1GIT-REBASE(1) Git Manual GIT-REBASE(1)
2
6 git-rebase - Reapply commits on top of another base tip
7
9 git rebase [-i | --interactive] [<options>] [--exec <cmd>]
10 [--onto <newbase> | --keep-base] [<upstream> [<branch>]]
11 git rebase [-i | --interactive] [<options>] [--exec <cmd>] [--onto <newbase>]
12 --root [<branch>]
13 git rebase (--continue | --skip | --abort | --quit | --edit-todo | --show-current-patch)
14
16 If <branch> is specified, git rebase will perform an automatic git
17 switch <branch> before doing anything else. Otherwise it remains on the
18 current branch.
19 If <upstream> is not specified, the upstream configured in
21 branch.<name>.remote and branch.<name>.merge options will be used (see
22 git-config(1) for details) and the --fork-point option is assumed. If
23 you are currently not on any branch or if the current branch does not
24 have a configured upstream, the rebase will abort.
25 All changes made by commits in the current branch but that are not in
27 <upstream> are saved to a temporary area. This is the same set of
28 commits that would be shown by git log <upstream>..HEAD; or by git log
29 'fork_point'..HEAD, if --fork-point is active (see the description on
30 --fork-point below); or by git log HEAD, if the --root option is
31 specified.
32 The current branch is reset to <upstream>, or <newbase> if the --onto
34 option was supplied. This has the exact same effect as git reset --hard
35 <upstream> (or <newbase>). ORIG_HEAD is set to point at the tip of the
36 branch before the reset.
37 The commits that were previously saved into the temporary area are then
39 reapplied to the current branch, one by one, in order. Note that any
40 commits in HEAD which introduce the same textual changes as a commit in
41 HEAD..<upstream> are omitted (i.e., a patch already accepted upstream
42 with a different commit message or timestamp will be skipped).
43 It is possible that a merge failure will prevent this process from
45 being completely automatic. You will have to resolve any such merge
46 failure and run git rebase --continue. Another option is to bypass the
47 commit that caused the merge failure with git rebase --skip. To check
48 out the original <branch> and remove the .git/rebase-apply working
49 files, use the command git rebase --abort instead.
50 Assume the following history exists and the current branch is "topic":
52 A---B---C topic
54 /
55 D---E---F---G master
56 From this point, the result of either of the following commands:
58 git rebase master
60 git rebase master topic
61 would be:
63 A'--B'--C' topic
65 /
66 D---E---F---G master
67 NOTE: The latter form is just a short-hand of git checkout topic
69 followed by git rebase master. When rebase exits topic will remain the
70 checked-out branch.
71 If the upstream branch already contains a change you have made (e.g.,
73 because you mailed a patch which was applied upstream), then that
74 commit will be skipped. For example, running git rebase master on the
75 following history (in which A' and A introduce the same set of changes,
76 but have different committer information):
77 A---B---C topic
79 /
80 D---E---A'---F master
81 will result in:
83 B'---C' topic
85 /
86 D---E---A'---F master
87 Here is how you would transplant a topic branch based on one branch to
89 another, to pretend that you forked the topic branch from the latter
90 branch, using rebase --onto.
91 First let’s assume your topic is based on branch next. For example, a
93 feature developed in topic depends on some functionality which is found
94 in next.
95 o---o---o---o---o master
97 \
98 o---o---o---o---o next
99 \
100 o---o---o topic
101 We want to make topic forked from branch master; for example, because
103 the functionality on which topic depends was merged into the more
104 stable master branch. We want our tree to look like this:
105 o---o---o---o---o master
107 | \
108 | o'--o'--o' topic
109 \
110 o---o---o---o---o next
111 We can get this using the following command:
113 git rebase --onto master next topic
115 Another example of --onto option is to rebase part of a branch. If we
117 have the following situation:
118 H---I---J topicB
120 /
121 E---F---G topicA
122 /
123 A---B---C---D master
124 then the command
126 git rebase --onto master topicA topicB
128 would result in:
130 H'--I'--J' topicB
132 /
133 | E---F---G topicA
134 |/
135 A---B---C---D master
136 This is useful when topicB does not depend on topicA.
138 A range of commits could also be removed with rebase. If we have the
140 following situation:
141 E---F---G---H---I---J topicA
143 then the command
145 git rebase --onto topicA~5 topicA~3 topicA
147 would result in the removal of commits F and G:
149 E---H'---I'---J' topicA
151 This is useful if F and G were flawed in some way, or should not be
153 part of topicA. Note that the argument to --onto and the <upstream>
154 parameter can be any valid commit-ish.
155 In case of conflict, git rebase will stop at the first problematic
157 commit and leave conflict markers in the tree. You can use git diff to
158 locate the markers (<<<<<<) and make edits to resolve the conflict. For
159 each file you edit, you need to tell Git that the conflict has been
160 resolved, typically this would be done with
161 git add <filename>
163 After resolving the conflict manually and updating the index with the
165 desired resolution, you can continue the rebasing process with
166 git rebase --continue
168 Alternatively, you can undo the git rebase with
170 git rebase --abort
172
174 rebase.useBuiltin
175 Unused configuration variable. Used in Git versions 2.20 and 2.21
176 as an escape hatch to enable the legacy shellscript implementation
177 of rebase. Now the built-in rewrite of it in C is always used.
178 Setting this will emit a warning, to alert any remaining users that
179 setting this now does nothing.
180 rebase.backend
182 Default backend to use for rebasing. Possible choices are apply or
183 merge. In the future, if the merge backend gains all remaining
184 capabilities of the apply backend, this setting may become unused.
185 rebase.stat
187 Whether to show a diffstat of what changed upstream since the last
188 rebase. False by default.
189 rebase.autoSquash
191 If set to true enable --autosquash option by default.
192 rebase.autoStash
194 When set to true, automatically create a temporary stash entry
195 before the operation begins, and apply it after the operation ends.
196 This means that you can run rebase on a dirty worktree. However,
197 use with care: the final stash application after a successful
198 rebase might result in non-trivial conflicts. This option can be
199 overridden by the --no-autostash and --autostash options of git-
200 rebase(1). Defaults to false.
