{"thread":{"id":"65771","subject":"[PATCH 0/3] Teach git-replay(1) to linearize merge commits","startedAt":"2026-06-08T18:37:45Z","lastAt":"2026-09-01T22:15:34Z","messageCount":75,"participants":["Toon Claes","Junio C Hamano","Justin Tobler","Elijah Newren","Patrick Steinhardt","Phillip Wood","Johannes Schindelin"],"isPatch":true,"patchVersion":1,"patchTotal":3},"messages":[{"id":"544955","messageId":"20260608-toon-git-replay-drop-merges-v1-0-e3ee71fce7b4@iotcl.com","threadId":"65771","inReplyTo":null,"subject":"[PATCH 0/3] Teach git-replay(1) to linearize merge commits","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-08T18:37:18Z","receivedAt":"2026-06-08T18:37:45Z","isPatch":true,"body":"As an alternative to dscho's patch series to replay merges[1], add\noption to git-replay(1) to linearize merges. This mimics wath\ngit-rebase(1) does too with --no-rebase-merges (the default).\n\nThe first two patches do some refactoring. The third patch implements\nthe actual change. I was kindly helped by dscho to implement this\nchange.\n\nThe --linearize option is only added to git-replay(1) and not to\ngit-history(1) because in my opinion doesn't make much sense to do so,\nbut I'm happy to hear if anyone disagrees.\n\nThis series might conflict with Kristoffer's series to make\ndocumentation changes[2], but should be trivial to resolve. And I don't\nthink there's a conflict with Patrick's series on adding \"drop\" to\ngit-history(1)[3].\n\ndscho's series to replay merges[1] need a bit of rework to fit on top of\nthis, but I'm happy to help figuring that out.\n\n[1]: \n[2]: \n[3]: <20260603-b4-pks-history-drop-v2-0-742cb5b5176d@pks.im>\n\nSigned-off-by: Toon Claes \n---\nJohannes Schindelin (1):\n replay: offer an option to linearize the commit topology\n\nToon Claes (2):\n replay: refactor enum replay_mode into a bool\n replay: add helper to put entry into mapped_commits\n\n Documentation/git-replay.adoc | 5 ++\n builtin/replay.c | 4 ++\n replay.c | 109 +++++++++++++++++++++++-------------------\n replay.h | 5 ++\n t/t3650-replay-basics.sh | 22 +++++++++\n 5 files changed, 97 insertions(+), 48 deletions(-)\n\n\n\n---\nbase-commit: 9ac3f193c05c2237e2b14ebaa1149e9fc8a1abe0\nchange-id: 20260604-toon-git-replay-drop-merges-807fa008d395\n\n"},{"id":"544956","messageId":"20260608-toon-git-replay-drop-merges-v1-1-e3ee71fce7b4@iotcl.com","threadId":"65771","inReplyTo":"20260608-toon-git-replay-drop-merges-v1-0-e3ee71fce7b4@iotcl.com","subject":"[PATCH 1/3] replay: refactor enum replay_mode into a bool","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-08T18:37:19Z","receivedAt":"2026-06-08T18:37:53Z","isPatch":true,"body":"In 2760ee4983 (replay: add --revert mode to reverse commit changes,\n2026-03-26) the enum `replay_mode` was introduced. This has two possible\nvalues:\n\n - The value `REPLAY_MODE_REVERT` is used when option `--revert` is\n passed to git-replay(1). When using this value the commits are\n possible in reverse order and the inverse of the changes are applied.\n\n - The value `REPLAY_MODE_PICK` is used when either option `--onto` or\n `--advance` is used. In both cases the commits are pocessed in normal\n order, and the changes are applied as-is.\n\nSince there are only two possible values of this enum, simplify the code\nby converting the enum into a bool. This avoid adding code paths that\ncheck for invalid vaues of the enum, and shortens code where the value\nis checked with a ternary operator.\n\nSigned-off-by: Toon Claes \n---\n replay.c | 59 +++++++++++++++++++++++++----------------------------------\n 1 file changed, 25 insertions(+), 34 deletions(-)\n\ndiff --git a/replay.c b/replay.c\nindex 4ef8abb607..1f8e5b083b 100644\n--- a/replay.c\n+++ b/replay.c\n@@ -18,11 +18,6 @@\n */\n #define the_repository DO_NOT_USE_THE_REPOSITORY\n \n-enum replay_mode {\n-\tREPLAY_MODE_PICK,\n-\tREPLAY_MODE_REVERT,\n-};\n-\n static const char *short_commit_name(struct repository *repo,\n \t\t\t\t struct commit *commit)\n {\n@@ -81,7 +76,7 @@ static struct commit *create_commit(struct repository *repo,\n \t\t\t\t struct tree *tree,\n \t\t\t\t struct commit *based_on,\n \t\t\t\t struct commit *parent,\n-\t\t\t\t enum replay_mode mode)\n+\t\t\t\t bool reverse)\n {\n \tstruct object_id ret;\n \tstruct object *obj = NULL;\n@@ -98,15 +93,13 @@ static struct commit *create_commit(struct repository *repo,\n \n \tcommit_list_insert(parent, &parents);\n \textra = read_commit_extra_headers(based_on, exclude_gpgsig);\n-\tif (mode == REPLAY_MODE_REVERT) {\n+\tif (reverse) {\n \t\tgenerate_revert_message(&msg, based_on, repo);\n \t\t/* For revert, use current user as author (NULL = use default) */\n-\t} else if (mode == REPLAY_MODE_PICK) {\n+\t} else {\n \t\tfind_commit_subject(message, &orig_message);\n \t\tstrbuf_addstr(&msg, orig_message);\n \t\tauthor = get_author(message);\n-\t} else {\n-\t\tBUG(\"unexpected replay mode %d\", mode);\n \t}\n \treset_ident_date();\n \tif (commit_tree_extended(msg.buf, msg.len, &tree->object.oid, parents,\n@@ -269,7 +262,7 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t struct commit *onto,\n \t\t\t\t\t struct merge_options *merge_opt,\n \t\t\t\t\t struct merge_result *result,\n-\t\t\t\t\t enum replay_mode mode,\n+\t\t\t\t\t bool reverse,\n \t\t\t\t\t enum replay_empty_commit_action empty)\n {\n \tstruct commit *base, *replayed_base;\n@@ -287,7 +280,21 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \treplayed_base_tree = repo_get_commit_tree(repo, replayed_base);\n \tpickme_tree = repo_get_commit_tree(repo, pickme);\n \n-\tif (mode == REPLAY_MODE_PICK) {\n+\tif (reverse) {\n+\t\t/* Revert: swap base and pickme to reverse the diff */\n+\t\tconst char *pickme_name = short_commit_name(repo, pickme);\n+\t\tmerge_opt->branch1 = short_commit_name(repo, replayed_base);\n+\t\tmerge_opt->branch2 = xstrfmt(\"parent of %s\", pickme_name);\n+\t\tmerge_opt->ancestor = pickme_name;\n+\n+\t\tmerge_incore_nonrecursive(merge_opt,\n+\t\t\t\t\t pickme_tree,\n+\t\t\t\t\t replayed_base_tree,\n+\t\t\t\t\t base_tree,\n+\t\t\t\t\t result);\n+\n+\t\tfree((char *)merge_opt->branch2);\n+\t} else {\n \t\t/* Cherry-pick: normal order */\n \t\tmerge_opt->branch1 = short_commit_name(repo, replayed_base);\n \t\tmerge_opt->branch2 = short_commit_name(repo, pickme);\n@@ -303,22 +310,6 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t result);\n \n \t\tfree((char *)merge_opt->ancestor);\n-\t} else if (mode == REPLAY_MODE_REVERT) {\n-\t\t/* Revert: swap base and pickme to reverse the diff */\n-\t\tconst char *pickme_name = short_commit_name(repo, pickme);\n-\t\tmerge_opt->branch1 = short_commit_name(repo, replayed_base);\n-\t\tmerge_opt->branch2 = xstrfmt(\"parent of %s\", pickme_name);\n-\t\tmerge_opt->ancestor = pickme_name;\n-\n-\t\tmerge_incore_nonrecursive(merge_opt,\n-\t\t\t\t\t pickme_tree,\n-\t\t\t\t\t replayed_base_tree,\n-\t\t\t\t\t base_tree,\n-\t\t\t\t\t result);\n-\n-\t\tfree((char *)merge_opt->branch2);\n-\t} else {\n-\t\tBUG(\"unexpected replay mode %d\", mode);\n \t}\n \tmerge_opt->ancestor = NULL;\n \tmerge_opt->branch2 = NULL;\n@@ -341,7 +332,7 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t}\n \t}\n \n-\treturn create_commit(repo, result->tree, pickme, replayed_base, mode);\n+\treturn create_commit(repo, result->tree, pickme, replayed_base, reverse);\n }\n \n void replay_result_release(struct replay_result *result)\n@@ -381,13 +372,13 @@ int replay_revisions(struct rev_info *revs,\n \tchar *revert;\n \tconst char *ref;\n \tstruct object_id old_oid;\n-\tenum replay_mode mode = REPLAY_MODE_PICK;\n+\tbool reverse;\n \tint ret;\n \n \tadvance = xstrdup_or_null(opts->advance);\n \trevert = xstrdup_or_null(opts->revert);\n-\tif (revert)\n-\t\tmode = REPLAY_MODE_REVERT;\n+\treverse = !!revert;\n+\n \tset_up_replay_mode(revs->repo, &revs->cmdline, opts->onto,\n \t\t\t &detached_head, &advance, &revert, &onto, &update_refs);\n \n@@ -430,8 +421,8 @@ int replay_revisions(struct rev_info *revs,\n \t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n \n \t\tlast_commit = pick_regular_commit(revs->repo, commit, replayed_commits,\n-\t\t\t\t\t\t mode == REPLAY_MODE_REVERT ? last_commit : onto,\n-\t\t\t\t\t\t &merge_opt, &result, mode, opts->empty);\n+\t\t\t\t\t\t reverse ? last_commit : onto,\n+\t\t\t\t\t\t &merge_opt, &result, reverse, opts->empty);\n \t\tif (!last_commit)\n \t\t\tbreak;\n \n\n-- \n2.53.0.1323.g189a785ab5\n\n"},{"id":"544957","messageId":"20260608-toon-git-replay-drop-merges-v1-2-e3ee71fce7b4@iotcl.com","threadId":"65771","inReplyTo":"20260608-toon-git-replay-drop-merges-v1-0-e3ee71fce7b4@iotcl.com","subject":"[PATCH 2/3] replay: add helper to put entry into mapped_commits","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-08T18:37:20Z","receivedAt":"2026-06-08T18:38:02Z","isPatch":true,"body":"The function replay_revisions() in replay.c is rather lengthy. Extract\nthe logic to put commit entry into mapped_commits into a helper\nfunction.