git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH v4 06/12] builtin/history: implement "reword" subcommand

From
Karthik Nayak <karthik.188@gmail.com>
Date
Oct 14, 2025, 11:04 UTC
Message-ID
<CAOLa=ZSU8yr9Gn0EZ7x705qPyVM-qiMjgMCNCb8p8SMGTToxqQ@mail.gmail.com>
In-Reply-To
<20251001-b4-pks-history-builtin-v4-6-8e61ddb86317@pks.im>
Patrick Steinhardt <ps@pks.im> writes:
> Implement a new "reword" subcommand for git-history(1). This subcommand
> is essentially the same as if a user performed an interactive rebase
> with a single commit changed to use the "reword" verb.
>
[snip]
Show 16 quoted lines
> diff --git a/builtin/history.c b/builtin/history.c
> index f6fe32610b..7b2a0023e8 100644
> --- a/builtin/history.c
> +++ b/builtin/history.c
> @@ -1,22 +1,389 @@
> +#define USE_THE_REPOSITORY_VARIABLE
> +
>  #include "builtin.h"
> +#include "commit-reach.h"
> +#include "commit.h"
> +#include "config.h"
> +#include "editor.h"
> +#include "environment.h"
>  #include "gettext.h"
> +#include "hex.h"
> +#include "oidmap.h"
Nit: This can be dropped, perhaps needed in a future patch?
Show 28 quoted lines
>  #include "parse-options.h"
> +#include "refs.h"
> +#include "replay.h"
> +#include "reset.h"
> +#include "revision.h"
> +#include "sequencer.h"
> +#include "strvec.h"
> +#include "tree.h"
> +#include "wt-status.h"
> +
> +static int collect_commits(struct repository *repo,
> +			   struct commit *old_commit,
> +			   struct commit *new_commit,
> +			   struct strvec *out)
> +{
> +	struct setup_revision_opt revision_opts = {
> +		.assume_dashdash = 1,
> +	};
> +	struct strvec revisions = STRVEC_INIT;
> +	struct commit_list *from_list = NULL;
> +	struct commit *child;
> +	struct rev_info rev = { 0 };
> +	int ret;
> +
> +	/*
> +	 * Check that the old commit actually is an ancestor of HEAD. If not
> +	 * the whole request becomes nonsensical.
> +	*/
Missing space here
Show 7 quoted lines
> +	if (old_commit) {
> +		commit_list_insert(old_commit, &from_list);
> +		if (!repo_is_descendant_of(repo, new_commit, from_list)) {
> +			ret = error(_("commit must be reachable from current HEAD commit"));
> +			goto out;
> +		}
> +	}

Makes sense. There is an inherent assumption using the 'git history' command that you want to modify the history of the current reference.

One question, wouldn't it make sense to parse and check that the commit to be reworded should be checked to be a descendant of HEAD earlier on in `cmd_history_reword()`?

This would ensure this function `collect_commits()` doesn't worry about how it is meant to be used, and simply worries about collecting commits.

Show 27 quoted lines
> +	repo_init_revisions(repo, &rev, NULL);
> +	strvec_push(&revisions, "");
> +	strvec_push(&revisions, oid_to_hex(&new_commit->object.oid));
> +	if (old_commit)
> +		strvec_pushf(&revisions, "^%s", oid_to_hex(&old_commit->object.oid));
> +
> +	setup_revisions_from_strvec(&revisions, &rev, &revision_opts);
> +	if (revisions.nr != 1 || prepare_revision_walk(&rev)) {
> +		ret = error(_("revision walk setup failed"));
> +		goto out;
> +	}
> +
> +	while ((child = get_revision(&rev))) {
> +		if (old_commit && !child->parents)
> +			BUG("revision walk did not find child commit");
> +		if (child->parents && child->parents->next) {
> +			ret = error(_("cannot rearrange commit history with merges"));
> +			goto out;
> +		}
> +
> +		strvec_push(out, oid_to_hex(&child->object.oid));
> +
> +		if (child->parents && old_commit &&
> +		    commit_list_contains(old_commit, child->parents))
> +			break;
> +	}
> +

Okay makes sense here, we collect all the commits we break as soon as we reach old_commit. Since we check for merges at the start of the loop, the history should be linear.

[snip]
Show 25 quoted lines
> +static void replace_commits(struct strvec *commits,
> +			    const struct object_id *commit_to_replace,
> +			    const struct object_id *replacements,
> +			    size_t replacements_nr)
> +{
> +	char commit_to_replace_oid[GIT_MAX_HEXSZ + 1];
> +	struct strvec replacement_oids = STRVEC_INIT;
> +	bool found = false;
> +
> +	oid_to_hex_r(commit_to_replace_oid, commit_to_replace);
> +	for (size_t i = 0; i < replacements_nr; i++)
> +		strvec_push(&replacement_oids, oid_to_hex(&replacements[i]));
> +
> +	for (size_t i = 0; i < commits->nr; i++) {
> +		if (strcmp(commits->v[i], commit_to_replace_oid))
> +			continue;
> +		strvec_splice(commits, i, 1, replacement_oids.v, replacement_oids.nr);
> +		found = true;
> +		break;
> +	}
> +	if (!found)
> +		BUG("could not find commit to replace");
> +
> +	strvec_clear(&replacement_oids);
> +}

So this basically goes over the commits that we pass and replaces a single commit with a set of commits. In our case that would be

  C0 C1 C2 ...  CN
   └─(reword)    └─(HEAD)
     ↓ (rewrites to)
  R0 R1 R2 ... RN C1 C2 ... CN
   └────────────┘            └─(HEAD)
    (new commits)

Makes sense. I assume we use a list of replacements here for future commands.

