committed Jan 26, 2011
  11. @schwern

    Remove the "last_test_seen" hack in TB1 (mostly) freeing its dependen…

    …cy on the TAP formatter.
    
    This is only necessary because the TAP formatter unsets the counter on stream end, so
    TB1 can't rely on its count being there for its ending analysis.  How the TAP formatter will
    handle multiple/nested streams is going to be redone, so just get rid of it.
    
    For #81
    schwern committed Jan 26, 2011
Commits on Jan 25, 2011
  1. @schwern
  2. @schwern
  3. @schwern
  4. @schwern
  5. @schwern

    Make the EventCoordinator singleton a wrapper around a real EventCoor…

    …dinator.
    
    This allows builders to conveniently store the singleton while allowing things like
    TB2::Tester to swap it out.
    schwern committed Jan 25, 2011
  6. @schwern

    Move stream_depth() from the TAP formatter to History so it's availab…

    …le to everyone.
    
    This necessitated...
    * Ensuring that accept_event/result is called with an event coordinator
      because things might want to get the history now
    * Letting history see events before the formatter to update the stream_depth
      (and any future "state of the test" statistics)
    
    There's no longer a need for the formatter to do anything special to an event, so
    INNER_accept_event was no longer necessary.  The restriction that a result must be
    inside a stream is not universal, so INNER_accept_result was no longer necessary.
    
    Closes #118
    schwern committed Jan 25, 2011
  7. @schwern
  8. @schwern

    Remove the indentation from the TAP formatter.

    Nothing is using it.  We can do it better.
    
    Closes #120
    schwern committed Jan 25, 2011
  9. @schwern

    Add TB2::CanLoad to protect $! and $@ and make everything use it.

    t/subtest/args.t would mysteriously fail due to a heisenbug.  That turned out to be
    because module loading would get delayed until in the middle of processing a test.  That
    would call require() which would blow away $@.  So this guards against that.
    schwern committed Jan 25, 2011
  10. @schwern
  11. @schwern

    Formatter, Counter and History no longer have to be singletons. You u…

    …se whatever is in your
    
    EventCoordinator.
    
    Closes #117
    schwern committed Jan 25, 2011
Commits on Jan 24, 2011
  1. @schwern

    Change the EventCoordinator to have one history object.

    There's no use case for multiple history objects and it just makes things a pain.
    
    Closes #115
    schwern committed Jan 24, 2011
  2. @schwern

    Version 2.00_05

    schwern committed Jan 24, 2011
  3. @schwern

    The newlines on pipes appear broken on Strawberry, not our problem.

    \n in Strawberry is \012 (I don't know why) but the pipe has \015\012.  Anyhow,
    this isn't important so we'll just regex around it.
    schwern committed Jan 24, 2011