\n\nSigned-off-by: Toon Claes \n---\n replay.c | 31 ++++++++++++++++++++-----------\n 1 file changed, 20 insertions(+), 11 deletions(-)\n\ndiff --git a/replay.c b/replay.c\nindex 1f8e5b083b..7921d7dba3 100644\n--- a/replay.c\n+++ b/replay.c\n@@ -243,9 +243,9 @@ static void set_up_replay_mode(struct repository *repo,\n \tstrset_clear(&rinfo.positive_refs);\n }\n \n-static struct commit *mapped_commit(kh_oid_map_t *replayed_commits,\n-\t\t\t\t struct commit *commit,\n-\t\t\t\t struct commit *fallback)\n+static struct commit *get_mapped_commit(kh_oid_map_t *replayed_commits,\n+\t\t\t\t\tstruct commit *commit,\n+\t\t\t\t\tstruct commit *fallback)\n {\n \tkhint_t pos;\n \tif (!commit)\n@@ -256,6 +256,21 @@ static struct commit *mapped_commit(kh_oid_map_t *replayed_commits,\n \treturn kh_value(replayed_commits, pos);\n }\n \n+static void put_mapped_commit(kh_oid_map_t *replayed_commits,\n+\t\t\t struct commit *commit,\n+\t\t\t struct commit *new_commit)\n+{\n+\tkhint_t pos;\n+\tint ret;\n+\n+\tpos = kh_put_oid_map(replayed_commits, commit->object.oid, &ret);\n+\tif (ret == 0)\n+\t\tBUG(\"Duplicate rewritten commit: %s\\n\",\n+\t\t oid_to_hex(&commit->object.oid));\n+\n+\tkh_value(replayed_commits, pos) = new_commit;\n+}\n+\n static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t struct commit *pickme,\n \t\t\t\t\t kh_oid_map_t *replayed_commits,\n@@ -276,7 +291,7 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\tbase_tree = lookup_tree(repo, repo->hash_algo->empty_tree);\n \t}\n \n-\treplayed_base = mapped_commit(replayed_commits, base, onto);\n+\treplayed_base = get_mapped_commit(replayed_commits, base, onto);\n \treplayed_base_tree = repo_get_commit_tree(repo, replayed_base);\n \tpickme_tree = repo_get_commit_tree(repo, pickme);\n \n@@ -414,8 +429,6 @@ int replay_revisions(struct rev_info *revs,\n \treplayed_commits = kh_init_oid_map();\n \twhile ((commit = get_revision(revs))) {\n \t\tconst struct name_decoration *decoration;\n-\t\tkhint_t pos;\n-\t\tint hr;\n \n \t\tif (commit->parents && commit->parents->next)\n \t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n@@ -427,11 +440,7 @@ int replay_revisions(struct rev_info *revs,\n \t\t\tbreak;\n \n \t\t/* Record commit -> last_commit mapping */\n-\t\tpos = kh_put_oid_map(replayed_commits, commit->object.oid, &hr);\n-\t\tif (hr == 0)\n-\t\t\tBUG(\"Duplicate rewritten commit: %s\\n\",\n-\t\t\t oid_to_hex(&commit->object.oid));\n-\t\tkh_value(replayed_commits, pos) = last_commit;\n+\t\tput_mapped_commit(replayed_commits, commit, last_commit);\n \n \t\t/* Update any necessary branches */\n \t\tif (ref)\n\n-- \n2.53.0.1323.g189a785ab5\n\n"},{"id":"544958","messageId":"20260608-toon-git-replay-drop-merges-v1-3-e3ee71fce7b4@iotcl.com","threadId":"65771","inReplyTo":"20260608-toon-git-replay-drop-merges-v1-0-e3ee71fce7b4@iotcl.com","subject":"[PATCH 3/3] replay: offer an option to linearize the commit topology","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-08T18:37:21Z","receivedAt":"2026-06-08T18:38:05Z","isPatch":true,"body":"From: Johannes Schindelin \n\nOne of the stated goals of git-replay(1) is to allow implementing the\ngit-rebase(1) functionality on the server side.\n\nThe default mode of git-rebase(1) is to act as if `--no-rebase-merges`\nwas given. This mode drops merge commits instead of replaying them, and\nlinearized the commit history into a sequence of the\nregular (single-parent) commits.\n\nAdd option `--linearize` to git-replay(1) do the same.\n\nCo-authored-by: Toon Claes \n---\n Documentation/git-replay.adoc | 5 +++++\n builtin/replay.c | 4 ++++\n replay.c | 25 +++++++++++++++++++------\n replay.h | 5 +++++\n t/t3650-replay-basics.sh | 22 ++++++++++++++++++++++\n 5 files changed, 55 insertions(+), 6 deletions(-)\n\ndiff --git a/Documentation/git-replay.adoc b/Documentation/git-replay.adoc\nindex a32f72aead..41c96c7061 100644\n--- a/Documentation/git-replay.adoc\n+++ b/Documentation/git-replay.adoc\n@@ -88,6 +88,11 @@ incompatible with `--contained` (which is a modifier for `--onto` only).\n +\n The default mode can be configured via the `replay.refAction` configuration variable.\n \n+--linearize::\n+\tIn this mode, `git replay` imitates `git rebase --no-rebase-merges`,\n+\ti.e. it cherry-picks only non-merge commits, each one on top of the\n+\tprevious one.\n+\n ::\n \tRange of commits to replay; see \"Specifying Ranges\" in\n \tlinkgit:git-rev-parse[1]. In `--advance=` or\ndiff --git a/builtin/replay.c b/builtin/replay.c\nindex 39e3a86f6c..fedfe46dc6 100644\n--- a/builtin/replay.c\n+++ b/builtin/replay.c\n@@ -111,6 +111,8 @@ int cmd_replay(int argc,\n \t\t\t N_(\"mode\"),\n \t\t\t N_(\"control ref update behavior (update|print)\"),\n \t\t\t PARSE_OPT_NONEG),\n+\t\tOPT_BOOL(0, \"linearize\", &opts.linearize,\n+\t\t\t N_(\"ignore merge commits instead of replaying them\")),\n \t\tOPT_END()\n \t};\n \n@@ -132,6 +134,8 @@ int cmd_replay(int argc,\n \t\t\t\t opts.contained, \"--contained\");\n \tdie_for_incompatible_opt2(!!opts.ref, \"--ref\",\n \t\t\t\t !!opts.contained, \"--contained\");\n+\tdie_for_incompatible_opt2(!!opts.revert, \"--revert\",\n+\t\t\t\t opts.linearize, \"--linearize\");\n \n \t/* Parse ref action mode from command line or config */\n \tref_mode = get_ref_action_mode(repo, ref_action);\ndiff --git