Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 1 | git-rerere(1) |
| 2 | ============= |
| 3 | |
| 4 | NAME |
| 5 | ---- |
Junio C Hamano | 7c73c66 | 2007-01-19 00:37:50 | [diff] [blame] | 6 | git-rerere - Reuse recorded resolution of conflicted merges |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 7 | |
| 8 | SYNOPSIS |
| 9 | -------- |
Junio C Hamano | ba4b928 | 2008-07-06 05:20:31 | [diff] [blame] | 10 | 'git rerere' ['clear'|'diff'|'status'|'gc'] |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 11 | |
| 12 | DESCRIPTION |
| 13 | ----------- |
| 14 | |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 15 | In a workflow employing relatively long lived topic branches, |
| 16 | the developer sometimes needs to resolve the same conflicts over |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 17 | and over again until the topic branches are done (either merged |
| 18 | to the "release" branch, or sent out and accepted upstream). |
| 19 | |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 20 | This command assists the developer in this process by recording |
| 21 | conflicted automerge results and corresponding hand resolve results |
| 22 | on the initial manual merge, and applying previously recorded |
| 23 | hand resolutions to their corresponding automerge results. |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 24 | |
Junio C Hamano | 0e47b23 | 2008-01-15 08:31:10 | [diff] [blame] | 25 | [NOTE] |
Junio C Hamano | 543f8d6 | 2009-07-29 08:35:21 | [diff] [blame^] | 26 | You need to set the configuration variable rerere.enabled in order to |
Junio C Hamano | 0e47b23 | 2008-01-15 08:31:10 | [diff] [blame] | 27 | enable this command. |
| 28 | |
Junio C Hamano | e7935c4 | 2006-12-13 21:32:17 | [diff] [blame] | 29 | |
| 30 | COMMANDS |
| 31 | -------- |
| 32 | |
Junio C Hamano | ba4b928 | 2008-07-06 05:20:31 | [diff] [blame] | 33 | Normally, 'git-rerere' is run without arguments or user-intervention. |
Junio C Hamano | e7935c4 | 2006-12-13 21:32:17 | [diff] [blame] | 34 | However, it has several commands that allow it to interact with |
| 35 | its working state. |
| 36 | |
| 37 | 'clear':: |
| 38 | |
| 39 | This resets the metadata used by rerere if a merge resolution is to be |
Junio C Hamano | ccc2f76 | 2008-07-24 04:03:33 | [diff] [blame] | 40 | aborted. Calling 'git-am [--skip|--abort]' or 'git-rebase [--skip|--abort]' |
Junio C Hamano | fce7c7e | 2008-07-02 03:06:38 | [diff] [blame] | 41 | will automatically invoke this command. |
Junio C Hamano | e7935c4 | 2006-12-13 21:32:17 | [diff] [blame] | 42 | |
| 43 | 'diff':: |
| 44 | |
| 45 | This displays diffs for the current state of the resolution. It is |
| 46 | useful for tracking what has changed while the user is resolving |
| 47 | conflicts. Additional arguments are passed directly to the system |
Junio C Hamano | ba4b928 | 2008-07-06 05:20:31 | [diff] [blame] | 48 | 'diff' command installed in PATH. |
Junio C Hamano | e7935c4 | 2006-12-13 21:32:17 | [diff] [blame] | 49 | |
| 50 | 'status':: |
| 51 | |
Junio C Hamano | ba4b928 | 2008-07-06 05:20:31 | [diff] [blame] | 52 | Like 'diff', but this only prints the filenames that will be tracked |
Junio C Hamano | e7935c4 | 2006-12-13 21:32:17 | [diff] [blame] | 53 | for resolutions. |
| 54 | |
| 55 | 'gc':: |
| 56 | |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 57 | This prunes records of conflicted merges that |
| 58 | occurred a long time ago. By default, unresolved conflicts older |
| 59 | than 15 days and resolved conflicts older than 60 |
| 60 | days are pruned. These defaults are controlled via the |
Junio C Hamano | 7464064 | 2006-12-27 10:59:55 | [diff] [blame] | 61 | `gc.rerereunresolved` and `gc.rerereresolved` configuration |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 62 | variables respectively. |
Junio C Hamano | e7935c4 | 2006-12-13 21:32:17 | [diff] [blame] | 63 | |
| 64 | |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 65 | DISCUSSION |
| 66 | ---------- |
| 67 | |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 68 | When your topic branch modifies an overlapping area that your |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 69 | master branch (or upstream) touched since your topic branch |
| 70 | forked from it, you may want to test it with the latest master, |
| 71 | even before your topic branch is ready to be pushed upstream: |
| 72 | |
| 73 | ------------ |
| 74 | o---*---o topic |