Show 15 quoted lines
> +static int apply_commits(struct repository *repo,
> +			 const struct strvec *commits,
> +			 struct commit *onto,
> +			 struct commit *orig_head,
> +			 const char *action)
> +{
> +	struct reset_head_opts reset_opts = { 0 };
> +	struct merge_options merge_opts = { 0 };
> +	struct merge_result result = { 0 };
> +	struct strbuf buf = STRBUF_INIT;
> +	kh_oid_map_t *replayed_commits;
> +	int ret;
> +
> +	replayed_commits = kh_init_oid_map();
> +

So this is used to the pass the replayed_commits list to the replay mechanism so that the appropriate base commit is selected.

Show 26 quoted lines
> +	init_basic_merge_options(&merge_opts, repo);
> +	merge_opts.show_rename_progress = 0;
> +
> +	for (size_t i = 0; i < commits->nr; i++) {
> +		struct object_id commit_id;
> +		struct commit *commit;
> +		const char *end;
> +		int hash_result;
> +		khint_t pos;
> +
> +		if (parse_oid_hex_algop(commits->v[i], &commit_id, &end,
> +					repo->hash_algo)) {
> +			ret = error(_("invalid object ID: %s"), commits->v[i]);
> +			goto out;
> +		}
> +
> +		commit = lookup_commit(repo, &commit_id);
> +		if (!commit || repo_parse_commit(repo, commit)) {
> +			ret = error(_("failed to look up commit: %s"), oid_to_hex(&commit_id));
> +			goto out;
> +		}
> +
> +		if (!onto) {
> +			onto = commit;
> +			result.clean = 1;
> +			result.tree = repo_get_commit_tree(repo, commit);

So if there is no onto, we're starting at the root commit and expect a clean merge.

Show 5 quoted lines
> +		} else {
> +			onto = replay_pick_regular_commit(repo, commit, replayed_commits,
> +							  onto, &merge_opts, &result);
> +			if (!onto)
> +				break;
Else we will replay the current commit onto the prev 'onto' commit.
Show 26 quoted lines
> +		}
> +
> +		pos = kh_put_oid_map(replayed_commits, commit->object.oid, &hash_result);
> +		if (hash_result == 0) {
> +			ret = error(_("duplicate rewritten commit: %s\n"),
> +				    oid_to_hex(&commit->object.oid));
> +			goto out;
> +		}
> +		kh_value(replayed_commits, pos) = onto;
> +	}
> +
> +	if (!result.clean) {
> +		ret = error(_("could not merge"));
> +		goto out;
> +	}
> +
> +	reset_opts.oid = &onto->object.oid;
> +	strbuf_addf(&buf, "%s: switch to rewritten %s", action, oid_to_hex(reset_opts.oid));
> +	reset_opts.flags = RESET_HEAD_REFS_ONLY | RESET_ORIG_HEAD;
> +	reset_opts.orig_head = &orig_head->object.oid;
> +	reset_opts.default_reflog_action = action;
> +	if (reset_head(repo, &reset_opts) < 0) {
> +		ret = error(_("could not switch to %s"), oid_to_hex(reset_opts.oid));
> +		goto out;
> +	}
> +
We finally update the HEAD reference also.
Show 29 quoted lines
> +	ret = 0;
> +
> +out:
> +	kh_destroy_oid_map(replayed_commits);
> +	merge_finalize(&merge_opts, &result);
> +	strbuf_release(&buf);
> +	return ret;
> +}
> +
> +static void change_data_free(void *util, const char *str UNUSED)
> +{
> +	struct wt_status_change_data *d = util;
> +	free(d->rename_source);
> +	free(d);
> +}
> +
> +static int fill_commit_message(struct repository *repo,
> +			       const struct object_id *old_tree,
> +			       const struct object_id *new_tree,
> +			       const char *default_message,
> +			       const char *provided_message,
> +			       const char *action,
> +			       struct strbuf *out)
> +{
> +	if (!provided_message) {
> +		const char *path = git_path_commit_editmsg();
> +		const char *hint =
> +			_("Please enter the commit message for the %s changes. Lines starting\n"
> +     "with '%s' will be kept; you may remove them yourself if you want to.\n");
Shouldn't this be s/kept/removed? Also this line needs to be aligned.
Show 27 quoted lines
> +		struct wt_status s;
> +
> +		strbuf_addstr(out, default_message);
> +		strbuf_addch(out, '\n');
> +		strbuf_commented_addf(out, comment_line_str, hint, action, comment_line_str);
> +		write_file_buf(path, out->buf, out->len);
> +
> +		wt_status_prepare(repo, &s);
> +		FREE_AND_NULL(s.branch);
> +		s.ahead_behind_flags = AHEAD_BEHIND_QUICK;
> +		s.commit_template = 1;
> +		s.colopts = 0;
> +		s.display_comment_prefix = 1;
> +		s.hints = 0;
> +		s.use_color = 0;
> +		s.whence = FROM_COMMIT;
> +		s.committable = 1;
> +
> +		s.fp = fopen(git_path_commit_editmsg(), "a");
> +		if (!s.fp)
> +			return error_errno(_("could not open '%s'"), git_path_commit_editmsg());
> +
> +		wt_status_collect_changes_trees(&s, old_tree, new_tree);
> +		wt_status_print(&s);
> +		wt_status_collect_free_buffers(&s);
> +		string_list_clear_func(&s.change, change_data_free);
> +

Nice, so we show the user the diff tree and then ask them to modify the existing message as they want.

