0.43 - added SystemLogger.pm module
       - not used by default
       - available by request (esp. by LocalConfig)
       - logs return status in a global location
         - can be used to check for recurring problems like failing node
           hardware
       - initialization of default_attributes was broken

0.42 - added verify_completion_state callback to allow user code to decide
       whether a stage finished successfully, should be retried, or failed
       independently of the stage's exit status
     - added final_job_state to stats to record pass/fail/retry decision
     - removed modules_to_load from tests - they are BoutrosLab specific
     - renamed t/01-templating-role.t to t/01-templating_role.t for consistency
     - fixed a test title
     - added test for forced_retries and fixed it (was broken)
     - removed unused line of code from ScriptSource - would fail if local
       config did not invoke HPCI::ModuleLoad

0.41 - made stage_dir the normal name for the stage creation attribute
     - added stage_dir_name to have the created stage_dir be created under
     group_dir

0.40 - rearranged retry logic
       - added should_choose_retry and choose_retries attributes
         - allows user-provided function to decide whether a retry
           should be attempted after a stage fails (with a limit to
           the maximum number of retries, default 1)

0.39 - Doc updates

0.38 - cleanup of HPCI.pm pod
     - minor typo bugfixes noticed in pod cleanup, no-body had actually
       tripped over them yet

0.37 - add force_retries attribute to stage objects to force a specified
         number of retries before deciding that a stage has really failed
     - some cleanup of pod, more yet to come

0.36 - tighten timeout processing to reduce critical section holes
     - protect against $? being over-written
     - added 'stage_defaults' attribute to group - a hash that will be used as
         attributes for every stage that gets created
     - added 'cluster_specific' attribute to both group and stage creation
         - a hash indexed by cluster type
             - if there is an entry for the actual cluster type
                 - that subhash will be used as attributes
             - other non-matching cluster types will be ignored
         - allows same code to be used for multipe cluster types more easily
     - added a default_attrs hash to HPCI for config over-rides to use

0.35 - added HPCI::LocalConfig support code to allow local configuration
     - cleaned up module dependencies, removed some use statements that are no
     longer used
     - refactored _register_status and _collect_job_stats to support drmaa in
     SGE
     - cleaner (non-Dumper) output for stage status
     - added required modules to Build.PL

0.34 - made dump of job status info more readble

0.33 -
     - use 'method' instead of 'sub' in parameterized roles - the doc says
         that this is required, although it seemed to be working before
     - added support for abort for cases when the driver can identify a
         situation in which the job was not attempted to be run
     - added direct support for kill, killsignal rather than leaving them
         tangled in exit_status and subject to local interpretation
     - rearranged a number of methods and modifiers so that they could be
         overridden in drivers smoothly
     - removed default from HPCI::ModuleLoad, made it an empty array that can
         be loaded by a config wrapper module
     - added support for extra roles to be provided for local environments
         and/or specific cluster types.  (e.g. ModuleLoad might be selected
         for labs that use it for all cluster types)
     - added ScriptSource extra role to allow a local environment to request
         adding a line at the beginning of the generated script for each run
         to source necessary config stuff
         It defaults to an empty array that can be provided by a config
         wrapper module
     - reordered stuff relating to running, needed to provide alternate
         control for different run styles in drivers (in particular, using DRMAA
         to manage SGE)
0.32 - major cleanup of existing tests
     - more tests added
     - bug fixes
       - mkdir initialization handling
       - timeout termination handling
       - failure handling choices all work (skip_deps, skip_all, ignore)
     - enhancements
       - made dep/deps and pre_req/pre_reqs be synonyms for add_deps method
         - both accept a scalar or list
       - added ChangeLog and README

0.31 - initial CPAN release
0.30 - separate SGE into an independent module preparing for CPAN release
0.01 .. 0.25 - BoutrosLab internal releases