201 rebase.missingCommitsCheck
203 If set to "warn", git rebase -i will print a warning if some
204 commits are removed (e.g. a line was deleted), however the rebase
205 will still proceed. If set to "error", it will print the previous
206 warning and stop the rebase, git rebase --edit-todo can then be
207 used to correct the error. If set to "ignore", no checking is done.
208 To drop a commit without warning or error, use the drop command in
209 the todo list. Defaults to "ignore".
210 rebase.instructionFormat
212 A format string, as specified in git-log(1), to be used for the
213 todo list during an interactive rebase. The format will
214 automatically have the long commit hash prepended to the format.
215 rebase.abbreviateCommands
217 If set to true, git rebase will use abbreviated command names in
218 the todo list resulting in something like this:
219 p deadbee The oneline of the commit
221 p fa1afe1 The oneline of the next commit
222 ...
223 instead of:
225 pick deadbee The oneline of the commit
227 pick fa1afe1 The oneline of the next commit
228 ...
229 Defaults to false.
231 rebase.rescheduleFailedExec
233 Automatically reschedule exec commands that failed. This only makes
234 sense in interactive mode (or when an --exec option was provided).
235 This is the same as specifying the --reschedule-failed-exec option.
236 rebase.forkPoint
238 If set to false set --no-fork-point option by default.
239 sequence.editor
241 Text editor used by git rebase -i for editing the rebase
242 instruction file. The value is meant to be interpreted by the shell
243 when it is used. It can be overridden by the GIT_SEQUENCE_EDITOR
244 environment variable. When not configured the default commit
245 message editor is used instead.
246
248 --onto <newbase>
249 Starting point at which to create the new commits. If the --onto
250 option is not specified, the starting point is <upstream>. May be
251 any valid commit, and not just an existing branch name.
252 As a special case, you may use "A...B" as a shortcut for the merge
254 base of A and B if there is exactly one merge base. You can leave
255 out at most one of A and B, in which case it defaults to HEAD.
256 --keep-base
258 Set the starting point at which to create the new commits to the
259 merge base of <upstream> <branch>. Running git rebase --keep-base
260 <upstream> <branch> is equivalent to running git rebase --onto
261 <upstream>... <upstream>.
262 This option is useful in the case where one is developing a feature
264 on top of an upstream branch. While the feature is being worked on,
265 the upstream branch may advance and it may not be the best idea to
266 keep rebasing on top of the upstream but to keep the base commit
267 as-is.
268 Although both this option and --fork-point find the merge base
270 between <upstream> and <branch>, this option uses the merge base as
271 the starting point on which new commits will be created, whereas
272 --fork-point uses the merge base to determine the set of commits
273 which will be rebased.
274 See also INCOMPATIBLE OPTIONS below.
276 <upstream>
278 Upstream branch to compare against. May be any valid commit, not
279 just an existing branch name. Defaults to the configured upstream
280 for the current branch.
281 <branch>
283 Working branch; defaults to HEAD.
284 --continue
286 Restart the rebasing process after having resolved a merge
287 conflict.
288 --abort
290 Abort the rebase operation and reset HEAD to the original branch.
291 If <branch> was provided when the rebase operation was started,
292 then HEAD will be reset to <branch>. Otherwise HEAD will be reset
293 to where it was when the rebase operation was started.
294 --quit
296 Abort the rebase operation but HEAD is not reset back to the
297 original branch. The index and working tree are also left unchanged
298 as a result. If a temporary stash entry was created using
299 --autostash, it will be saved to the stash list.
300 --apply
302 Use applying strategies to rebase (calling git-am internally). This
303 option may become a no-op in the future once the merge backend
304 handles everything the apply one does.
305 See also INCOMPATIBLE OPTIONS below.
307 --empty={drop,keep,ask}
309 How to handle commits that are not empty to start and are not clean
310 cherry-picks of any upstream commit, but which become empty after
311 rebasing (because they contain a subset of already upstream
312 changes). With drop (the default), commits that become empty are
313 dropped. With keep, such commits are kept. With ask (implied by
314 --interactive), the rebase will halt when an empty commit is
315 applied allowing you to choose whether to drop it, edit files more,
316 or just commit the empty changes. Other options, like --exec, will
317 use the default of drop unless -i/--interactive is explicitly
318 specified.
319 Note that commits which start empty are kept (unless
321 --no-keep-empty is specified), and commits which are clean
322 cherry-picks (as determined by git log --cherry-mark ...) are
323 detected and dropped as a preliminary step (unless
324 --reapply-cherry-picks is passed).
325 See also INCOMPATIBLE OPTIONS below.
327 --no-keep-empty, --keep-empty
329 Do not keep commits that start empty before the rebase (i.e. that
330 do not change anything from its parent) in the result. The default
331 is to keep commits which start empty, since creating such commits
332 requires passing the --allow-empty override flag to git commit,
333 signifying that a user is very intentionally creating such a commit
334 and thus wants to keep it.
335 Usage of this flag will probably be rare, since you can get rid of
337 commits that start empty by just firing up an interactive rebase
338 and removing the lines corresponding to the commits you don’t want.
339 This flag exists as a convenient shortcut, such as for cases where
340 external tools generate many empty commits and you want them all
341 removed.
342 For commits which do not start empty but become empty after
344 rebasing, see the --empty flag.
345 See also INCOMPATIBLE OPTIONS below.
347 --reapply-cherry-picks, --no-reapply-cherry-picks
349 Reapply all clean cherry-picks of any upstream commit instead of
350 preemptively dropping them. (If these commits then become empty
351 after rebasing, because they contain a subset of already upstream
352 changes, the behavior towards them is controlled by the --empty
353 flag.)
354 By default (or if --no-reapply-cherry-picks is given), these
356 commits will be automatically dropped. Because this necessitates
357 reading all upstream commits, this can be expensive in repos with a
358 large number of upstream commits that need to be read.
359 --reapply-cherry-picks allows rebase to forgo reading all upstream
361 commits, potentially improving performance.
362 See also INCOMPATIBLE OPTIONS below.
364 --allow-empty-message
366 No-op. Rebasing commits with an empty message used to fail and this
367 option would override that behavior, allowing commits with empty
368 messages to be rebased. Now commits with an empty message do not
369 cause rebasing to halt.
