Junio C Hamano | 1a4e841 | 2005-12-27 08:17:23 | [diff] [blame] | 1 | From: Junio C Hamano <junkio@cox.net> |
| 2 | To: git@vger.kernel.org |
| 3 | Subject: [HOWTO] Reverting an existing commit |
| 4 | Abstract: In this article, JC gives a small real-life example of using |
| 5 | 'git revert' command, and using a temporary branch and tag for safety |
| 6 | and easier sanity checking. |
| 7 | Date: Mon, 29 Aug 2005 21:39:02 -0700 |
| 8 | Content-type: text/asciidoc |
| 9 | Message-ID: <7voe7g3uop.fsf@assigned-by-dhcp.cox.net> |
| 10 | |
| 11 | Reverting an existing commit |
| 12 | ============================ |
| 13 | |
| 14 | One of the changes I pulled into the 'master' branch turns out to |
| 15 | break building GIT with GCC 2.95. While they were well intentioned |
| 16 | portability fixes, keeping things working with gcc-2.95 was also |
| 17 | important. Here is what I did to revert the change in the 'master' |
| 18 | branch and to adjust the 'pu' branch, using core GIT tools and |
| 19 | barebone Porcelain. |
| 20 | |
| 21 | First, prepare a throw-away branch in case I screw things up. |
| 22 | |
| 23 | ------------------------------------------------ |
| 24 | $ git checkout -b revert-c99 master |
| 25 | ------------------------------------------------ |
| 26 | |
| 27 | Now I am on the 'revert-c99' branch. Let's figure out which commit to |
| 28 | revert. I happen to know that the top of the 'master' branch is a |
| 29 | merge, and its second parent (i.e. foreign commit I merged from) has |
| 30 | the change I would want to undo. Further I happen to know that that |
| 31 | merge introduced 5 commits or so: |
| 32 | |
| 33 | ------------------------------------------------ |
| 34 | $ git show-branch --more=4 master master^2 | head |
Junio C Hamano | 38064d5 | 2006-01-15 10:13:49 | [diff] [blame] | 35 | * [master] Merge refs/heads/portable from http://www.cs.berkeley.... |
Junio C Hamano | 1a4e841 | 2005-12-27 08:17:23 | [diff] [blame] | 36 | ! [master^2] Replace C99 array initializers with code. |
| 37 | -- |
Junio C Hamano | 38064d5 | 2006-01-15 10:13:49 | [diff] [blame] | 38 | - [master] Merge refs/heads/portable from http://www.cs.berkeley.... |
| 39 | *+ [master^2] Replace C99 array initializers with code. |
| 40 | *+ [master^2~1] Replace unsetenv() and setenv() with older putenv(). |
| 41 | *+ [master^2~2] Include sys/time.h in daemon.c. |
| 42 | *+ [master^2~3] Fix ?: statements. |
| 43 | *+ [master^2~4] Replace zero-length array decls with []. |
| 44 | * [master~1] tutorial note about git branch |
Junio C Hamano | 1a4e841 | 2005-12-27 08:17:23 | [diff] [blame] | 45 | ------------------------------------------------ |
| 46 | |
| 47 | The '--more=4' above means "after we reach the merge base of refs, |
| 48 | show until we display four more common commits". That last commit |
| 49 | would have been where the "portable" branch was forked from the main |
| 50 | git.git repository, so this would show everything on both branches |
| 51 | since then. I just limited the output to the first handful using |
| 52 | 'head'. |
| 53 | |
| 54 | Now I know 'master^2~4' (pronounce it as "find the second parent of |
| 55 | the 'master', and then go four generations back following the first |
| 56 | parent") is the one I would want to revert. Since I also want to say |
| 57 | why I am reverting it, the '-n' flag is given to 'git revert'. This |
| 58 | prevents it from actually making a commit, and instead 'git revert' |
