[MERGE] Faster commit

[Ian Clatworthy]

Aaron Bentley wrote:
> Aaron Bentley has voted +1.
> Status is now: Semi-approved
> Comment:
> This is fine as an incremental improvement, but I don't like the way
> we're invoking _note_before_complete unconditionally.  If
> ReportCommitToLog.deleted is invoked *after* ReportCommitToLog.complete,
> then we'll invoke _note_before_complete *after* complete.
> 
> I may just have a naming neurosis, but absent a guarantee that
> ReportCommitToLog.deleted cannot be invoked after
> ReportCommitToLog.complete, I think it should be called _note_buffered
> or something.

Well picked up. note_before_complete() would have been better named as
note_before_finished() (or note_buffered() as you suggested): the
association with the completed() call was accidental from a naming
perspective.

Naming neurosis is a very good thing. Understandability is a slippery
slope and clear names are vital. I'm always extremely happy to use
better names or make any other changes that improve code clarity.

As luck would have it, note_before_completed() will simply become note()
as I've been asked to remove the complexity introduced by the buffering.
The problem it was solving will be addressed by logging just to stdout
directly, rather than stderr+.bzr.log as implied by bzrlib.note. Updated
patch coming soon ...

Ian C.