370 See also INCOMPATIBLE OPTIONS below.
372 --skip
374 Restart the rebasing process by skipping the current patch.
375 --edit-todo
377 Edit the todo list during an interactive rebase.
378 --show-current-patch
380 Show the current patch in an interactive rebase or when rebase is
381 stopped because of conflicts. This is the equivalent of git show
382 REBASE_HEAD.
383 -m, --merge
385 Use merging strategies to rebase. When the recursive (default)
386 merge strategy is used, this allows rebase to be aware of renames
387 on the upstream side. This is the default.
388 Note that a rebase merge works by replaying each commit from the
390 working branch on top of the <upstream> branch. Because of this,
391 when a merge conflict happens, the side reported as ours is the
392 so-far rebased series, starting with <upstream>, and theirs is the
393 working branch. In other words, the sides are swapped.
394 See also INCOMPATIBLE OPTIONS below.
396 -s <strategy>, --strategy=<strategy>
398 Use the given merge strategy. If there is no -s option git
399 merge-recursive is used instead. This implies --merge.
400 Because git rebase replays each commit from the working branch on
402 top of the <upstream> branch using the given strategy, using the
403 ours strategy simply empties all patches from the <branch>, which
404 makes little sense.
405 See also INCOMPATIBLE OPTIONS below.
407 -X <strategy-option>, --strategy-option=<strategy-option>
409 Pass the <strategy-option> through to the merge strategy. This
410 implies --merge and, if no strategy has been specified, -s
411 recursive. Note the reversal of ours and theirs as noted above for
412 the -m option.
413 See also INCOMPATIBLE OPTIONS below.
415 --rerere-autoupdate, --no-rerere-autoupdate
417 Allow the rerere mechanism to update the index with the result of
418 auto-conflict resolution if possible.
419 -S[<keyid>], --gpg-sign[=<keyid>], --no-gpg-sign
421 GPG-sign commits. The keyid argument is optional and defaults to
422 the committer identity; if specified, it must be stuck to the
423 option without a space. --no-gpg-sign is useful to countermand
424 both commit.gpgSign configuration variable, and earlier --gpg-sign.
425 -q, --quiet
427 Be quiet. Implies --no-stat.
428 -v, --verbose
430 Be verbose. Implies --stat.
431 --stat
433 Show a diffstat of what changed upstream since the last rebase. The
434 diffstat is also controlled by the configuration option
435 rebase.stat.
436 -n, --no-stat
438 Do not show a diffstat as part of the rebase process.
439 --no-verify
441 This option bypasses the pre-rebase hook. See also githooks(5).
442 --verify
444 Allows the pre-rebase hook to run, which is the default. This
445 option can be used to override --no-verify. See also githooks(5).
446 -C<n>
448 Ensure at least <n> lines of surrounding context match before and
449 after each change. When fewer lines of surrounding context exist
450 they all must match. By default no context is ever ignored. Implies
451 --apply.
452 See also INCOMPATIBLE OPTIONS below.
454 --no-ff, --force-rebase, -f
456 Individually replay all rebased commits instead of fast-forwarding
457 over the unchanged ones. This ensures that the entire history of
458 the rebased branch is composed of new commits.
459 You may find this helpful after reverting a topic branch merge, as
461 this option recreates the topic branch with fresh commits so it can
462 be remerged successfully without needing to "revert the reversion"
463 (see the revert-a-faulty-merge How-To[1] for details).
464 --fork-point, --no-fork-point
466 Use reflog to find a better common ancestor between <upstream> and
467 <branch> when calculating which commits have been introduced by
468 <branch>.
469 When --fork-point is active, fork_point will be used instead of
471 <upstream> to calculate the set of commits to rebase, where
472 fork_point is the result of git merge-base --fork-point <upstream>
473 <branch> command (see git-merge-base(1)). If fork_point ends up
474 being empty, the <upstream> will be used as a fallback.
475 If <upstream> is given on the command line, then the default is
477 --no-fork-point, otherwise the default is --fork-point.
478 If your branch was based on <upstream> but <upstream> was rewound
480 and your branch contains commits which were dropped, this option
481 can be used with --keep-base in order to drop those commits from
482 your branch.
483 See also INCOMPATIBLE OPTIONS below.
485 --ignore-whitespace
487 Ignore whitespace differences when trying to reconcile differences.
488 Currently, each backend implements an approximation of this
489 behavior:
490 apply backend: When applying a patch, ignore changes in whitespace
492 in context lines. Unfortunately, this means that if the "old" lines
493 being replaced by the patch differ only in whitespace from the
494 existing file, you will get a merge conflict instead of a
495 successful patch application.
496 merge backend: Treat lines with only whitespace changes as
498 unchanged when merging. Unfortunately, this means that any patch
499 hunks that were intended to modify whitespace and nothing else will
500 be dropped, even if the other side had no changes that conflicted.
501 --whitespace=<option>
503 This flag is passed to the git apply program (see git-apply(1))
504 that applies the patch. Implies --apply.
505 See also INCOMPATIBLE OPTIONS below.
507 --committer-date-is-author-date
509 Instead of using the current time as the committer date, use the
510 author date of the commit being rebased as the committer date. This
511 option implies --force-rebase.
512 --ignore-date, --reset-author-date
514 Instead of using the author date of the original commit, use the
515 current time as the author date of the rebased commit. This option
516 implies --force-rebase.
517 See also INCOMPATIBLE OPTIONS below.
519 --signoff
521 Add a Signed-off-by trailer to all the rebased commits. Note that
522 if --interactive is given then only commits marked to be picked,
523 edited or reworded will have the trailer added.
524 See also INCOMPATIBLE OPTIONS below.
526 -i, --interactive
528 Make a list of the commits which are about to be rebased. Let the
529 user edit that list before rebasing. This mode can also be used to
530 split commits (see SPLITTING COMMITS below).
531 The commit list format can be changed by setting the configuration
533 option rebase.instructionFormat. A customized instruction format
534 will automatically have the long commit hash prepended to the
535 format.
536 See also INCOMPATIBLE OPTIONS below.
538 -r, --rebase-merges[=(rebase-cousins|no-rebase-cousins)]
540 By default, a rebase will simply drop merge commits from the todo
541 list, and put the rebased commits into a single, linear branch.