a/replay.c b/replay.c\nindex 7921d7dba3..3e36908131 100644\n--- a/replay.c\n+++ b/replay.c\n@@ -277,12 +277,16 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t struct commit *onto,\n \t\t\t\t\t struct merge_options *merge_opt,\n \t\t\t\t\t struct merge_result *result,\n+\t\t\t\t\t struct commit *replayed_base,\n \t\t\t\t\t bool reverse,\n \t\t\t\t\t enum replay_empty_commit_action empty)\n {\n-\tstruct commit *base, *replayed_base;\n+\tstruct commit *base;\n \tstruct tree *pickme_tree, *base_tree, *replayed_base_tree;\n \n+\tif (replayed_base && reverse)\n+\t\tBUG(\"Linearizing commits is not supported when replaying in reverse\");\n+\n \tif (pickme->parents) {\n \t\tbase = pickme->parents->item;\n \t\tbase_tree = repo_get_commit_tree(repo, base);\n@@ -291,7 +295,8 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\tbase_tree = lookup_tree(repo, repo->hash_algo->empty_tree);\n \t}\n \n-\treplayed_base = get_mapped_commit(replayed_commits, base, onto);\n+\tif (!replayed_base)\n+\t\treplayed_base = get_mapped_commit(replayed_commits, base, onto);\n \treplayed_base_tree = repo_get_commit_tree(repo, replayed_base);\n \tpickme_tree = repo_get_commit_tree(repo, pickme);\n \n@@ -430,12 +435,20 @@ int replay_revisions(struct rev_info *revs,\n \twhile ((commit = get_revision(revs))) {\n \t\tconst struct name_decoration *decoration;\n \n-\t\tif (commit->parents && commit->parents->next)\n+\t\tif (opts->linearize && (!commit->parents || commit->parents->next))\n+\t\t\t; /* map current commit to the same as the previous commit */\n+\t\telse if (commit->parents && commit->parents->next)\n \t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n+\t\telse {\n+\t\t\tstruct commit *to_pick = reverse ? last_commit : onto;\n+\t\t\tlast_commit =\n+\t\t\t\tpick_regular_commit(revs->repo, commit,\n+\t\t\t\t\t\t replayed_commits, to_pick,\n+\t\t\t\t\t\t &merge_opt, &result,\n+\t\t\t\t\t\t opts->linearize ? last_commit : NULL,\n+\t\t\t\t\t\t reverse, opts->empty);\n+\t\t}\n \n-\t\tlast_commit = pick_regular_commit(revs->repo, commit, replayed_commits,\n-\t\t\t\t\t\t reverse ? last_commit : onto,\n-\t\t\t\t\t\t &merge_opt, &result, reverse, opts->empty);\n \t\tif (!last_commit)\n \t\t\tbreak;\n \ndiff --git a/replay.h b/replay.h\nindex 1851a07705..07e6fdcca3 100644\n--- a/replay.h\n+++ b/replay.h\n@@ -62,6 +62,11 @@ struct replay_revisions_options {\n \t * Defaults to REPLAY_EMPTY_COMMIT_DROP.\n \t */\n \tenum replay_empty_commit_action empty;\n+\n+\t/*\n+\t * Whether to linearize the commits (i.e. drop merge commits).\n+\t */\n+\tint linearize;\n };\n \n /* This struct is used as an out-parameter by `replay_revisions()`. */\ndiff --git a/t/t3650-replay-basics.sh b/t/t3650-replay-basics.sh\nindex 3353bc4a4d..c781a3bb1b 100755\n--- a/t/t3650-replay-basics.sh\n+++ b/t/t3650-replay-basics.sh\n@@ -565,4 +565,26 @@ test_expect_success '--onto with --ref rejects multiple revision ranges' '\n \ttest_grep \"cannot be used with multiple revision ranges\" err\n '\n \n+test_expect_success 'linearize the commit topology' '\n+\ttest_tick &&\n+\tN=$(git commit-tree -m N -p L -p I L:) &&\n+\tN=$(git commit-tree -m N-child -p $N L:) &&\n+\tgit update-ref refs/heads/N $N &&\n+\n+\tgit replay --ref-action=print --linearize \\\n+\t\t--onto A B..refs/heads/N >out &&\n+\n+\ttest_line_count = 1 out &&\n+\tread N1 N2 N3 N4 expect <<-EOF &&\n+\t* N-child\n+\t* I\n+\t* L\n+\to A\n+\tEOF\n+\tgit log --format=%s --graph --boundary A...$N3 >actual &&\n+\ttest_cmp expect actual\n+'\n+\n test_done\n\n-- \n2.53.0.1323.g189a785ab5\n\n"},{"id":"544964","messageId":"xmqqtsrcvnjw.fsf@gitster.g","threadId":"65771","inReplyTo":"20260608-toon-git-replay-drop-merges-v1-3-e3ee71fce7b4@iotcl.com","subject":"Re: [PATCH 3/3] replay: offer an option to linearize the commit topology","fromName":"Junio C Hamano","fromEmail":"gitster@pobox.com","sentAt":"2026-06-08T19:29:07Z","receivedAt":"2026-06-08T19:29:10Z","isPatch":true,"body":"Toon Claes writes:\n\n> From: Johannes Schindelin \n>\n> One of the stated goals of git-replay(1) is to allow implementing the\n> git-rebase(1) functionality on the server side.\n>\n> The default mode of git-rebase(1) is to act as if `--no-rebase-merges`\n> was given. This mode drops merge commits instead of replaying them, and\n> linearized the commit history into a sequence of the\n> regular (single-parent) commits.\n\n\"linearized\" -> \"linearizes\"?\n\n>\n> Add option `--linearize` to git-replay(1) do the same.\n\n\"do the same\" -> \"to do the same\"?\n\n> Co-authored-by: Toon Claes \n\nThere is no sign-off by any of the authors?