[LLVMdev] git

David A. Greene greened at obbligato.org
Sat Jul 23 05:04:46 PDT 2011

Matthieu Moy <Matthieu.Moy at grenoble-inp.fr> writes:

> With Git's best practice, you may have such history locally, but you
> should not push it as it is. You usually rework your history (things
> like "git rebase -i" are just great for that) to clean it up, and then
> have it accepted by upstream. Ideally, you make sur the testsuite passes
> for each commit before pushing.
> If you like incremental development, you're going to love DVCS.

Let me report a recent experience here.

Some of you may recall that not long ago I proposed a rather intrusive
change to TableGen with the goal of making Inits unique.  My original
patch was large and I was asked to break it up.

Initially, I was quite resistant.  There was much wailing and gnashing
of teeth.  Not really because I thought it was the wrong thing to do,
but because of the pain I forsaw in actually accomplishing the task.

You may have also noted that as the review process went on, I got less
and less resistant.  That's not because I gave up.  :) It's because I
gained a lot of experience with git and things like rebase -i and add
-i.  It was *much* easier to respond to review feedback with git than
with svn because git directly supports the kind of review iteration on
patches that we want to see.  It's designed to make that process less

I think this is another killer feature of git.


More information about the llvm-dev mailing list