542 With --rebase-merges, the rebase will instead try to preserve the
543 branching structure within the commits that are to be rebased, by
544 recreating the merge commits. Any resolved merge conflicts or
545 manual amendments in these merge commits will have to be
546 resolved/re-applied manually.
547 By default, or when no-rebase-cousins was specified, commits which
549 do not have <upstream> as direct ancestor will keep their original
550 branch point, i.e. commits that would be excluded by git-log(1)'s
551 --ancestry-path option will keep their original ancestry by
552 default. If the rebase-cousins mode is turned on, such commits are
553 instead rebased onto <upstream> (or <onto>, if specified).
554 The --rebase-merges mode is similar in spirit to the deprecated
556 --preserve-merges but works with interactive rebases, where commits
557 can be reordered, inserted and dropped at will.
558 It is currently only possible to recreate the merge commits using
560 the recursive merge strategy; Different merge strategies can be
561 used only via explicit exec git merge -s <strategy> [...]
562 commands.
563 See also REBASING MERGES and INCOMPATIBLE OPTIONS below.
565 -p, --preserve-merges
567 [DEPRECATED: use --rebase-merges instead] Recreate merge commits
568 instead of flattening the history by replaying commits a merge
569 commit introduces. Merge conflict resolutions or manual amendments
570 to merge commits are not preserved.
571 This uses the --interactive machinery internally, but combining it
573 with the --interactive option explicitly is generally not a good
574 idea unless you know what you are doing (see BUGS below).
575 See also INCOMPATIBLE OPTIONS below.
577 -x <cmd>, --exec <cmd>
579 Append "exec <cmd>" after each line creating a commit in the final
580 history. <cmd> will be interpreted as one or more shell commands.
581 Any command that fails will interrupt the rebase, with exit code 1.
582 You may execute several commands by either using one instance of
584 --exec with several commands:
585 git rebase -i --exec "cmd1 && cmd2 && ..."
587 or by giving more than one --exec:
589 git rebase -i --exec "cmd1" --exec "cmd2" --exec ...
591 If --autosquash is used, "exec" lines will not be appended for the
593 intermediate commits, and will only appear at the end of each
594 squash/fixup series.
595 This uses the --interactive machinery internally, but it can be run
597 without an explicit --interactive.
598 See also INCOMPATIBLE OPTIONS below.
600 --root
602 Rebase all commits reachable from <branch>, instead of limiting
603 them with an <upstream>. This allows you to rebase the root
604 commit(s) on a branch. When used with --onto, it will skip changes
605 already contained in <newbase> (instead of <upstream>) whereas
606 without --onto it will operate on every change. When used together
607 with both --onto and --preserve-merges, all root commits will be
608 rewritten to have <newbase> as parent instead.
609 See also INCOMPATIBLE OPTIONS below.
611 --autosquash, --no-autosquash
613 When the commit log message begins with "squash! ..." (or "fixup!
614 ..."), and there is already a commit in the todo list that matches
615 the same ..., automatically modify the todo list of rebase -i so
616 that the commit marked for squashing comes right after the commit
617 to be modified, and change the action of the moved commit from pick
618 to squash (or fixup). A commit matches the ... if the commit
619 subject matches, or if the ... refers to the commit’s hash. As a
620 fall-back, partial matches of the commit subject work, too. The
621 recommended way to create fixup/squash commits is by using the
622 --fixup/--squash options of git-commit(1).
623 If the --autosquash option is enabled by default using the
625 configuration variable rebase.autoSquash, this option can be used
626 to override and disable this setting.
627 See also INCOMPATIBLE OPTIONS below.
629 --autostash, --no-autostash
631 Automatically create a temporary stash entry before the operation
632 begins, and apply it after the operation ends. This means that you
633 can run rebase on a dirty worktree. However, use with care: the
634 final stash application after a successful rebase might result in
635 non-trivial conflicts.
636 --reschedule-failed-exec, --no-reschedule-failed-exec
638 Automatically reschedule exec commands that failed. This only makes
639 sense in interactive mode (or when an --exec option was provided).
640
642 The following options:
643 • --apply
645 • --whitespace
647 • -C
649 are incompatible with the following options:
651 • --merge
653 • --strategy
655 • --strategy-option
657 • --allow-empty-message
659 • --[no-]autosquash
661 • --rebase-merges
663 • --preserve-merges
665 • --interactive
667 • --exec
669 • --no-keep-empty
671 • --empty=
673 • --reapply-cherry-picks
675 • --edit-todo
677 • --root when used in combination with --onto
679 In addition, the following pairs of options are incompatible:
681 • --preserve-merges and --interactive
683 • --preserve-merges and --signoff
685 • --preserve-merges and --rebase-merges
687 • --preserve-merges and --empty=
689 • --preserve-merges and --ignore-whitespace
691 • --preserve-merges and --committer-date-is-author-date
693 • --preserve-merges and --ignore-date
695 • --keep-base and --onto
697 • --keep-base and --root
699 • --fork-point and --root
701
703 git rebase has two primary backends: apply and merge. (The apply
704 backend used to be known as the am backend, but the name led to
705 confusion as it looks like a verb instead of a noun. Also, the merge
706 backend used to be known as the interactive backend, but it is now used
707 for non-interactive cases as well. Both were renamed based on
708 lower-level functionality that underpinned each.) There are some subtle
709 differences in how these two backends behave:
710 Empty commits
712 The apply backend unfortunately drops intentionally empty commits, i.e.
713 commits that started empty, though these are rare in practice. It also
714 drops commits that become empty and has no option for controlling this
715 behavior.
716 The merge backend keeps intentionally empty commits by default (though
718 with -i they are marked as empty in the todo list editor, or they can
719 be dropped automatically with --no-keep-empty).
720 Similar to the apply backend, by default the merge backend drops
722 commits that become empty unless -i/--interactive is specified (in
723 which case it stops and asks the user what to do). The merge backend
724 also has an --empty={drop,keep,ask} option for changing the behavior of
725 handling commits that become empty.