\n\n> @@ -430,12 +435,20 @@ int replay_revisions(struct rev_info *revs,\n> \twhile ((commit = get_revision(revs))) {\n> \t\tconst struct name_decoration *decoration;\n> \n> -\t\tif (commit->parents && commit->parents->next)\n> +\t\tif (opts->linearize && (!commit->parents || commit->parents->next))\n> +\t\t\t; /* map current commit to the same as the previous commit */\n\nThis uses the same treatment on either root commits or merge\ncommits? If this were a mistake and this wants to handle merges but\nnot roots, shouldn't it be more like\n\n\t\tif (opts->linearize && (commit->parents && commit->parents->next))\n\t\t\t; /* map the merge to the previous */\n\n> +\t\telse if (commit->parents && commit->parents->next)\n> \t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n\nAnd because the next one is also about merges, perhaps the early\npart of this if/else if cascade can be written\n\n\t\tif (commit->parents && commit->parents->next) {\n\t\t\t/* We have a merge */\n\t\t\tif (!opts->linearize)\n\t\t\t\tdie(_(\"can't replay a merge (yet)\"));\n\t\t\t; /* map current to the previous */\n\t\t} else {\n\t\t\t...\n\nwouldn't it?\n\nIf the \"map current to prev\" is applicable to root, any root are\nmapped to the last_commit in the above, and if we saw a root as the\nfirst thing in the loop, last_commit is NULL, we do not do anything\nhere, and after the if/else if/else cascade, we see last_commit is\nNULL and break out of the loop.\n\n> +\t\telse {\n> +\t\t\tstruct commit *to_pick = reverse ? last_commit : onto;\n> +\t\t\tlast_commit =\n> +\t\t\t\tpick_regular_commit(revs->repo, commit,\n> +\t\t\t\t\t\t replayed_commits, to_pick,\n> +\t\t\t\t\t\t &merge_opt, &result,\n> +\t\t\t\t\t\t opts->linearize ? last_commit : NULL,\n> +\t\t\t\t\t\t reverse, opts->empty);\n> +\t\t}\n> \n> -\t\tlast_commit = pick_regular_commit(revs->repo, commit, replayed_commits,\n> -\t\t\t\t\t\t reverse ? last_commit : onto,\n> -\t\t\t\t\t\t &merge_opt, &result, reverse, opts->empty);\n> \t\tif (!last_commit)\n> \t\t\tbreak;\n\n> diff --git a/replay.h b/replay.h\n> index 1851a07705..07e6fdcca3 100644\n> --- a/replay.h\n> +++ b/replay.h\n> @@ -62,6 +62,11 @@ struct replay_revisions_options {\n> \t * Defaults to REPLAY_EMPTY_COMMIT_DROP.\n> \t */\n> \tenum replay_empty_commit_action empty;\n> +\n> +\t/*\n> +\t * Whether to linearize the commits (i.e. drop merge commits).\n> +\t */\n> +\tint linearize;\n> };\n> \n> /* This struct is used as an out-parameter by `replay_revisions()`. */\n> diff --git a/t/t3650-replay-basics.sh b/t/t3650-replay-basics.sh\n> index 3353bc4a4d..c781a3bb1b 100755\n> --- a/t/t3650-replay-basics.sh\n> +++ b/t/t3650-replay-basics.sh\n> @@ -565,4 +565,26 @@ test_expect_success '--onto with --ref rejects multiple revision ranges' '\n> \ttest_grep \"cannot be used with multiple revision ranges\" err\n> '\n> \n> +test_expect_success 'linearize the commit topology' '\n> +\ttest_tick &&\n> +\tN=$(git commit-tree -m N -p L -p I L:) &&\n> +\tN=$(git commit-tree -m N-child -p $N L:) &&\n> +\tgit update-ref refs/heads/N $N &&\n> +\n> +\tgit replay --ref-action=print --linearize \\\n> +\t\t--onto A B..refs/heads/N >out &&\n> +\n> +\ttest_line_count = 1 out &&\n> +\tread N1 N2 N3 N4 +\n> +\tcat >expect <<-EOF &&\n> +\t* N-child\n> +\t* I\n> +\t* L\n> +\to A\n> +\tEOF\n> +\tgit log --format=%s --graph --boundary A...$N3 >actual &&\n> +\ttest_cmp expect actual\n> +'\n\nPerhaps we would want to have a test that replays all the way down\nto the root commit?\n"},{"id":"545146","messageId":"87zf12zd33.fsf@emacs.iotcl.com","threadId":"65771","inReplyTo":"xmqqtsrcvnjw.fsf@gitster.g","subject":"Re: [PATCH 3/3] replay: offer an option to linearize the commit topology","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-10T14:26:08Z","receivedAt":"2026-06-10T14:26:27Z","isPatch":true,"body":"Junio C Hamano writes:\n\n> Toon Claes writes:\n>\n>> From: Johannes Schindelin \n>>\n>> One of the stated goals of git-replay(1) is to allow implementing the\n>> git-rebase(1) functionality on the server side.\n>>\n>> The default mode of git-rebase(1) is to act as if `--no-rebase-merges`\n>> was given. This mode drops merge commits instead of replaying them, and\n>> linearized the commit history into a sequence of the\n>> regular (single-parent) commits.\n>\n> \"linearized\" -> \"linearizes\"?\n\nThanks.\n\n>>\n>> Add option `--linearize` to git-replay(1) do the same.\n>\n> \"do the same\" -> \"to do the same\"?\n\nAck.\n\n>> Co-authored-by: Toon Claes \n>\n> There is no sign-off by any of the authors?\n\nMy bad. I'll add mine.\n\n@Johannes, can I re-add yours? I've removed it because I've made some\nchanges on top of the patch you wrote, but if you agree, I'll add your\nSign-off back.