Show 26 quoted lines
> +		strbuf_reset(out);
> +		if (launch_editor(path, out, NULL)) {
> +			fprintf(stderr, _("Please supply the message using the -m option.\n"));
> +			return -1;
> +		}
> +		strbuf_stripspace(out, comment_line_str);
> +	} else {
> +		strbuf_addstr(out, provided_message);
> +	}
> +
> +	cleanup_message(out, COMMIT_MSG_CLEANUP_ALL, 0);
> +
> +	if (!out->len) {
> +		fprintf(stderr, _("Aborting commit due to empty commit message.\n"));
> +		return -1;
> +	}
> +
> +	return 0;
> +}
> +
> +static int cmd_history_reword(int argc,
> +			      const char **argv,
> +			      const char *prefix,
> +			      struct repository *repo)
> +{
> +	const char * const usage[] = {
Same as the prev commit, this should be '*const'.
Show 25 quoted lines
> +		N_("git history reword [<options>] <commit>"),
> +		NULL,
> +	};
> +	const char *commit_message = NULL;
> +	struct option options[] = {
> +		OPT_STRING('m', "message", &commit_message, N_("message"), N_("commit message")),
> +		OPT_END(),
> +	};
> +	struct strbuf final_message = STRBUF_INIT;
> +	struct commit *original_commit, *parent, *head;
> +	struct strvec commits = STRVEC_INIT;
> +	struct object_id parent_tree_oid, original_commit_tree_oid;
> +	struct object_id rewritten_commit;
> +	const char *original_message, *original_body, *ptr;
> +	char *original_author = NULL;
> +	size_t len;
> +	int ret;
> +
> +	argc = parse_options(argc, argv, prefix, options, usage, 0);
> +	if (argc != 1) {
> +		ret = error(_("command expects a single revision"));
> +		goto out;
> +	}
> +	repo_config(repo, git_default_config, NULL);
> +

Right. We only expect something like 'git history reword @~10', so we allow only one commit to be re-worded at a given time.

Show 5 quoted lines
> +	original_commit = lookup_commit_reference_by_name(argv[0]);
> +	if (!original_commit) {
> +		ret = error(_("commit to be reworded cannot be found: %s"), argv[0]);
> +		goto out;
> +	}
So here we want to parse the commit-ish name to find the commit.
Show 5 quoted lines
> +	if (repo_parse_commit(repo, original_commit)) {
> +		ret = error(_("unable to parse commit %s"),
> +			    oid_to_hex(&original_commit->object.oid));
> +		goto out;
> +	}

Isn't this already done as part of `lookup_commit_reference_by_name_gently()` which is called by `lookup_commit_reference_by_name()` ?

> +	original_commit_tree_oid = repo_get_commit_tree(repo, original_commit)->object.oid;
> +

Okay we get the tree of the commit, I assume this is used to create the new commit and replay the children on top of it.

Show 27 quoted lines
> +	parent = original_commit->parents ? original_commit->parents->item : NULL;
> +	if (parent) {
> +		if (repo_parse_commit(repo, parent)) {
> +			ret = error(_("unable to parse commit %s"),
> +				    oid_to_hex(&parent->object.oid));
> +			goto out;
> +		}
> +		parent_tree_oid = repo_get_commit_tree(repo, parent)->object.oid;
> +	} else {
> +		oidcpy(&parent_tree_oid, repo->hash_algo->empty_tree);
> +	}
> +
> +	head = lookup_commit_reference_by_name("HEAD");
> +	if (!head) {
> +		ret = error(_("could not resolve HEAD to a commit"));
> +		goto out;
> +	}
> +
> +	/*
> +	 * Collect the list of commits that we'll have to reapply now already.
> +	 * This ensures that we'll abort early on in case the range of commits
> +	 * contains merges, which we do not yet handle.
> +	 */
> +	ret = collect_commits(repo, parent, head, &commits);
> +	if (ret < 0)
> +		goto out;
> +
The user is currently at HEAD and wishes to reword a commit.
  PN ... P2 P1 C0 C1 ... CN
               └─(reword) └─(HEAD)
So we want to re-word C0, so we need to collect P1..CN.
Show 11 quoted lines
> +	/* We retain authorship of the original commit. */
> +	original_message = repo_logmsg_reencode(repo, original_commit, NULL, NULL);
> +	ptr = find_commit_header(original_message, "author", &len);
> +	if (ptr)
> +		original_author = xmemdupz(ptr, len);
> +	find_commit_subject(original_message, &original_body);
> +
> +	ret = fill_commit_message(repo, &parent_tree_oid, &original_commit_tree_oid,
> +				  original_body, commit_message, "reworded", &final_message);
> +	if (ret < 0)
> +		goto out;
We obtain the new message from the user.
> +
> +	ret = commit_tree(final_message.buf, final_message.len,
> +			  &repo_get_commit_tree(repo, original_commit)->object.oid,
Can't we use original_commit_tree_oid here?
Show 8 quoted lines
> +			  original_commit->parents, &rewritten_commit, original_author, NULL);
> +	if (ret < 0) {
> +		ret = error(_("failed writing reworded commit"));
> +		goto out;
> +	}
> +
> +	replace_commits(&commits, &original_commit->object.oid, &rewritten_commit, 1);
> +
Okay we replace the commits we obtained with the new rewritten commit.
Show 6 quoted lines
> +	ret = apply_commits(repo, &commits, parent, head, "reword");
> +	if (ret < 0)
> +		goto out;
> +
> +	ret = 0;
> +