726 Directory rename detection
728 Due to the lack of accurate tree information (arising from constructing
729 fake ancestors with the limited information available in patches),
730 directory rename detection is disabled in the apply backend. Disabled
731 directory rename detection means that if one side of history renames a
732 directory and the other adds new files to the old directory, then the
733 new files will be left behind in the old directory without any warning
734 at the time of rebasing that you may want to move these files into the
735 new directory.
736 Directory rename detection works with the merge backend to provide you
738 warnings in such cases.
739 Context
741 The apply backend works by creating a sequence of patches (by calling
742 format-patch internally), and then applying the patches in sequence
743 (calling am internally). Patches are composed of multiple hunks, each
744 with line numbers, a context region, and the actual changes. The line
745 numbers have to be taken with some fuzz, since the other side will
746 likely have inserted or deleted lines earlier in the file. The context
747 region is meant to help find how to adjust the line numbers in order to
748 apply the changes to the right lines. However, if multiple areas of the
749 code have the same surrounding lines of context, the wrong one can be
750 picked. There are real-world cases where this has caused commits to be
751 reapplied incorrectly with no conflicts reported. Setting diff.context
752 to a larger value may prevent such types of problems, but increases the
753 chance of spurious conflicts (since it will require more lines of
754 matching context to apply).
755 The merge backend works with a full copy of each relevant file,
757 insulating it from these types of problems.
758 Labelling of conflicts markers
760 When there are content conflicts, the merge machinery tries to annotate
761 each side’s conflict markers with the commits where the content came
762 from. Since the apply backend drops the original information about the
763 rebased commits and their parents (and instead generates new fake
764 commits based off limited information in the generated patches), those
765 commits cannot be identified; instead it has to fall back to a commit
766 summary. Also, when merge.conflictStyle is set to diff3, the apply
767 backend will use "constructed merge base" to label the content from the
768 merge base, and thus provide no information about the merge base commit
769 whatsoever.
770 The merge backend works with the full commits on both sides of history
772 and thus has no such limitations.
773 Hooks
775 The apply backend has not traditionally called the post-commit hook,
776 while the merge backend has. Both have called the post-checkout hook,
777 though the merge backend has squelched its output. Further, both
778 backends only call the post-checkout hook with the starting point
779 commit of the rebase, not the intermediate commits nor the final
780 commit. In each case, the calling of these hooks was by accident of
781 implementation rather than by design (both backends were originally
782 implemented as shell scripts and happened to invoke other commands like
783 git checkout or git commit that would call the hooks). Both backends
784 should have the same behavior, though it is not entirely clear which,
785 if any, is correct. We will likely make rebase stop calling either of
786 these hooks in the future.
787 Interruptability
789 The apply backend has safety problems with an ill-timed interrupt; if
790 the user presses Ctrl-C at the wrong time to try to abort the rebase,
791 the rebase can enter a state where it cannot be aborted with a
792 subsequent git rebase --abort. The merge backend does not appear to
793 suffer from the same shortcoming. (See
794 https://lore.kernel.org/git/20200207132152.GC2868@szeder.dev/ for
795 details.)
796 Commit Rewording
798 When a conflict occurs while rebasing, rebase stops and asks the user
799 to resolve. Since the user may need to make notable changes while
800 resolving conflicts, after conflicts are resolved and the user has run
801 git rebase --continue, the rebase should open an editor and ask the
802 user to update the commit message. The merge backend does this, while
803 the apply backend blindly applies the original commit message.
804 Miscellaneous differences
806 There are a few more behavioral differences that most folks would
807 probably consider inconsequential but which are mentioned for
808 completeness:
809 • Reflog: The two backends will use different wording when describing
811 the changes made in the reflog, though both will make use of the
812 word "rebase".
813 • Progress, informational, and error messages: The two backends
815 provide slightly different progress and informational messages.
816 Also, the apply backend writes error messages (such as "Your files
817 would be overwritten...") to stdout, while the merge backend writes
818 them to stderr.
819 • State directories: The two backends keep their state in different
821 directories under .git/
822
824 The merge mechanism (git merge and git pull commands) allows the
825 backend merge strategies to be chosen with -s option. Some strategies
826 can also take their own options, which can be passed by giving
827 -X<option> arguments to git merge and/or git pull.
828 resolve
830 This can only resolve two heads (i.e. the current branch and
831 another branch you pulled from) using a 3-way merge algorithm. It
832 tries to carefully detect criss-cross merge ambiguities and is
833 considered generally safe and fast.
834 recursive
836 This can only resolve two heads using a 3-way merge algorithm. When
837 there is more than one common ancestor that can be used for 3-way
838 merge, it creates a merged tree of the common ancestors and uses
839 that as the reference tree for the 3-way merge. This has been
840 reported to result in fewer merge conflicts without causing
841 mismerges by tests done on actual merge commits taken from Linux
842 2.6 kernel development history. Additionally this can detect and
843 handle merges involving renames, but currently cannot make use of
844 detected copies. This is the default merge strategy when pulling or
845 merging one branch.
846 The recursive strategy can take the following options:
848 ours
850 This option forces conflicting hunks to be auto-resolved
851 cleanly by favoring our version. Changes from the other tree
852 that do not conflict with our side are reflected in the merge
853 result. For a binary file, the entire contents are taken from
854 our side.
855 This should not be confused with the ours merge strategy, which
857 does not even look at what the other tree contains at all. It
858 discards everything the other tree did, declaring our history
859 contains all that happened in it.
860 theirs
862 This is the opposite of ours; note that, unlike ours, there is
863 no theirs merge strategy to confuse this merge option with.
864 patience
866 With this option, merge-recursive spends a little extra time to
867 avoid mismerges that sometimes occur due to unimportant
868 matching lines (e.g., braces from distinct functions). Use this
869 when the branches to be merged have diverged wildly. See also
870 git-diff(1) --patience.
871 diff-algorithm=[patience|minimal|histogram|myers]
873 Tells merge-recursive to use a different diff algorithm, which
874 can help avoid mismerges that occur due to unimportant matching
875 lines (such as braces from distinct functions). See also git-
876 diff(1) --diff-algorithm.
877 ignore-space-change, ignore-all-space, ignore-space-at-eol,
879 ignore-cr-at-eol
880 Treats lines with the indicated type of whitespace change as
881 unchanged for the sake of a three-way merge. Whitespace changes
882 mixed with other changes to a line are not ignored. See also
883 git-diff(1) -b, -w, --ignore-space-at-eol, and
884 --ignore-cr-at-eol.