\n\n>> @@ -430,12 +435,20 @@ int replay_revisions(struct rev_info *revs,\n>> \twhile ((commit = get_revision(revs))) {\n>> \t\tconst struct name_decoration *decoration;\n>> \n>> -\t\tif (commit->parents && commit->parents->next)\n>> +\t\tif (opts->linearize && (!commit->parents || commit->parents->next))\n>> +\t\t\t; /* map current commit to the same as the previous commit */\n>\n> This uses the same treatment on either root commits or merge\n> commits? If this were a mistake and this wants to handle merges but\n> not roots, shouldn't it be more like\n>\n> \t\tif (opts->linearize && (commit->parents && commit->parents->next))\n> \t\t\t; /* map the merge to the previous */\n>\n>> +\t\telse if (commit->parents && commit->parents->next)\n>> \t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n>\n> And because the next one is also about merges, perhaps the early\n> part of this if/else if cascade can be written\n>\n> \t\tif (commit->parents && commit->parents->next) {\n> \t\t\t/* We have a merge */\n> \t\t\tif (!opts->linearize)\n> \t\t\t\tdie(_(\"can't replay a merge (yet)\"));\n> \t\t\t; /* map current to the previous */\n> \t\t} else {\n> \t\t\t...\n>\n> wouldn't it?\n\nThe way it was written in v1 was maybe a bit too smart and hard to\nfollow. I agree with your suggestion and will adopt this (with some\ntweaks) in the next version.\n\n> If the \"map current to prev\" is applicable to root, any root are\n> mapped to the last_commit in the above, and if we saw a root as the\n> first thing in the loop, last_commit is NULL, we do not do anything\n> here, and after the if/else if/else cascade, we see last_commit is\n> NULL and break out of the loop.\n\nYes, good observation. I did not test this.\n\n> Perhaps we would want to have a test that replays all the way down\n> to the root commit?\n\nI'll add it.\n\n-- \nCheers,\nToon\n"},{"id":"545148","messageId":"20260610-toon-git-replay-drop-merges-v2-0-5714a71c6d83@iotcl.com","threadId":"65771","inReplyTo":"20260608-toon-git-replay-drop-merges-v1-0-e3ee71fce7b4@iotcl.com","subject":"[PATCH v2 0/3] Teach git-replay(1) to linearize merge commits","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-10T14:49:11Z","receivedAt":"2026-06-10T14:49:26Z","isPatch":true,"body":"As an alternative to dscho's patch series to replay merges[1], add\noption to git-replay(1) to linearize merges. This mimics wath\ngit-rebase(1) does too with --no-rebase-merges (the default).\n\nThe first two patches do some refactoring. The third patch implements\nthe actual change. I was kindly helped by dscho to implement this\nchange.\n\nThe --linearize option is only added to git-replay(1) and not to\ngit-history(1) because in my opinion doesn't make much sense to do so,\nbut I'm happy to hear if anyone disagrees.\n\nThis series might conflict with Kristoffer's series to make\ndocumentation changes[2], but should be trivial to resolve. And I don't\nthink there's a conflict with Patrick's series on adding \"drop\" to\ngit-history(1)[3].\n\ndscho's series to replay merges[1] need a bit of rework to fit on top of\nthis, but I'm happy to help figuring that out. We've been discussing to\neither name the option --flatten or --linearize, but I've decided on\n\"linearize\" because the documentation of git-rebase(1) also mentions\n\"linearize\".\n\n[1]: \n[2]: \n[3]: <20260603-b4-pks-history-drop-v2-0-742cb5b5176d@pks.im>\n\nSigned-off-by: Toon Claes \n---\nChanges in v2:\n- Restructured the conditions to detect merge commits and added a line\n of comment why the loop continues.\n- Rewrote tests to use the history from the setup step and added a few\n test cases.\n- Re-added Johannes's Signed-off-by trailer. Johannes gave me the\n patches with this trailer, and if I understand correctly, I can keep\n it. Please let me know if that wrong.\n- Link to v1: https://patch.msgid.link/20260608-toon-git-replay-drop-merges-v1-0-e3ee71fce7b4@iotcl.com\n\n---\nJohannes Schindelin (1):\n replay: offer an option to linearize the commit topology\n\nToon Claes (2):\n replay: refactor enum replay_mode into a bool\n replay: add helper to put entry into mapped_commits\n\n Documentation/git-replay.adoc | 5 ++\n builtin/replay.c | 4 ++\n replay.c | 114 ++++++++++++++++++++++++------------------\n replay.h | 5 ++\n t/t3650-replay-basics.sh | 26 ++++++++++\n 5 files changed, 105 insertions(+), 49 deletions(-)\n\nRange-diff versus v1:\n\n1: 7f3bc6f425 ! 1: 0975b142e3 replay: refactor enum replay_mode into a bool\n @@ Commit message\n \n - The value `REPLAY_MODE_REVERT` is used when option `--revert` is\n passed to git-replay(1). When using this value the commits are\n - possible in reverse order and the inverse of the changes are applied.\n + processed in reverse order and the inverse of the changes are\n + applied.\n \n - The value `REPLAY_MODE_PICK` is used when either option `--onto` or\n - `--advance` is used. In both cases the commits are pocessed in normal\n - order, and the changes are applied as-is.