So we're now asking to replay the commits onto parent and update HEAD. Looks good.

Show 16 quoted lines
> +out:
> +	strbuf_release(&final_message);
> +	strvec_clear(&commits);
> +	free(original_author);
> +	return ret;
> +}
>
>  int cmd_history(int argc,
>  		const char **argv,
>  		const char *prefix,
> -		struct repository *repo UNUSED)
> +		struct repository *repo)
>  {
>  	const char * const usage[] = {
>  		N_("git history [<options>]"),
> +		N_("git history reword [<options>] <commit>"),
This string is used twice, perhaps we move it to a macro?
[snip]
The tests looked good too!
Previous: Patrick SteinhardtNext: Patrick Steinhardt
Message 156 of 325 in “Introduce git-history(1) command for easy history editing”
  1. 00/11 Introduce git-history(1) command for easy history editingPatrick Steinhardt, Aug 19, 2025
  2. 01/11 sequencer: optionally skip printing commit summaryPatrick Steinhardt, Aug 19, 2025
  3. 02/11 sequencer: add option to rewind HEAD after picking commitsPatrick Steinhardt, Aug 19, 2025
  4. 03/11 cache-tree: allow writing in-memory index as treePatrick Steinhardt, Aug 19, 2025
  5. 04/11 builtin: add new "history" commandPatrick Steinhardt, Aug 19, 2025
  6. 05/11 builtin/history: implement "drop" subcommandPatrick Steinhardt, Aug 19, 2025
  7. Ben KnobleAug 20, 2025
  8. Patrick SteinhardtAug 22, 2025
  9. Jean-Noël AVILAAug 23, 2025
  10. Patrick SteinhardtAug 24, 2025
  11. 06/11 builtin/history: implement "reorder" subcommandPatrick Steinhardt, Aug 19, 2025
  12. Jean-Noël AVILAAug 23, 2025
  13. Kristoffer HaugsbakkAug 24, 2025
  14. Patrick SteinhardtAug 24, 2025
  15. 07/11 add-patch: split out header from "add-interactive.h"Patrick Steinhardt, Aug 19, 2025
  16. 08/11 add-patch: split out `struct interactive_options`Patrick Steinhardt, Aug 19, 2025
  17. 09/11 add-patch: remove dependency on "add-interactive" subsystemPatrick Steinhardt, Aug 19, 2025
  18. 10/11 add-patch: add support for in-memory index patchingPatrick Steinhardt, Aug 19, 2025
  19. D. Ben KnobleAug 20, 2025
  20. Patrick SteinhardtAug 22, 2025
  21. 11/11 builtin/history: implement "split" subcommandPatrick Steinhardt, Aug 19, 2025
  22. D. Ben KnobleAug 20, 2025
  23. Patrick SteinhardtAug 22, 2025
  24. Junio C HamanoAug 22, 2025
  25. Patrick SteinhardtAug 24, 2025
  26. Jean-Noël AVILAAug 23, 2025
  27. Patrick SteinhardtAug 24, 2025
  28. D. Ben KnobleAug 19, 2025
  29. Patrick SteinhardtAug 20, 2025
  30. Ben KnobleAug 20, 2025
  31. Junio C HamanoAug 20, 2025
  32. Ben KnobleAug 20, 2025
  33. Patrick SteinhardtAug 22, 2025
  34. Junio C HamanoAug 22, 2025
  35. Sergey OrganovAug 21, 2025
  36. Ben KnobleAug 21, 2025
  37. Sergey OrganovAug 21, 2025
  38. Martin von ZweigbergkAug 24, 2025
  39. Patrick SteinhardtAug 24, 2025
  40. SZEDER GáborSep 17, 2025
  41. Matthias BeyerDec 3, 2025
  42. Phillip WoodDec 10, 2025
  43. Matthias BeyerDec 10, 2025
  44. Phillip WoodDec 10, 2025
  45. Junio C HamanoDec 10, 2025
  46. Patrick SteinhardtDec 19, 2025
  47. SZEDER GáborDec 19, 2025
  48. Patrick SteinhardtDec 19, 2025
  49. Elijah NewrenDec 19, 2025
  50. Elijah NewrenDec 20, 2025
  51. Patrick SteinhardtJan 6, 2026
  52. Phillip WoodDec 22, 2025
  53. Patrick SteinhardtJan 6, 2026
  54. D. Ben KnobleDec 22, 2025
  55. Martin von ZweigbergkDec 10, 2025
  56. Elijah NewrenDec 10, 2025
  57. Martin von ZweigbergkDec 10, 2025
  58. Elijah NewrenDec 10, 2025
  59. Kristoffer HaugsbakkDec 15, 2025
  60. Kristoffer HaugsbakkAug 24, 2025
  61. Patrick SteinhardtAug 24, 2025
  62. 00/16 Introduce git-history(1) command for easy history editingPatrick Steinhardt, Aug 24, 2025
  63. 01/16 sequencer: optionally skip printing commit summaryPatrick Steinhardt, Aug 24, 2025
  64. 02/16 sequencer: add option to rewind HEAD after picking commitsPatrick Steinhardt, Aug 24, 2025