885 • If their version only introduces whitespace changes to a
887 line, our version is used;
888 • If our version introduces whitespace changes but their
890 version includes a substantial change, their version is
891 used;
892 • Otherwise, the merge proceeds in the usual way.
894 renormalize
896 This runs a virtual check-out and check-in of all three stages
897 of a file when resolving a three-way merge. This option is
898 meant to be used when merging branches with different clean
899 filters or end-of-line normalization rules. See "Merging
900 branches with differing checkin/checkout attributes" in
901 gitattributes(5) for details.
902 no-renormalize
904 Disables the renormalize option. This overrides the
905 merge.renormalize configuration variable.
906 no-renames
908 Turn off rename detection. This overrides the merge.renames
909 configuration variable. See also git-diff(1) --no-renames.
910 find-renames[=<n>]
912 Turn on rename detection, optionally setting the similarity
913 threshold. This is the default. This overrides the
914 merge.renames configuration variable. See also git-diff(1)
915 --find-renames.
916 rename-threshold=<n>
918 Deprecated synonym for find-renames=<n>.
919 subtree[=<path>]
921 This option is a more advanced form of subtree strategy, where
922 the strategy makes a guess on how two trees must be shifted to
923 match with each other when merging. Instead, the specified path
924 is prefixed (or stripped from the beginning) to make the shape
925 of two trees to match.
926 octopus
928 This resolves cases with more than two heads, but refuses to do a
929 complex merge that needs manual resolution. It is primarily meant
930 to be used for bundling topic branch heads together. This is the
931 default merge strategy when pulling or merging more than one
932 branch.
933 ours
935 This resolves any number of heads, but the resulting tree of the
936 merge is always that of the current branch head, effectively
937 ignoring all changes from all other branches. It is meant to be
938 used to supersede old development history of side branches. Note
939 that this is different from the -Xours option to the recursive
940 merge strategy.
941 subtree
943 This is a modified recursive strategy. When merging trees A and B,
944 if B corresponds to a subtree of A, B is first adjusted to match
945 the tree structure of A, instead of reading the trees at the same
946 level. This adjustment is also done to the common ancestor tree.
947 With the strategies that use 3-way merge (including the default,
949 recursive), if a change is made on both branches, but later reverted on
950 one of the branches, that change will be present in the merged result;
951 some people find this behavior confusing. It occurs because only the
952 heads and the merge base are considered when performing a merge, not
953 the individual commits. The merge algorithm therefore considers the
954 reverted change as no change at all, and substitutes the changed
955 version instead.
956
958 You should understand the implications of using git rebase on a
959 repository that you share. See also RECOVERING FROM UPSTREAM REBASE
960 below.
961 When the git-rebase command is run, it will first execute a
963 "pre-rebase" hook if one exists. You can use this hook to do sanity
964 checks and reject the rebase if it isn’t appropriate. Please see the
965 template pre-rebase hook script for an example.
966 Upon completion, <branch> will be the current branch.
968
970 Rebasing interactively means that you have a chance to edit the commits
971 which are rebased. You can reorder the commits, and you can remove them
972 (weeding out bad or otherwise unwanted patches).
973 The interactive mode is meant for this type of workflow:
975 1. have a wonderful idea
977 2. hack on the code
979 3. prepare a series for submission
981 4. submit
983 where point 2. consists of several instances of
985 a) regular use
987 1. finish something worthy of a commit
989 2. commit
991 b) independent fixup
993 1. realize that something does not work
995 2. fix that
997 3. commit it
999 Sometimes the thing fixed in b.2. cannot be amended to the not-quite
1001 perfect commit it fixes, because that commit is buried deeply in a
1002 patch series. That is exactly what interactive rebase is for: use it
1003 after plenty of "a"s and "b"s, by rearranging and editing commits, and
1004 squashing multiple commits into one.
1005 Start it with the last commit you want to retain as-is:
1007 git rebase -i <after-this-commit>
1009 An editor will be fired up with all the commits in your current branch
1011 (ignoring merge commits), which come after the given commit. You can
1012 reorder the commits in this list to your heart’s content, and you can
1013 remove them. The list looks more or less like this:
1014 pick deadbee The oneline of this commit
1016 pick fa1afe1 The oneline of the next commit
1017 ...
1018 The oneline descriptions are purely for your pleasure; git rebase will
1020 not look at them but at the commit names ("deadbee" and "fa1afe1" in
1021 this example), so do not delete or edit the names.
1022 By replacing the command "pick" with the command "edit", you can tell
1024 git rebase to stop after applying that commit, so that you can edit the
1025 files and/or the commit message, amend the commit, and continue
1026 rebasing.
1027 To interrupt the rebase (just like an "edit" command would do, but
1029 without cherry-picking any commit first), use the "break" command.
1030 If you just want to edit the commit message for a commit, replace the
1032 command "pick" with the command "reword".
1033 To drop a commit, replace the command "pick" with "drop", or just
1035 delete the matching line.
1036 If you want to fold two or more commits into one, replace the command
1038 "pick" for the second and subsequent commits with "squash" or "fixup".
1039 If the commits had different authors, the folded commit will be
1040 attributed to the author of the first commit. The suggested commit
1041 message for the folded commit is the concatenation of the commit
1042 messages of the first commit and of those with the "squash" command,
1043 but omits the commit messages of commits with the "fixup" command.
1044 git rebase will stop when "pick" has been replaced with "edit" or when
1046 a command fails due to merge errors. When you are done editing and/or
1047 resolving conflicts you can continue with git rebase --continue.