| 75 | / |
| 76 | o---o---o---*---o---o master |
| 77 | ------------ |
| 78 | |
| 79 | For such a test, you need to merge master and topic somehow. |
| 80 | One way to do it is to pull master into the topic branch: |
| 81 | |
| 82 | ------------ |
| 83 | $ git checkout topic |
Junio C Hamano | edd2b0a | 2007-01-15 06:12:45 | [diff] [blame] | 84 | $ git merge master |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 85 | |
| 86 | o---*---o---+ topic |
| 87 | / / |
| 88 | o---o---o---*---o---o master |
| 89 | ------------ |
| 90 | |
| 91 | The commits marked with `*` touch the same area in the same |
| 92 | file; you need to resolve the conflicts when creating the commit |
Junio C Hamano | faa1e50 | 2008-08-10 03:55:58 | [diff] [blame] | 93 | marked with `{plus}`. Then you can test the result to make sure your |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 94 | work-in-progress still works with what is in the latest master. |
| 95 | |
| 96 | After this test merge, there are two ways to continue your work |
| 97 | on the topic. The easiest is to build on top of the test merge |
Junio C Hamano | faa1e50 | 2008-08-10 03:55:58 | [diff] [blame] | 98 | commit `{plus}`, and when your work in the topic branch is finally |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 99 | ready, pull the topic branch into master, and/or ask the |
| 100 | upstream to pull from you. By that time, however, the master or |
Junio C Hamano | faa1e50 | 2008-08-10 03:55:58 | [diff] [blame] | 101 | the upstream might have been advanced since the test merge `{plus}`, |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 102 | in which case the final commit graph would look like this: |
| 103 | |
| 104 | ------------ |
| 105 | $ git checkout topic |
Junio C Hamano | edd2b0a | 2007-01-15 06:12:45 | [diff] [blame] | 106 | $ git merge master |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 107 | $ ... work on both topic and master branches |
| 108 | $ git checkout master |
Junio C Hamano | edd2b0a | 2007-01-15 06:12:45 | [diff] [blame] | 109 | $ git merge topic |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 110 | |
| 111 | o---*---o---+---o---o topic |
| 112 | / / \ |
| 113 | o---o---o---*---o---o---o---o---+ master |
| 114 | ------------ |
| 115 | |
| 116 | When your topic branch is long-lived, however, your topic branch |
| 117 | would end up having many such "Merge from master" commits on it, |
| 118 | which would unnecessarily clutter the development history. |
| 119 | Readers of the Linux kernel mailing list may remember that Linus |
| 120 | complained about such too frequent test merges when a subsystem |
| 121 | maintainer asked to pull from a branch full of "useless merges". |
| 122 | |
| 123 | As an alternative, to keep the topic branch clean of test |
| 124 | merges, you could blow away the test merge, and keep building on |
| 125 | top of the tip before the test merge: |
| 126 | |
| 127 | ------------ |
| 128 | $ git checkout topic |
Junio C Hamano | edd2b0a | 2007-01-15 06:12:45 | [diff] [blame] | 129 | $ git merge master |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 130 | $ git reset --hard HEAD^ ;# rewind the test merge |
| 131 | $ ... work on both topic and master branches |
| 132 | $ git checkout master |
Junio C Hamano | edd2b0a | 2007-01-15 06:12:45 | [diff] [blame] | 133 | $ git merge topic |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 134 | |
| 135 | o---*---o-------o---o topic |
| 136 | / \ |
| 137 | o---o---o---*---o---o---o---o---+ master |
| 138 | ------------ |
| 139 | |
| 140 | This would leave only one merge commit when your topic branch is |
| 141 | finally ready and merged into the master branch. This merge |
| 142 | would require you to resolve the conflict, introduced by the |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 143 | commits marked with `*`. However, this conflict is often the |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 144 | same conflict you resolved when you created the test merge you |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 145 | blew away. 'git-rerere' helps you resolve this final |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 146 | conflicted merge using the information from your earlier hand |
| 147 | resolve. |
| 148 | |
Junio C Hamano | ba4b928 | 2008-07-06 05:20:31 | [diff] [blame] | 149 | Running the 'git-rerere' command immediately after a conflicted |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 150 | automerge records the conflicted working tree files, with the |