  65. 03/16 sequencer: introduce new history editing modePatrick Steinhardt, Aug 24, 2025
  66. D. Ben KnobleAug 26, 2025
  67. Patrick SteinhardtSep 3, 2025
  68. 04/16 sequencer: stop using `the_repository` in `sequencer_remove_state()`Patrick Steinhardt, Aug 24, 2025
  69. 05/16 cache-tree: allow writing in-memory index as treePatrick Steinhardt, Aug 24, 2025
  70. Junio C HamanoAug 25, 2025
  71. Patrick SteinhardtSep 3, 2025
  72. 06/16 builtin: add new "history" commandPatrick Steinhardt, Aug 24, 2025
  73. 07/16 builtin/history: introduce subcommands to manage interrupted rewritesPatrick Steinhardt, Aug 24, 2025
  74. 08/16 builtin/history: implement "drop" subcommandPatrick Steinhardt, Aug 24, 2025
  75. 09/16 builtin/history: implement "reorder" subcommandPatrick Steinhardt, Aug 24, 2025
  76. D. Ben KnobleAug 26, 2025
  77. Patrick SteinhardtSep 3, 2025
  78. 10/16 add-patch: split out header from "add-interactive.h"Patrick Steinhardt, Aug 24, 2025
  79. Junio C HamanoAug 25, 2025
  80. 11/16 add-patch: split out `struct interactive_options`Patrick Steinhardt, Aug 24, 2025
  81. 12/16 add-patch: remove dependency on "add-interactive" subsystemPatrick Steinhardt, Aug 24, 2025
  82. Junio C HamanoAug 25, 2025
  83. 13/16 add-patch: add support for in-memory index patchingPatrick Steinhardt, Aug 24, 2025
  84. 14/16 wt-status: provide function to expose status for treesPatrick Steinhardt, Aug 24, 2025
  85. 15/16 builtin/history: implement "split" subcommandPatrick Steinhardt, Aug 24, 2025
  86. Kristoffer HaugsbakkAug 24, 2025
  87. Patrick SteinhardtSep 3, 2025
  88. D. Ben KnobleAug 26, 2025
  89. Patrick SteinhardtSep 3, 2025
  90. D. Ben KnobleSep 3, 2025
  91. Patrick SteinhardtSep 4, 2025
  92. D. Ben KnobleSep 12, 2025
  93. Patrick SteinhardtSep 15, 2025
  94. Ben KnobleSep 15, 2025
  95. 16/16 builtin/history: implement "reword" subcommandPatrick Steinhardt, Aug 24, 2025
  96. Kristoffer HaugsbakkAug 24, 2025
  97. Patrick SteinhardtSep 3, 2025
  98. D. Ben KnobleSep 3, 2025
  99. Patrick SteinhardtSep 4, 2025
  100. 00/18 Introduce git-history(1) command for easy history editingPatrick Steinhardt, Sep 4, 2025
  101. 01/18 sequencer: optionally skip printing commit summaryPatrick Steinhardt, Sep 4, 2025
  102. Phillip WoodSep 10, 2025
  103. Patrick SteinhardtSep 15, 2025
  104. 02/18 sequencer: add option to rewind HEAD after picking commitsPatrick Steinhardt, Sep 4, 2025
  105. Phillip WoodSep 10, 2025
  106. Patrick SteinhardtSep 15, 2025
  107. Phillip WoodSep 15, 2025
  108. 03/18 sequencer: introduce new history editing modePatrick Steinhardt, Sep 4, 2025
  109. 04/18 sequencer: stop using `the_repository` in `sequencer_remove_state()`Patrick Steinhardt, Sep 4, 2025
  110. 05/18 sequencer: wire up "rewritten-hook" for REPLAY_HISTORY_EDITPatrick Steinhardt, Sep 4, 2025
  111. 06/18 cache-tree: allow writing in-memory index as treePatrick Steinhardt, Sep 4, 2025
  112. 07/18 builtin: add new "history" commandPatrick Steinhardt, Sep 4, 2025
  113. 08/18 builtin/history: introduce subcommands to manage interrupted rewritesPatrick Steinhardt, Sep 4, 2025
  114. 09/18 builtin/history: implement "drop" subcommandPatrick Steinhardt, Sep 4, 2025
  115. 10/18 builtin/history: implement "reorder" subcommandPatrick Steinhardt, Sep 4, 2025
  116. 11/18 add-patch: split out header from "add-interactive.h"Patrick Steinhardt, Sep 4, 2025
  117. 12/18 add-patch: split out `struct interactive_options`Patrick Steinhardt, Sep 4, 2025
  118. 13/18 add-patch: remove dependency on "add-interactive" subsystemPatrick Steinhardt, Sep 4, 2025
  119. 14/18 add-patch: add support for in-memory index patchingPatrick Steinhardt, Sep 4, 2025
  120. 15/18 wt-status: provide function to expose status for treesPatrick Steinhardt, Sep 4, 2025