1048 For example, if you want to reorder the last 5 commits, such that what
1050 was HEAD~4 becomes the new HEAD. To achieve that, you would call git
1051 rebase like this:
1052 $ git rebase -i HEAD~5
1054 And move the first patch to the end of the list.
1056 You might want to recreate merge commits, e.g. if you have a history
1058 like this:
1059 X
1061 \
1062 A---M---B
1063 /
1064 ---o---O---P---Q
1065 Suppose you want to rebase the side branch starting at "A" to "Q". Make
1067 sure that the current HEAD is "B", and call
1068 $ git rebase -i -r --onto Q O
1070 Reordering and editing commits usually creates untested intermediate
1072 steps. You may want to check that your history editing did not break
1073 anything by running a test, or at least recompiling at intermediate
1074 points in history by using the "exec" command (shortcut "x"). You may
1075 do so by creating a todo list like this one:
1076 pick deadbee Implement feature XXX
1078 fixup f1a5c00 Fix to feature XXX
1079 exec make
1080 pick c0ffeee The oneline of the next commit
1081 edit deadbab The oneline of the commit after
1082 exec cd subdir; make test
1083 ...
1084 The interactive rebase will stop when a command fails (i.e. exits with
1086 non-0 status) to give you an opportunity to fix the problem. You can
1087 continue with git rebase --continue.
1088 The "exec" command launches the command in a shell (the one specified
1090 in $SHELL, or the default shell if $SHELL is not set), so you can use
1091 shell features (like "cd", ">", ";" ...). The command is run from the
1092 root of the working tree.
1093 $ git rebase -i --exec "make test"
1095 This command lets you check that intermediate commits are compilable.
1097 The todo list becomes like that:
1098 pick 5928aea one
1100 exec make test
1101 pick 04d0fda two
1102 exec make test
1103 pick ba46169 three
1104 exec make test
1105 pick f4593f9 four
1106 exec make test
1107
1109 In interactive mode, you can mark commits with the action "edit".
1110 However, this does not necessarily mean that git rebase expects the
1111 result of this edit to be exactly one commit. Indeed, you can undo the
1112 commit, or you can add other commits. This can be used to split a
1113 commit into two:
1114 • Start an interactive rebase with git rebase -i <commit>^, where
1116 <commit> is the commit you want to split. In fact, any commit range
1117 will do, as long as it contains that commit.
1118 • Mark the commit you want to split with the action "edit".
1120 • When it comes to editing that commit, execute git reset HEAD^. The
1122 effect is that the HEAD is rewound by one, and the index follows
1123 suit. However, the working tree stays the same.
1124 • Now add the changes to the index that you want to have in the first
1126 commit. You can use git add (possibly interactively) or git gui (or
1127 both) to do that.
1128 • Commit the now-current index with whatever commit message is
1130 appropriate now.
1131 • Repeat the last two steps until your working tree is clean.
1133 • Continue the rebase with git rebase --continue.
1135 If you are not absolutely sure that the intermediate revisions are
1137 consistent (they compile, pass the testsuite, etc.) you should use git
1138 stash to stash away the not-yet-committed changes after each commit,
1139 test, and amend the commit if fixes are necessary.
1140
1142 Rebasing (or any other form of rewriting) a branch that others have
1143 based work on is a bad idea: anyone downstream of it is forced to
1144 manually fix their history. This section explains how to do the fix
1145 from the downstream’s point of view. The real fix, however, would be to
1146 avoid rebasing the upstream in the first place.
1147 To illustrate, suppose you are in a situation where someone develops a
1149 subsystem branch, and you are working on a topic that is dependent on
1150 this subsystem. You might end up with a history like the following:
1151 o---o---o---o---o---o---o---o master
1153 \
1154 o---o---o---o---o subsystem
1155 \
1156 *---*---* topic
1157 If subsystem is rebased against master, the following happens:
1159 o---o---o---o---o---o---o---o master
1161 \ \
1162 o---o---o---o---o o'--o'--o'--o'--o' subsystem
1163 \
1164 *---*---* topic
1165 If you now continue development as usual, and eventually merge topic to
1167 subsystem, the commits from subsystem will remain duplicated forever:
1168 o---o---o---o---o---o---o---o master
1170 \ \
1171 o---o---o---o---o o'--o'--o'--o'--o'--M subsystem
1172 \ /
1173 *---*---*-..........-*--* topic
1174 Such duplicates are generally frowned upon because they clutter up
1176 history, making it harder to follow. To clean things up, you need to
1177 transplant the commits on topic to the new subsystem tip, i.e., rebase
1178 topic. This becomes a ripple effect: anyone downstream from topic is
1179 forced to rebase too, and so on!
1180 There are two kinds of fixes, discussed in the following subsections:
1182 Easy case: The changes are literally the same.
1184 This happens if the subsystem rebase was a simple rebase and had no
1185 conflicts.
1186 Hard case: The changes are not the same.
1188 This happens if the subsystem rebase had conflicts, or used
1189 --interactive to omit, edit, squash, or fixup commits; or if the
1190 upstream used one of commit --amend, reset, or a full history
1191 rewriting command like filter-repo[2].
1192 The easy case
1194 Only works if the changes (patch IDs based on the diff contents) on
1195 subsystem are literally the same before and after the rebase subsystem
1196 did.
1197 In that case, the fix is easy because git rebase knows to skip changes
1199 that are already present in the new upstream (unless
1200 --reapply-cherry-picks is given). So if you say (assuming you’re on
1201 topic)
1202 $ git rebase subsystem
1204 you will end up with the fixed history
1206 o---o---o---o---o---o---o---o master
1208 \
1209 o'--o'--o'--o'--o' subsystem
1210 \
1211 *---*---* topic
1212 The hard case
1214 Things get more complicated if the subsystem changes do not exactly
1215 correspond to the ones before the rebase.
1216 Note
1218 While an "easy case recovery" sometimes appears to be successful
1219 even in the hard case, it may have unintended consequences. For
1220 example, a commit that was removed via git rebase --interactive
1221 will be resurrected!
1222 The idea is to manually tell git rebase "where the old subsystem ended
1224 and your topic began", that is, what the old merge base between them
1225 was. You will have to find a way to name the last commit of the old
1226 subsystem, for example:
1227 • With the subsystem reflog: after git fetch, the old tip of
1229 subsystem is at subsystem@{1}. Subsequent fetches will increase the
1230 number. (See git-reflog(1).)
1231 • Relative to the tip of topic: knowing that your topic has three
1233 commits, the old tip of subsystem must be topic~3.
