From: Tim Henigan Date: Sat, 17 Mar 2012 02:26:11 GMT Subject: Re: [PATCH 0/9] difftool: teach command to perform directory diffs Message-ID: In-Reply-To: <7vlin0dlrm.fsf@alter.siamese.dyndns.org> On Fri, Mar 16, 2012 at 10:06 PM, Junio C Hamano wrote: > Tim Henigan writes: > >> This is the second phase of development for this feature. The first phase >> was added as a separate command (git diffall) in 1252bbe (contrib: add >> git-diffall script). > > Ummm.  I do not think the first step is not even done yet; fixing the > whitespace-in-pathspec and whitespace-or-lf-in-paths issues now the script > is in contrib/, while having people play with it. I began fixing those issues in 'git-diffall' by rewriting the loop that populates the tmp dirs in Perl. By the time I changed the loop to use 'git diff --raw -z' and handle working copy changes and submodules, it seemed I was very near to finishing the things I wanted to accomplish in "phase 2". Finishing the work by updating 'difftool' seemed to actually be easier than trying to update 'diffall'. I realize the feature hasn't had any time to cook in contrib/...perhaps I am over-enthusiastic. However, the 'diffall' concept has been in steady use by a small audience for nearly two years (by people who downloaded/forked the GitHub project). This past week I was lucky enough to have some time to study the problem and propose a solution. If it needs to wait while 'diffall' cooks in contrib, I will try to be more patient ;) The changes in patches 1-6 may be useful regardless. They cut the length of the script in half without changing how it behaves. The only functional change in those patches is that `git difftool -h` now finishes with exit(0) instead of exit(1). I understand that may not be enough reason to accept such a large change in working code...but like I said, I was lucky enough to have time to study the problem in detail this week.