Re: add -e, was Re: What's cooking in git.git (Apr 2009, #02; Sun, 12)

From: Junio C Hamano <gitster@pobox.com>
Date: Apr 14, 2009, 18:40 UTC
Message-ID: <7vhc0r3wpq.fsf@gitster.siamese.dyndns.org>
In-Reply-To: <alpine.DEB.1.00.0904141943060.10279@pacific.mpi-cbg.de>

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

Show 10 quoted lines

> On Sun, 12 Apr 2009, Junio C Hamano wrote:
>
>> * js/add-edit (Wed Apr 8 23:30:24 2009 +0200) 1 commit
>>  - git-add: introduce --edit (to edit the diff vs. the index)
>> 
>> I am Ok with the general idea, but the error detection needs to be more 
>> robust than merely relying on --recount.
>
> You mean something like saving an extra copy of the patch, and checking if 
> common or removed lines were either removed or kept intact?

No, editing a removed line and changing it to an unchanged line is perfectly fine.

I was thinking more about people touching the lines near the hunk boundary (e.g. insert a new line at the beginning of the hunk) which would not be compatible without --unidiff-zero hack while applying, and --unidiff-zero hack should not be used if we care about the correctness.