| 59 | leaves the commit log message it wanted to use in '.msg' file: |
| 60 | |
| 61 | ------------------------------------------------ |
| 62 | $ git revert -n master^2~4 |
| 63 | $ cat .msg |
| 64 | Revert "Replace zero-length array decls with []." |
| 65 | |
| 66 | This reverts 6c5f9baa3bc0d63e141e0afc23110205379905a4 commit. |
| 67 | $ git diff HEAD ;# to make sure what we are reverting makes sense. |
| 68 | $ make CC=gcc-2.95 clean test ;# make sure it fixed the breakage. |
| 69 | $ make clean test ;# make sure it did not cause other breakage. |
| 70 | ------------------------------------------------ |
| 71 | |
| 72 | The reverted change makes sense (from reading the 'diff' output), does |
| 73 | fix the problem (from 'make CC=gcc-2.95' test), and does not cause new |
| 74 | breakage (from the last 'make test'). I'm ready to commit: |
| 75 | |
| 76 | ------------------------------------------------ |
| 77 | $ git commit -a -s ;# read .msg into the log, |
| 78 | # and explain why I am reverting. |
| 79 | ------------------------------------------------ |
| 80 | |
| 81 | I could have screwed up in any of the above steps, but in the worst |
| 82 | case I could just have done 'git checkout master' to start over. |
| 83 | Fortunately I did not have to; what I have in the current branch |
| 84 | 'revert-c99' is what I want. So merge that back into 'master': |
| 85 | |
| 86 | ------------------------------------------------ |
| 87 | $ git checkout master |
| 88 | $ git resolve master revert-c99 fast ;# this should be a fast forward |
| 89 | Updating from 10d781b9caa4f71495c7b34963bef137216f86a8 to e3a693c... |
| 90 | cache.h | 8 ++++---- |
| 91 | commit.c | 2 +- |
| 92 | ls-files.c | 2 +- |
| 93 | receive-pack.c | 2 +- |
| 94 | server-info.c | 2 +- |
| 95 | 5 files changed, 8 insertions(+), 8 deletions(-) |
| 96 | ------------------------------------------------ |
| 97 | |
| 98 | The 'fast' in the above 'git resolve' is not a magic. I knew this |
| 99 | 'resolve' would result in a fast forward merge, and if not, there is |
| 100 | something very wrong (so I would do 'git reset' on the 'master' branch |
| 101 | and examine the situation). When a fast forward merge is done, the |
| 102 | message parameter to 'git resolve' is discarded, because no new commit |
| 103 | is created. You could have said 'junk' or 'nothing' there as well. |
| 104 | |
| 105 | There is no need to redo the test at this point. We fast forwarded |
| 106 | and we know 'master' matches 'revert-c99' exactly. In fact: |
| 107 | |
| 108 | ------------------------------------------------ |
| 109 | $ git diff master..revert-c99 |
| 110 | ------------------------------------------------ |
| 111 | |
| 112 | says nothing. |
| 113 | |
| 114 | Then we rebase the 'pu' branch as usual. |
| 115 | |
| 116 | ------------------------------------------------ |
| 117 | $ git checkout pu |
| 118 | $ git tag pu-anchor pu |
| 119 | $ git rebase master |
| 120 | * Applying: Redo "revert" using three-way merge machinery. |
| 121 | First trying simple merge strategy to cherry-pick. |
| 122 | Finished one cherry-pick. |
| 123 | * Applying: Remove git-apply-patch-script. |
| 124 | First trying simple merge strategy to cherry-pick. |
| 125 | Simple cherry-pick fails; trying Automatic cherry-pick. |
| 126 | Removing Documentation/git-apply-patch-script.txt |
| 127 | Removing git-apply-patch-script |
| 128 | Finished one cherry-pick. |
| 129 | * Applying: Document "git cherry-pick" and "git revert" |
| 130 | First trying simple merge strategy to cherry-pick. |