  121. 16/18 sequencer: allow callers to provide mappings for the old commitPatrick Steinhardt, Sep 4, 2025
  122. 17/18 builtin/history: implement "split" subcommandPatrick Steinhardt, Sep 4, 2025
  123. Phillip WoodSep 10, 2025
  124. Patrick SteinhardtSep 15, 2025
  125. 18/18 builtin/history: implement "reword" subcommandPatrick Steinhardt, Sep 4, 2025
  126. Phillip WoodSep 10, 2025
  127. Patrick SteinhardtSep 15, 2025
  128. Phillip WoodSep 15, 2025
  129. Patrick SteinhardtSep 16, 2025
  130. Phillip WoodSep 16, 2025
  131. Kristoffer HaugsbakkSep 5, 2025
  132. Patrick SteinhardtSep 5, 2025
  133. Elijah NewrenSep 7, 2025
  134. Phillip WoodSep 10, 2025
  135. Phillip WoodSep 10, 2025
  136. Patrick SteinhardtSep 15, 2025
  137. Oswald BuddenhagenSep 16, 2025
  138. Junio C HamanoSep 10, 2025
  139. Patrick SteinhardtSep 15, 2025
  140. 00/12 Introduce git-history(1) command for easy history editingPatrick Steinhardt, Oct 1, 2025
  141. 01/12 wt-status: provide function to expose status for treesPatrick Steinhardt, Oct 1, 2025
  142. Karthik NayakOct 14, 2025
  143. Patrick SteinhardtOct 21, 2025
  144. 02/12 replay: extract logic to pick commitsPatrick Steinhardt, Oct 1, 2025
  145. 03/12 replay: stop using `the_repository`Patrick Steinhardt, Oct 1, 2025
  146. Karthik NayakOct 14, 2025
  147. 04/12 replay: parse commits before dereferencing themPatrick Steinhardt, Oct 1, 2025
  148. Karthik NayakOct 14, 2025
  149. 05/12 builtin: add new "history" commandPatrick Steinhardt, Oct 1, 2025
  150. Kristoffer HaugsbakkOct 2, 2025
  151. Karthik NayakOct 14, 2025
  152. Patrick SteinhardtOct 21, 2025
  153. Junio C HamanoOct 22, 2025
  154. Karthik NayakOct 22, 2025
  155. 06/12 builtin/history: implement "reword" subcommandPatrick Steinhardt, Oct 1, 2025
  156. Karthik NayakOct 14, 2025
  157. Patrick SteinhardtOct 21, 2025
  158. 07/12 add-patch: split out header from "add-interactive.h"Patrick Steinhardt, Oct 1, 2025
  159. 08/12 add-patch: split out `struct interactive_options`Patrick Steinhardt, Oct 1, 2025
  160. Kristoffer HaugsbakkOct 2, 2025
  161. Karthik NayakOct 14, 2025
  162. Patrick SteinhardtOct 21, 2025
  163. 09/12 add-patch: remove dependency on "add-interactive" subsystemPatrick Steinhardt, Oct 1, 2025
  164. 10/12 add-patch: add support for in-memory index patchingPatrick Steinhardt, Oct 1, 2025
  165. Kristoffer HaugsbakkOct 2, 2025
  166. Patrick SteinhardtOct 2, 2025
  167. Karthik NayakOct 14, 2025
  168. 11/12 cache-tree: allow writing in-memory index as treePatrick Steinhardt, Oct 1, 2025
  169. 12/12 builtin/history: implement "split" subcommandPatrick Steinhardt, Oct 1, 2025
  170. Karthik NayakOct 14, 2025
  171. Patrick SteinhardtOct 21, 2025
  172. D. Ben KnobleOct 21, 2025
  173. Patrick SteinhardtOct 27, 2025
  174. Karthik NayakOct 14, 2025
  175. Junio C HamanoOct 14, 2025
  176. 00/12 Introduce git-history(1) command for easy history editingPatrick Steinhardt, Oct 21, 2025
  177. 01/12 wt-status: provide function to expose status for treesPatrick Steinhardt, Oct 21, 2025
  178. Junio C HamanoOct 21, 2025
  179. 02/12 replay: extract logic to pick commitsPatrick Steinhardt, Oct 21, 2025
  180. Junio C HamanoOct 21, 2025
  181. 03/12 replay: stop using `the_repository`Patrick Steinhardt, Oct 21, 2025
  182. Junio C HamanoOct 21, 2025
  183. Junio C HamanoOct 21, 2025
  184. 04/12 replay: parse commits before dereferencing themPatrick Steinhardt, Oct 21, 2025
  185. Junio C HamanoOct 21, 2025
  186. Patrick SteinhardtOct 27, 2025
  187. 05/12 builtin: add new "history" commandPatrick Steinhardt, Oct 21, 2025
  188. Junio C HamanoOct 21, 2025
  189. Patrick SteinhardtOct 27, 2025
  190. 06/12 builtin/history: implement "reword" subcommandPatrick Steinhardt, Oct 21, 2025
