log --date question

[Matt Mackall]

On Wed, Dec 06, 2006 at 02:44:47PM -0800, Danek Duvall wrote:
> On Wed, Dec 06, 2006 at 11:00:09PM +0100, Mercurial Commits wrote:
> 
> > http://www.selenic.com/hg/rev/fc5ba0ab7f45
> > changeset:   3813:fc5ba0ab7f45
> > tag:         tip
> > user:        Matt Mackall <mpm at selenic.com>
> > date:        Wed Dec 06 15:29:17 2006 -0600
> > summary:     Add --date support to log
> 
> Should 
> 
>     http://www.selenic.com/mercurial/bts/issue77
> 
> ever be implemented, how would one distinguish between a request to list
> changesets whose commit dates matched the given date and a request to list
> changesets which were added to the repo on date matching the given date?

If I understand the distinction you're trying to make, you're looking
to talk about the dates that a changeset was pulled into a given repo.

We don't keep that information. Nor do I think that there exists a
good way to track it. As many pulls are between short-lived
repositories or mirrors, we'd have to keep an arbitrarily deep history
of pulls per changeset for this information to be useful.

Consider: is the date you want the date it showed up in the "canonical
repo" (a nonexistent concept to hg), the date the canonical repo got
cloned to make the development branch, the date it got pulled to your
department mirror, the date you pulled it onto your laptop, the date
you cloned your laptop mirror to your feature branch, the date you
pulled your feature branch into your integration tree for testing, the
various dates you pushed and pulled your work between your laptop and
your desktop, or the date you pushed your testing tree to your
published tree for integration? That of course is ignoring all the
steps a changeset took to get to the "canonical repo" in the first place.

All of these transaction are equivalent to Mercurial.

-- 
Mathematics is the supreme nostalgia of our time.