1234 You can then transplant the old subsystem..topic to the new tip by
1236 saying (for the reflog case, and assuming you are on topic already):
1237 $ git rebase --onto subsystem subsystem@{1}
1239 The ripple effect of a "hard case" recovery is especially bad: everyone
1241 downstream from topic will now have to perform a "hard case" recovery
1242 too!
1243
1245 The interactive rebase command was originally designed to handle
1246 individual patch series. As such, it makes sense to exclude merge
1247 commits from the todo list, as the developer may have merged the
1248 then-current master while working on the branch, only to rebase all the
1249 commits onto master eventually (skipping the merge commits).
1250 However, there are legitimate reasons why a developer may want to
1252 recreate merge commits: to keep the branch structure (or "commit
1253 topology") when working on multiple, inter-related branches.
1254 In the following example, the developer works on a topic branch that
1256 refactors the way buttons are defined, and on another topic branch that
1257 uses that refactoring to implement a "Report a bug" button. The output
1258 of git log --graph --format=%s -5 may look like this:
1259 * Merge branch 'report-a-bug'
1261 |\
1262 | * Add the feedback button
1263 * | Merge branch 'refactor-button'
1264 |\ \
1265 | |/
1266 | * Use the Button class for all buttons
1267 | * Extract a generic Button class from the DownloadButton one
1268 The developer might want to rebase those commits to a newer master
1270 while keeping the branch topology, for example when the first topic
1271 branch is expected to be integrated into master much earlier than the
1272 second one, say, to resolve merge conflicts with changes to the
1273 DownloadButton class that made it into master.
1274 This rebase can be performed using the --rebase-merges option. It will
1276 generate a todo list looking like this:
1277 label onto
1279 # Branch: refactor-button
1281 reset onto
1282 pick 123456 Extract a generic Button class from the DownloadButton one
1283 pick 654321 Use the Button class for all buttons
1284 label refactor-button
1285 # Branch: report-a-bug
1287 reset refactor-button # Use the Button class for all buttons
1288 pick abcdef Add the feedback button
1289 label report-a-bug
1290 reset onto
1292 merge -C a1b2c3 refactor-button # Merge 'refactor-button'
1293 merge -C 6f5e4d report-a-bug # Merge 'report-a-bug'
1294 In contrast to a regular interactive rebase, there are label, reset and
1296 merge commands in addition to pick ones.
1297 The label command associates a label with the current HEAD when that
1299 command is executed. These labels are created as worktree-local refs
1300 (refs/rewritten/<label>) that will be deleted when the rebase finishes.
1301 That way, rebase operations in multiple worktrees linked to the same
1302 repository do not interfere with one another. If the label command
1303 fails, it is rescheduled immediately, with a helpful message how to
1304 proceed.
1305 The reset command resets the HEAD, index and worktree to the specified
1307 revision. It is similar to an exec git reset --hard <label>, but
1308 refuses to overwrite untracked files. If the reset command fails, it is
1309 rescheduled immediately, with a helpful message how to edit the todo
1310 list (this typically happens when a reset command was inserted into the
1311 todo list manually and contains a typo).
1312 The merge command will merge the specified revision(s) into whatever is
1314 HEAD at that time. With -C <original-commit>, the commit message of the
1315 specified merge commit will be used. When the -C is changed to a
1316 lower-case -c, the message will be opened in an editor after a
1317 successful merge so that the user can edit the message.
1318 If a merge command fails for any reason other than merge conflicts
1320 (i.e. when the merge operation did not even start), it is rescheduled
1321 immediately.
1322 At this time, the merge command will always use the recursive merge
1324 strategy for regular merges, and octopus for octopus merges, with no
1325 way to choose a different one. To work around this, an exec command can
1326 be used to call git merge explicitly, using the fact that the labels
1327 are worktree-local refs (the ref refs/rewritten/onto would correspond
1328 to the label onto, for example).
1329 Note: the first command (label onto) labels the revision onto which the
1331 commits are rebased; The name onto is just a convention, as a nod to
1332 the --onto option.
1333 It is also possible to introduce completely new merge commits from
1335 scratch by adding a command of the form merge <merge-head>. This form
1336 will generate a tentative commit message and always open an editor to
1337 let the user edit it. This can be useful e.g. when a topic branch turns
1338 out to address more than a single concern and wants to be split into
1339 two or even more topic branches. Consider this todo list:
1340 pick 192837 Switch from GNU Makefiles to CMake
1342 pick 5a6c7e Document the switch to CMake
1343 pick 918273 Fix detection of OpenSSL in CMake
1344 pick afbecd http: add support for TLS v1.3
1345 pick fdbaec Fix detection of cURL in CMake on Windows
1346 The one commit in this list that is not related to CMake may very well
1348 have been motivated by working on fixing all those bugs introduced by
1349 switching to CMake, but it addresses a different concern. To split this
1350 branch into two topic branches, the todo list could be edited like
1351 this:
1352 label onto
1354 pick afbecd http: add support for TLS v1.3
1356 label tlsv1.3
1357 reset onto
1359 pick 192837 Switch from GNU Makefiles to CMake
1360 pick 918273 Fix detection of OpenSSL in CMake
1361 pick fdbaec Fix detection of cURL in CMake on Windows
1362 pick 5a6c7e Document the switch to CMake
1363 label cmake
1364 reset onto
1366 merge tlsv1.3
1367 merge cmake
1368
1370 The todo list presented by the deprecated --preserve-merges
1371 --interactive does not represent the topology of the revision graph
1372 (use --rebase-merges instead). Editing commits and rewording their
1373 commit messages should work fine, but attempts to reorder commits tend
1374 to produce counterintuitive results. Use --rebase-merges in such
1375 scenarios instead.
1376 For example, an attempt to rearrange
1378 1 --- 2 --- 3 --- 4 --- 5
1380 to
1382 1 --- 2 --- 4 --- 3 --- 5
1384 by moving the "pick 4" line will result in the following history:
1386 3
1388 /
1389 1 --- 2 --- 4 --- 5
1390
1392 Part of the git(1) suite
1393
1395 1. revert-a-faulty-merge How-To
1396 file:///usr/share/doc/git/howto/revert-a-faulty-merge.html
1397 2. filter-repo
1399 https://github.com/newren/git-filter-repo
1400Git 2.31.1 2021-03-26 GIT-REBASE(1)