On Mon, Jan 30, 2017 at 7:11 PM, Junio C Hamano <gitster@xxxxxxxxx> wrote: > Benjamin Fuchs <email@xxxxxxxxxxxxxxxx> writes: > >> In [2/4] I got rid of the loop by feedback of Gábor. >> Sorry if my patch wasn't well formed. > > While it might be the way other development communities work, in the > Git development community, we do not work that way when presenting > your second and subsequent attempt to the community. > > Having the initial draft from the original developers that records > the bugs and misdesigns in an earlier parts of a series and separate > patches that record how the problems were fixed to arrive at the > final shape of the codebase might be interesting to the original > developers, and they may even find such a history valuable, but in > the context of the history that will be recorded in the official > tree of the project for eternity, that just adds useless noise. > > Instead of keeping the original, in which problems were pointed out, > and adding later commits to correct them incrementally, a patch is > "rerolled". That is, you are expected to learn from the review > comments and pretend as if you did the work from scratch and you > already possessed the wisdom lent by the reviewers when you started > your work. In the "rerolled" patches you send, you pretend as if > you worked without making mistakes you made in the earlier rounds at > all, producing (more) perfect patches from the beginning. > > In reality, you may locally be using Git tools like rebase, > cherry-pick and "add -p" while preparing these "rerolled" rounds of > patches, but the name of the game is to hide that effort from the > public and pretend to be a perfect human, recording the result of > exercising your best ability in the official history ;-). > > So this is OK: > > 0/3: I want to improve X, and for that I identified that I need > A, B and C done. A or B alone is already an improvement, but A > and B together makes it even more useful, and implementation of > C requires patches to do A and B. > > 1/3: do A > 2/3: do B > 3/3: do C, building on A and B > > This is not: > > 0/3: I want to improve X, and for that I need to do C. > 1/3: I couldn't do C, and I did A instead. > 2/3: A was totally useless. I fix it to do B. > 3/3: B is not very useful, either. I fix it to do C. > I agree with Junio here, "git rebase --interactive" and then editing/squashing commits is your friend. (unrelated side note:) At GitMerge facebook presented their improvements on mercurial and one of the things was "hg absorb". It would take the dirty hunks/lines of the working tree and amend them into the "stack of commits", i.e. into your local unpublished history. So instead of making fixup commits and doing the whole interactive rebase thing, it would do it automatically for you. I think that is a neat time saver. Thanks, Stefan