Re: What's cooking in git.git (Apr 2013, #05; Mon, 15)

From: Jeff King <peff@peff.net>
Date: Apr 18, 2013, 20:30 UTC
Message-ID: <20130418203035.GB24690@sigill.intra.peff.net>
In-Reply-To: <7v61zj66wu.fsf@alter.siamese.dyndns.org>

On Thu, Apr 18, 2013 at 11:16:33AM -0700, Junio C Hamano wrote:

Show 11 quoted lines

> Jeff King <peff@peff.net> writes:
> 
> > could work for both cases. Something like "not considering" (or another
> > synonym for "considering") might be even more accurate. It is not just
> > that we did not stage it; it is what we did not even consider it an item
> > for staging under the current rules.
> 
> Yes, "not considering" is much more sensible, while side-stepping
> the dryrun issue.  Or
> 
>        warning("ignoring removal of '%s'")

I like that much better than either of my suggestions.

Show 5 quoted lines

> > Note that the "not staging" warnings may potentially be interspersed
> > with the normal dry-run output. I think that's OK.
> 
> As long as the top-text makes it clear what "not considering" (or
> "ignoring") in the following text means, I think it is fine.

Agreed, and I think the current text is fine for that (though neither of us is the best judge at this point of how a less familiar user would interpret it).

Show 5 quoted lines

> But I think we are doing users a disservice by listing tons of
> paths.  Where the difference of versions matters _most_ is when the
> user has tons of removed paths in the working tree.  Either with one
> warning per path, or a block of collected paths at the end, we are
> scrolling the more important part of the message up.

I'm not sure I agree. Even with a handful, it made me wonder why one was mentioned and not others. That _could_ be cleared up by rewording (i.e., making it clear that this is an example, and there may be more). But somehow listing them is what I would expect. Perhaps because it gives the user a clue about what to do next; they ask themselves "did I want those updated or not?".

In the orphaned-commit message when leaving a detached HEAD, we collect the answer, say "you are leaving N commits", and show the first 5 five of them, with an ellipsis at the end if we didn't show them all. Would it makes sense to do that here?

Yet another alternative would be to print a warning for each path, but hold the main warning for the end, so that it is the first thing the user sees. That has the added bonus that regular "--dry-run" output will not scroll it away, either.

-Peff