Cancelling certain commits

Hi,

If I'm not happy with the changes I made in certain commits during development, how do I cancel those commits and remove them from git records most cleanly?

For example if I did commits 1 to 10, and want to get rid of commit 3 and 7, such that, the other commits are irrelevant to changes made in 3 and 7, but may involve changes in the same file as 3 and 7 changed, how do I handle it?

Furthermore, how would I handle it if there was a commit 4, that depended partially on commit 3? (For example if it uses a type that was changed in commit 3?) Would I hand-edit commit 4 to fix it?

Many thanks, Bahadir

Bahadir Balban wrote:

As a general rule, it's desirable to do

	$ git revert --edit

if you've made some programmatical error which is non-obvious (not a typo, fe), as you can then make clear why that code wasn't proper for implementation. It will also hinder others from doing the same mistake.

Enough lessons though.

# make sure "git status" doesn't show any changes

$ git branch anchor # to make a safe point to reset to if things go bad $ git reset --hard HEAD~7 # undo commits 3 - 10 # replay commits 4, 5, 6, 8, 9 $ for i in 6 5 4 2 1; do git cherry-pick -r anchor~$i || break; done # replay commit 10 $ git cherry-pick -r anchor

# done

Another variant: $ git format-patch --mbox -k HEAD~7 $ git reset --hard HEAD~7 # remove undesired commits from the files created $ git am -k 000*

# done

If some commits depend on other commits both of these will fail and you'll have to fix up by hand.

Some more lecturing then; You should have used a topic-branch to introduce such changes. This is always desirable when making a change that requires a series of patches.

If this didn't actually require several commits you should have done them as one (I always try to make sure my projects at least compile cleanly, even in topic-branches although they may contain warnings about static but unused functions and variables).

As for fixing up commits by hand when they're already written to the object database, you can't really do that (in an easy way, at least). What you can do is use "git format-patch" to get patches in mbox format for all the commits you eventually wish to keep, then do

$ git reset --hard <one-committish-*after*-undesired> $ git cat-file commit HEAD > msg $ git reset --soft HEAD~1 # edit changed files and msg $ git commit -F msg changed_file1.c changed_file2.c changed_filen.c

Note that in all cases you should create an anchor branch to get you back to where you were in case something goes wrong, and never *ever* do

$ git prune

on a repo where you're juggling loose commits like this (although that should actually be safe when you create 'anchor' as a branch rather than as a tag, which Junio usually recommends).

Andreas Ericsson <ae@op5.se> writes:

 ... (all the good stuff not repeated).

As a general rule, first of all, make sure that this kind of "cleaning up the history" can be done only before you make your history available to others, and not after.

I think "6 5 4 2 1 0" would be sweeter.  rev~0 == rev

Sorry, I do not understand this part. tags and branch heads are used in the same way to determine what objects to protect from "prune", so I thought there was no difference from the safety point of view between the anchor being a branch or a tag.

Junio C Hamano wrote:

Didn't know that. The day isn't a complete loss already. :)

You're right. That's weird though, because I distinctly remember being bitten by this when I was very new to git. OTOH, I probably did something stupid that broke the repo before, and my girlfriend provided plenty distraction so I didn't do the pruning and the reset'ing on the same day.

Would a tag with the same name in a central repo overwrite a local one when "git fetch --tags" is used? I think I did this over http, but I'm not sure. I know I had to clean up un-annotated tags that got pushed by accident after I had restored the lost commits though, so this could be a cause (there was an 'anchor' tag in the shared repo as well).