Re: [PROBE] cg-commit: show and enable editing of changes with --review

From: Jonas Fonseca <fonseca@diku.dk>
Date: Dec 1, 2005, 13:20 UTC
Message-ID: <20051201132025.GA3874@diku.dk>
In-Reply-To: <7vbr03564t.fsf@assigned-by-dhcp.cox.net>

Junio C Hamano <junkio@cox.net> wrote Tue, Nov 29, 2005:

Show 13 quoted lines

> Jonas Fonseca <fonseca@diku.dk> writes:
> 
> > Show changes being commited as a patch appended to the commit message
> > buffer. If the original patch is different from the patch extracted from
> > the commit message file the original patch will be reverted and the edited
> > patch applied before completing the commit.
> >
> > Due to limitations with cg-patch this can only be used when commiting
> > from the project root directory. The error handling if the either the
> > original patch or the edited patch does not apply is not optimal, since
> > cg-patch will not report errors properly.
> 
> I do not do Porcelains,

I am curious since you have also used this phrase in the past: didn't you do your own jit porcelain once and are you saying that git does not ship with it's own porcelain?

Show 5 quoted lines

> I am not a Cogito user, and I generally
> do not like encouraging people who are playing an individual
> developer role to commit something that has never existed in
> their working tree (hence by definition never been tested),
> but...

I see your point. I mostly intend to use it for fixing small typos in comments and documentation. It can help to cure some "post-commit stress". Also I hope you will agree that it is always a good idea to review your changes one more time although I agree it might be a little silly.

Show 11 quoted lines

> > +		echo "Updating changes to edited patch"
> > +		# FIXME: Can only be run from the top level
> > +		# FIXME: Is very 'fragile' error handling. We should probably
> > +		# save the original patch in a local file for recovery?
> > +		if ! cg-patch -R < $PATCH; then
> > +			backup=$(mktemp commit-backup.XXXXXX)
> > +			cp $PATCH $backup
> > +			error_msg="unable to revert original patch, backup saved to $backup"
> 
> I suspect at least you should be able to use checkout-index for
> the first one instead of cg-patch.

Oh, great, thanks. I guess I should work some more on the git docs to get to know these kind of things.

-- 
Jonas Fonseca