| 151 | usual conflict markers `<<<<<<<`, `=======`, and `>>>>>>>` in |
| 152 | them. Later, after you are done resolving the conflicts, |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 153 | running 'git-rerere' again will record the resolved state of these |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 154 | files. Suppose you did this when you created the test merge of |
| 155 | master into the topic branch. |
| 156 | |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 157 | Next time, after seeing the same conflicted automerge, |
| 158 | running 'git-rerere' will perform a three-way merge between the |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 159 | earlier conflicted automerge, the earlier manual resolution, and |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 160 | the current conflicted automerge. |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 161 | If this three-way merge resolves cleanly, the result is written |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 162 | out to your working tree file, so you do not have to manually |
Junio C Hamano | ba4b928 | 2008-07-06 05:20:31 | [diff] [blame] | 163 | resolve it. Note that 'git-rerere' leaves the index file alone, |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 164 | so you still need to do the final sanity checks with `git diff` |
Junio C Hamano | ba4b928 | 2008-07-06 05:20:31 | [diff] [blame] | 165 | (or `git diff -c`) and 'git-add' when you are satisfied. |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 166 | |
Junio C Hamano | ba4b928 | 2008-07-06 05:20:31 | [diff] [blame] | 167 | As a convenience measure, 'git-merge' automatically invokes |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 168 | 'git-rerere' upon exiting with a failed automerge and 'git-rerere' |
| 169 | records the hand resolve when it is a new conflict, or reuses the earlier hand |
Junio C Hamano | ba4b928 | 2008-07-06 05:20:31 | [diff] [blame] | 170 | resolve when it is not. 'git-commit' also invokes 'git-rerere' |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 171 | when committing a merge result. What this means is that you do |
| 172 | not have to do anything special yourself (besides enabling |
| 173 | the rerere.enabled config variable). |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 174 | |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 175 | In our example, when you do the test merge, the manual |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 176 | resolution is recorded, and it will be reused when you do the |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 177 | actual merge later with the updated master and topic branch, as long |
| 178 | as the recorded resolution is still applicable. |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 179 | |
Junio C Hamano | ba4b928 | 2008-07-06 05:20:31 | [diff] [blame] | 180 | The information 'git-rerere' records is also used when running |
| 181 | 'git-rebase'. After blowing away the test merge and continuing |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 182 | development on the topic branch: |
| 183 | |
| 184 | ------------ |
| 185 | o---*---o-------o---o topic |
| 186 | / |
| 187 | o---o---o---*---o---o---o---o master |
| 188 | |
| 189 | $ git rebase master topic |
| 190 | |
| 191 | o---*---o-------o---o topic |
| 192 | / |
| 193 | o---o---o---*---o---o---o---o master |
| 194 | ------------ |
| 195 | |
Junio C Hamano | 25e7ba6 | 2009-06-14 01:03:22 | [diff] [blame] | 196 | you could run `git rebase master topic`, to bring yourself |
| 197 | up-to-date before your topic is ready to be sent upstream. |
| 198 | This would result in falling back to a three-way merge, and it |
| 199 | would conflict the same way as the test merge you resolved earlier. |
| 200 | 'git-rerere' will be run by 'git-rebase' to help you resolve this |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 201 | conflict. |
| 202 | |
| 203 | |
| 204 | Author |
| 205 | ------ |
Junio C Hamano | 0868a30 | 2008-07-22 09:20:44 | [diff] [blame] | 206 | Written by Junio C Hamano <gitster@pobox.com> |
Junio C Hamano | 40f2f8d | 2006-02-07 08:04:39 | [diff] [blame] | 207 | |
| 208 | GIT |
| 209 | --- |
Junio C Hamano | f7c042d | 2008-06-06 22:50:53 | [diff] [blame] | 210 | Part of the linkgit:git[1] suite |