Re: [PATCH] More permissive "git-rm --cached" behavior without -f.

From: Jeff King <peff@peff.net>
Date: Jul 14, 2007, 03:42 UTC
Message-ID: <20070714034221.GA23328@coredump.intra.peff.net>
In-Reply-To: <vpq8x9kp231.fsf@bauges.imag.fr>

On Fri, Jul 13, 2007 at 08:53:38PM +0200, Matthieu Moy wrote:

Show 7 quoted lines

> do.". The behavior without --cached is not modified, except for the
> error message, and the previous was to require -f whenever the index
> doesn't match the head, *or* doesn't match the file. So, without
> --cached, you need to have file=index=HEAD to be able to git-rm.
> 
> If I missunderstand you, please, provide a senario where my patch
> doesn't do the expected.

Right, my point was that there is a case where running without --cached could lose content: when there is no working tree file. However, thinking about it more, I recall that Junio made the point that allowing that behavior means the CVS idiom of "rm file; git-rm file" will just work.

Not that that was a problem you introduced; I merely wanted to push for total consistency rather than just handling --cached. But I think the non --cached behavior is actually right now, so let me retract my complaint.

And assuming the "git-rm when no working tree file" current behavior is OK, then I think your patch removes the last consistency problem that I mentioned in my state table here:

  http://article.gmane.org/gmane.comp.version-control.git/51449

So in a round-about way, I totally approve of your patch. Sorry for the confusion.

-Peff