  191. Junio C HamanoOct 21, 2025
  192. D. Ben KnobleOct 21, 2025
  193. Patrick SteinhardtOct 27, 2025
  194. Patrick SteinhardtOct 27, 2025
  195. 07/12 add-patch: split out header from "add-interactive.h"Patrick Steinhardt, Oct 21, 2025
  196. 08/12 add-patch: split out `struct interactive_options`Patrick Steinhardt, Oct 21, 2025
  197. 09/12 add-patch: remove dependency on "add-interactive" subsystemPatrick Steinhardt, Oct 21, 2025
  198. 10/12 add-patch: add support for in-memory index patchingPatrick Steinhardt, Oct 21, 2025
  199. 11/12 cache-tree: allow writing in-memory index as treePatrick Steinhardt, Oct 21, 2025
  200. 12/12 builtin/history: implement "split" subcommandPatrick Steinhardt, Oct 21, 2025
  201. Junio C HamanoOct 21, 2025
  202. 00/11 Introduce git-history(1) command for easy history editingPatrick Steinhardt, Oct 27, 2025
  203. 01/11 wt-status: provide function to expose status for treesPatrick Steinhardt, Oct 27, 2025
  204. 02/11 replay: extract logic to pick commitsPatrick Steinhardt, Oct 27, 2025
  205. Phillip WoodNov 17, 2025
  206. Elijah NewrenNov 20, 2025
  207. 03/11 replay: stop using `the_repository`Patrick Steinhardt, Oct 27, 2025
  208. Elijah NewrenNov 20, 2025
  209. Patrick SteinhardtDec 2, 2025
  210. 04/11 builtin: add new "history" commandPatrick Steinhardt, Oct 27, 2025
  211. Phillip WoodNov 17, 2025
  212. Patrick SteinhardtDec 2, 2025
  213. Elijah NewrenNov 20, 2025
  214. Patrick SteinhardtDec 2, 2025
  215. D. Ben KnobleDec 2, 2025
  216. Patrick SteinhardtDec 3, 2025
  217. 05/11 builtin/history: implement "reword" subcommandPatrick Steinhardt, Oct 27, 2025
  218. Phillip WoodNov 17, 2025
  219. Patrick SteinhardtDec 2, 2025
  220. Phillip WoodDec 10, 2025
  221. Elijah NewrenNov 20, 2025
  222. Patrick SteinhardtDec 2, 2025
  223. SZEDER GáborNov 25, 2025
  224. Patrick SteinhardtDec 2, 2025
  225. 06/11 add-patch: split out header from "add-interactive.h"Patrick Steinhardt, Oct 27, 2025
  226. Elijah NewrenNov 20, 2025
  227. 07/11 add-patch: split out `struct interactive_options`Patrick Steinhardt, Oct 27, 2025
  228. Elijah NewrenNov 20, 2025
  229. Phillip WoodNov 20, 2025
  230. Patrick SteinhardtDec 2, 2025
  231. 08/11 add-patch: remove dependency on "add-interactive" subsystemPatrick Steinhardt, Oct 27, 2025
  232. Elijah NewrenNov 20, 2025
  233. Phillip WoodNov 20, 2025
  234. 09/11 add-patch: add support for in-memory index patchingPatrick Steinhardt, Oct 27, 2025
  235. Elijah NewrenNov 20, 2025
  236. Phillip WoodNov 20, 2025
  237. Patrick SteinhardtDec 2, 2025
  238. 10/11 cache-tree: allow writing in-memory index as treePatrick Steinhardt, Oct 27, 2025
  239. Elijah NewrenNov 20, 2025
  240. 11/11 builtin/history: implement "split" subcommandPatrick Steinhardt, Oct 27, 2025
  241. Elijah NewrenNov 20, 2025
  242. Patrick SteinhardtDec 2, 2025
  243. Phillip WoodNov 21, 2025
  244. Patrick SteinhardtDec 2, 2025
  245. Phillip WoodDec 10, 2025
  246. Patrick SteinhardtDec 19, 2025
  247. Sergey OrganovNov 12, 2025
  248. Elijah NewrenNov 20, 2025
  249. Junio C HamanoNov 20, 2025
  250. Elijah NewrenNov 20, 2025
  251. Junio C HamanoNov 20, 2025
  252. Elijah NewrenNov 20, 2025
  253. Phillip WoodNov 21, 2025
  254. Junio C HamanoNov 21, 2025
  255. Elijah NewrenNov 23, 2025
  256. Patrick SteinhardtDec 2, 2025
  257. Elijah NewrenDec 5, 2025
  258. Patrick SteinhardtDec 9, 2025
  259. Martin von ZweigbergkDec 9, 2025
  260. Phillip WoodDec 10, 2025
  261. Elijah NewrenDec 10, 2025
  262. Kristoffer HaugsbakkDec 9, 2025
  263. Working on top of mega mergesD. Ben Knoble, Dec 12, 2025
  264. Elijah NewrenNov 23, 2025
  265. Phillip WoodNov 24, 2025
  266. Elijah NewrenNov 25, 2025