| 131 | Finished one cherry-pick. |
| 132 | * Applying: mailinfo and applymbox updates |
| 133 | First trying simple merge strategy to cherry-pick. |
| 134 | Finished one cherry-pick. |
| 135 | * Applying: Show commits in topo order and name all commits. |
| 136 | First trying simple merge strategy to cherry-pick. |
| 137 | Finished one cherry-pick. |
| 138 | * Applying: More documentation updates. |
| 139 | First trying simple merge strategy to cherry-pick. |
| 140 | Finished one cherry-pick. |
| 141 | ------------------------------------------------ |
| 142 | |
| 143 | The temporary tag 'pu-anchor' is me just being careful, in case 'git |
| 144 | rebase' screws up. After this, I can do these for sanity check: |
| 145 | |
| 146 | ------------------------------------------------ |
| 147 | $ git diff pu-anchor..pu ;# make sure we got the master fix. |
| 148 | $ make CC=gcc-2.95 clean test ;# make sure it fixed the breakage. |
| 149 | $ make clean test ;# make sure it did not cause other breakage. |
| 150 | ------------------------------------------------ |
| 151 | |
| 152 | Everything is in the good order. I do not need the temporary branch |
| 153 | nor tag anymore, so remove them: |
| 154 | |
| 155 | ------------------------------------------------ |
| 156 | $ rm -f .git/refs/tags/pu-anchor |
| 157 | $ git branch -d revert-c99 |
| 158 | ------------------------------------------------ |
| 159 | |
| 160 | It was an emergency fix, so we might as well merge it into the |
| 161 | 'release candidate' branch, although I expect the next release would |
| 162 | be some days off: |
| 163 | |
| 164 | ------------------------------------------------ |
| 165 | $ git checkout rc |
| 166 | $ git pull . master |
| 167 | Packing 0 objects |
| 168 | Unpacking 0 objects |
| 169 | |
| 170 | * committish: e3a693c... refs/heads/master from . |
| 171 | Trying to merge e3a693c... into 8c1f5f0... using 10d781b... |
| 172 | Committed merge 7fb9b7262a1d1e0a47bbfdcbbcf50ce0635d3f8f |
| 173 | cache.h | 8 ++++---- |
| 174 | commit.c | 2 +- |
| 175 | ls-files.c | 2 +- |
| 176 | receive-pack.c | 2 +- |
| 177 | server-info.c | 2 +- |
| 178 | 5 files changed, 8 insertions(+), 8 deletions(-) |
| 179 | ------------------------------------------------ |
| 180 | |
| 181 | And the final repository status looks like this: |
| 182 | |
| 183 | ------------------------------------------------ |
| 184 | $ git show-branch --more=1 master pu rc |
| 185 | ! [master] Revert "Replace zero-length array decls with []." |
| 186 | ! [pu] git-repack: Add option to repack all objects. |
| 187 | * [rc] Merge refs/heads/master from . |
| 188 | --- |
| 189 | + [pu] git-repack: Add option to repack all objects. |
| 190 | + [pu~1] More documentation updates. |
| 191 | + [pu~2] Show commits in topo order and name all commits. |
| 192 | + [pu~3] mailinfo and applymbox updates |
| 193 | + [pu~4] Document "git cherry-pick" and "git revert" |
| 194 | + [pu~5] Remove git-apply-patch-script. |
| 195 | + [pu~6] Redo "revert" using three-way merge machinery. |
Junio C Hamano | 38064d5 | 2006-01-15 10:13:49 | [diff] [blame] | 196 | - [rc] Merge refs/heads/master from . |
| 197 | ++* [master] Revert "Replace zero-length array decls with []." |
| 198 | - [rc~1] Merge refs/heads/master from . |
| 199 | ... [master~1] Merge refs/heads/portable from http://www.cs.berkeley.... |
Junio C Hamano | 1a4e841 | 2005-12-27 08:17:23 | [diff] [blame] | 200 | ------------------------------------------------ |