\n + `--advance` is used. In both cases the commits are processed in\n + normal order, and the changes are applied as-is.\n \n Since there are only two possible values of this enum, simplify the code\n - by converting the enum into a bool. This avoid adding code paths that\n - check for invalid vaues of the enum, and shortens code where the value\n + by converting the enum into a bool. This avoids adding code paths that\n + check for invalid values of the enum, and shortens code where the value\n is checked with a ternary operator.\n \n Signed-off-by: Toon Claes \n2: 0868871c78 ! 2: db88193624 replay: add helper to put entry into mapped_commits\n @@ Commit message\n replay: add helper to put entry into mapped_commits\n \n The function replay_revisions() in replay.c is rather lengthy. Extract\n - the logic to put commit entry into mapped_commits into a helper\n - function.\n + the logic to put a commit entry into mapped_commits into a helper\n + function put_mapped_commit().\n +\n + While at it, rename mapped_commit() to get_mapped_commit() to pair with\n + this new function.\n \n Signed-off-by: Toon Claes \n \n3: a432ae753b ! 3: d0c220ec8e replay: offer an option to linearize the commit topology\n @@ Commit message\n \n The default mode of git-rebase(1) is to act as if `--no-rebase-merges`\n was given. This mode drops merge commits instead of replaying them, and\n - linearized the commit history into a sequence of the\n + linearizes the commit history into a sequence of the\n regular (single-parent) commits.\n \n - Add option `--linearize` to git-replay(1) do the same.\n + Add option `--linearize` to git-replay(1) to do the same.\n \n Co-authored-by: Toon Claes \n + Signed-off-by: Johannes Schindelin \n + Signed-off-by: Toon Claes \n \n ## Documentation/git-replay.adoc ##\n @@ Documentation/git-replay.adoc: incompatible with `--contained` (which is a modifier for `--onto` only).\n @@ replay.c: int replay_revisions(struct rev_info *revs,\n \t\tconst struct name_decoration *decoration;\n \n -\t\tif (commit->parents && commit->parents->next)\n -+\t\tif (opts->linearize && (!commit->parents || commit->parents->next))\n -+\t\t\t; /* map current commit to the same as the previous commit */\n -+\t\telse if (commit->parents && commit->parents->next)\n - \t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n -+\t\telse {\n +-\t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n ++\t\tif (commit->parents && commit->parents->next) {\n ++\t\t\tif (!opts->linearize)\n ++\t\t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n ++\t\t\t/*\n ++\t\t\t * When linearizing, a merge commit itself is not picked,\n ++\t\t\t * but refs that point to it might need updating.\n ++\t\t\t */\n ++\t\t} else {\n +\t\t\tstruct commit *to_pick = reverse ? last_commit : onto;\n +\t\t\tlast_commit =\n +\t\t\t\tpick_regular_commit(revs->repo, commit,\n @@ t/t3650-replay-basics.sh: test_expect_success '--onto with --ref rejects multipl\n \ttest_grep \"cannot be used with multiple revision ranges\" err\n '\n \n -+test_expect_success 'linearize the commit topology' '\n -+\ttest_tick &&\n -+\tN=$(git commit-tree -m N -p L -p I L:) &&\n -+\tN=$(git commit-tree -m N-child -p $N L:) &&\n -+\tgit update-ref refs/heads/N $N &&\n ++test_expect_success 'replay merge commit fails' '\n ++\techo \"fatal: replaying merge commits is not supported yet!\" >expect &&\n ++\ttest_must_fail git replay --ref-action=print --onto main I..P 2>actual &&\n ++\ttest_cmp expect actual\n ++'\n ++\n ++test_expect_success 'replay to rebase merge commit with --linearize' '\n ++\tgit replay --ref-action=print --linearize --onto main I..topic-with-merge >result &&\n ++\n ++\ttest_line_count = 1 result &&\n ++\n ++\tgit log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n ++\ttest_write_lines O N J M L B A >expect &&\n ++\ttest_cmp expect actual\n ++'\n +\n -+\tgit replay --ref-action=print --linearize \\\n -+\t\t--onto A B..refs/heads/N >out &&\n ++test_expect_success 'replay to rebase merge commit with --linearize down to root commit' '\n ++\tgit replay --ref-action=print --linearize --onto main A..topic-with-merge >result &&\n +\n -+\ttest_line_count = 1 out &&\n -+\tread N1 N2 N3 N4 expect <<-EOF &&\n -+\t* N-child\n -+\t* I\n -+\t* L\n -+\to A\n -+\tEOF\n -+\tgit log --format=%s --graph --boundary A...