  267. 00/12 Introduce git-history(1) command for easy history editingPatrick Steinhardt, Dec 3, 2025
  268. 01/12 wt-status: provide function to expose status for treesPatrick Steinhardt, Dec 3, 2025
  269. 02/12 replay: extract logic to pick commitsPatrick Steinhardt, Dec 3, 2025
  270. 03/12 replay: stop using `the_repository`Patrick Steinhardt, Dec 3, 2025
  271. 04/12 builtin: add new "history" commandPatrick Steinhardt, Dec 3, 2025
  272. Kristoffer HaugsbakkDec 22, 2025
  273. Patrick SteinhardtJan 6, 2026
  274. 05/12 builtin/history: implement "reword" subcommandPatrick Steinhardt, Dec 3, 2025
  275. 06/12 add-patch: split out header from "add-interactive.h"Patrick Steinhardt, Dec 3, 2025
  276. 07/12 add-patch: split out `struct interactive_options`Patrick Steinhardt, Dec 3, 2025
  277. 08/12 add-patch: remove dependency on "add-interactive" subsystemPatrick Steinhardt, Dec 3, 2025
  278. 09/12 add-patch: add support for in-memory index patchingPatrick Steinhardt, Dec 3, 2025
  279. 10/12 add-patch: allow disabling editing of hunksPatrick Steinhardt, Dec 3, 2025
  280. 11/12 cache-tree: allow writing in-memory index as treePatrick Steinhardt, Dec 3, 2025
  281. 12/12 builtin/history: implement "split" subcommandPatrick Steinhardt, Dec 3, 2025
  282. 0/7 Introduce git-history(1) command for easy history editingPatrick Steinhardt, Jan 9, 2026
  283. 1/7 builtin/replay: extract core logic to replay revisionsPatrick Steinhardt, Jan 9, 2026
  284. Elijah NewrenJan 10, 2026
  285. Patrick SteinhardtJan 12, 2026
  286. 2/7 builtin/replay: move core logic into "libgit.a"Patrick Steinhardt, Jan 9, 2026
  287. Elijah NewrenJan 10, 2026
  288. Patrick SteinhardtJan 12, 2026
  289. Elijah NewrenJan 13, 2026
  290. Patrick SteinhardtJan 13, 2026
  291. 3/7 replay: small set of cleanupsPatrick Steinhardt, Jan 9, 2026
  292. 4/7 replay: yield the object ID of the final rewritten commitPatrick Steinhardt, Jan 9, 2026
  293. Elijah NewrenJan 10, 2026
  294. Patrick SteinhardtJan 12, 2026
  295. 5/7 wt-status: provide function to expose status for treesPatrick Steinhardt, Jan 9, 2026
  296. 6/7 builtin: add new "history" commandPatrick Steinhardt, Jan 9, 2026
  297. Elijah NewrenJan 10, 2026
  298. Patrick SteinhardtJan 12, 2026
  299. 7/7 builtin/history: implement "reword" subcommandPatrick Steinhardt, Jan 9, 2026
  300. Elijah NewrenJan 10, 2026
  301. SZEDER GáborJan 10, 2026
  302. Patrick SteinhardtJan 12, 2026
  303. 0/2 SomeElijah Newren, Jan 11, 2026
  304. 8/7 SQUASH ME: FixupsElijah Newren, Jan 11, 2026
  305. 9/7 history: fix detached HEAD handlingElijah Newren, Jan 11, 2026
  306. Elijah NewrenJan 11, 2026
  307. Patrick SteinhardtJan 12, 2026
  308. 0/8 Introduce git-history(1) command for easy history editingPatrick Steinhardt, Jan 13, 2026
  309. 1/8 builtin/replay: extract core logic to replay revisionsPatrick Steinhardt, Jan 13, 2026
  310. 2/8 builtin/replay: move core logic into "libgit.a"Patrick Steinhardt, Jan 13, 2026
  311. 3/8 replay: small set of cleanupsPatrick Steinhardt, Jan 13, 2026
  312. 4/8 replay: support empty commit rangesPatrick Steinhardt, Jan 13, 2026
  313. 5/8 replay: support updating detached HEADPatrick Steinhardt, Jan 13, 2026
  314. 6/8 wt-status: provide function to expose status for treesPatrick Steinhardt, Jan 13, 2026
  315. 7/8 builtin: add new "history" commandPatrick Steinhardt, Jan 13, 2026
  316. 8/8 builtin/history: implement "reword" subcommandPatrick Steinhardt, Jan 13, 2026
  317. SZEDER GáborJan 16, 2026
  318. Elijah NewrenJan 17, 2026
  319. Junio C HamanoJan 18, 2026
  320. Junio C HamanoFeb 3, 2026
  321. Patrick SteinhardtFeb 5, 2026
  322. Elijah NewrenJan 16, 2026
  323. Junio C HamanoJan 16, 2026
  324. SZEDER GáborJan 16, 2026
  325. Junio C HamanoJan 21, 2026

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.