Re: VC top of the tree diff and log

[Miles Bader <miles@gnu.org>]

Dan Nicolaescu <dann@ics.uci.edu> writes:
>   > Would be useful, though I dunno what you'd bind it too... at some point
>   > having too many bindings for essentially-similar-commands becomes kind
>   > of annoying/confusing.
>
> I am thinking that we could have 2 modes for the log: 
> - a short one that would be used by default for directory logs
> - the current long one that would be used for file logs
> (I haven't completely thought this through)...

I normally am interested in exactly the same thing from either file or
project logs:  detailed info about the last few changesets.  I use short
logs too (when I'm trying to discover longer-term trends instead of
detailed info), but there seems no connection with whether I restrict
the scope to a file or not.

What sounds nice to me would be a quick "toggle log mode" binding in the
log display buffer (with caching so it's fast, and smart preservation of
point position in the log, etc), or maybe even better, a "toggle
detailed info for this entry" binding, but maybe those aren't so easy to
implement in emacs...  [Hmm... "*" comes to mind for some reason, though
I have no idea why...]

As far as global vs. file logs, I think maybe I'd prefer to _always_ get
global logs in git/svn/... but with some slight hint about the file when
invoked from a file buffer -- e.g., maybe display entries not related to
the file in a "dim" face, and initially position the cursor on the first
modification to the file, etc.  My reasoning is that I don't think I
really mean "show me the history of this file", when I use "C-x v l",
but rather "show me recent history of the project, given that I happen
to be in a file in that project".  [Real file logs would be nice too,
when there's too much chaff, but they seem secondary, rather than
primary.]

[Incidentally, one thing that definitely _should_ work, but doesn't, is
that "C-x v l" in a dired buffer...]

I suppose maybe the real question underlying this thread is how much vc
should try to emulate historical CVS-oriented behavior when used with
newer source control systems that have different models.  Personally I
think it should be a bit looser about this, and do what's "natural"
rather than what's "historically compatible".  But maybe that's just
me...

-Miles

-- 
P.S.  All information contained in the above letter is false,
      for reasons of military security.