$N3 >actual &&\n ++\tgit log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n ++\ttest_write_lines O N J I M L B A >expect &&\n +\ttest_cmp expect actual\n +'\n +\n\n\n---\nbase-commit: 9ac3f193c05c2237e2b14ebaa1149e9fc8a1abe0\nchange-id: 20260604-toon-git-replay-drop-merges-807fa008d395\n\n"},{"id":"545149","messageId":"20260610-toon-git-replay-drop-merges-v2-1-5714a71c6d83@iotcl.com","threadId":"65771","inReplyTo":"20260610-toon-git-replay-drop-merges-v2-0-5714a71c6d83@iotcl.com","subject":"[PATCH v2 1/3] replay: refactor enum replay_mode into a bool","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-10T14:49:12Z","receivedAt":"2026-06-10T14:49:30Z","isPatch":true,"body":"In 2760ee4983 (replay: add --revert mode to reverse commit changes,\n2026-03-26) the enum `replay_mode` was introduced. This has two possible\nvalues:\n\n - The value `REPLAY_MODE_REVERT` is used when option `--revert` is\n passed to git-replay(1). When using this value the commits are\n processed in reverse order and the inverse of the changes are\n applied.\n\n - The value `REPLAY_MODE_PICK` is used when either option `--onto` or\n `--advance` is used. In both cases the commits are processed in\n normal order, and the changes are applied as-is.\n\nSince there are only two possible values of this enum, simplify the code\nby converting the enum into a bool. This avoids adding code paths that\ncheck for invalid values of the enum, and shortens code where the value\nis checked with a ternary operator.\n\nSigned-off-by: Toon Claes \n---\n replay.c | 59 +++++++++++++++++++++++++----------------------------------\n 1 file changed, 25 insertions(+), 34 deletions(-)\n\ndiff --git a/replay.c b/replay.c\nindex 4ef8abb607..1f8e5b083b 100644\n--- a/replay.c\n+++ b/replay.c\n@@ -18,11 +18,6 @@\n */\n #define the_repository DO_NOT_USE_THE_REPOSITORY\n \n-enum replay_mode {\n-\tREPLAY_MODE_PICK,\n-\tREPLAY_MODE_REVERT,\n-};\n-\n static const char *short_commit_name(struct repository *repo,\n \t\t\t\t struct commit *commit)\n {\n@@ -81,7 +76,7 @@ static struct commit *create_commit(struct repository *repo,\n \t\t\t\t struct tree *tree,\n \t\t\t\t struct commit *based_on,\n \t\t\t\t struct commit *parent,\n-\t\t\t\t enum replay_mode mode)\n+\t\t\t\t bool reverse)\n {\n \tstruct object_id ret;\n \tstruct object *obj = NULL;\n@@ -98,15 +93,13 @@ static struct commit *create_commit(struct repository *repo,\n \n \tcommit_list_insert(parent, &parents);\n \textra = read_commit_extra_headers(based_on, exclude_gpgsig);\n-\tif (mode == REPLAY_MODE_REVERT) {\n+\tif (reverse) {\n \t\tgenerate_revert_message(&msg, based_on, repo);\n \t\t/* For revert, use current user as author (NULL = use default) */\n-\t} else if (mode == REPLAY_MODE_PICK) {\n+\t} else {\n \t\tfind_commit_subject(message, &orig_message);\n \t\tstrbuf_addstr(&msg, orig_message);\n \t\tauthor = get_author(message);\n-\t} else {\n-\t\tBUG(\"unexpected replay mode %d\", mode);\n \t}\n \treset_ident_date();\n \tif (commit_tree_extended(msg.buf, msg.len, &tree->object.oid, parents,\n@@ -269,7 +262,7 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t struct commit *onto,\n \t\t\t\t\t struct merge_options *merge_opt,\n \t\t\t\t\t struct merge_result *result,\n-\t\t\t\t\t enum replay_mode mode,\n+\t\t\t\t\t bool reverse,\n \t\t\t\t\t enum replay_empty_commit_action empty)\n {\n \tstruct commit *base, *replayed_base;\n@@ -287,7 +280,21 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \treplayed_base_tree = repo_get_commit_tree(repo, replayed_base);\n \tpickme_tree = repo_get_commit_tree(repo, pickme);\n \n-\tif (mode == REPLAY_MODE_PICK) {\n+\tif (reverse) {\n+\t\t/* Revert: swap base and pickme to reverse the diff */\n+\t\tconst char *pickme_name = short_commit_name(repo, pickme);\n+\t\tmerge_opt->branch1 = short_commit_name(repo, replayed_base);\n+\t\tmerge_opt->branch2 = xstrfmt(\"parent of %s\", pickme_name);\n+\t\tmerge_opt->ancestor = pickme_name;\n+\n+\t\tmerge_incore_nonrecursive(merge_opt,\n+\t\t\t\t\t pickme_tree,\n+\t\t\t\t\t replayed_base_tree,\n+\t\t\t\t\t base_tree,\n+\t\t\t\t\t result);\n+\n+\t\tfree((char *)merge_opt->branch2);\n+\t} else {\n \t\t/* Cherry-pick: normal order */\n \t\tmerge_opt->branch1 = short_commit_name(repo, replayed_base);\n \t\tmerge_opt->branch2 = short_commit_name(repo, pickme);\n@@ -303,22 +310,6 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t result);\n \n \t\tfree((char *)merge_opt->ancestor);\n-\t} else if (mode == REPLAY_MODE_REVERT) {\n-\t\t/* Revert: swap base and pickme to reverse the diff */\n-\t\tconst char *pickme_name = short_commit_name(repo, pickme);\n-\t\tmerge_opt->branch1 = short_commit_name(repo, replayed_base);\n-\t\tmerge_opt->branch2 = xstrfmt(\"parent of %s\", pickme_name);\n-\t\tmerge_opt->ancestor = pickme_name;\n-\n-\t\tmerge_incore_nonrecursive(merge_opt,\n-\t\t\t\t\t pickme_tree,\n-\t\t\t\t\t replayed_base_tree,\n-\t\t\t\t\t base_tree,\n-\t\t\t\t\t result);\n-\n-\t\tfree((char *)merge_opt->branch2);\n-\t} else {\n-\t\tBUG(\"unexpected replay mode %d\", mode);\n \t}\n \tmerge_opt->ancestor = NULL;\n \tmerge_opt->branch2 = NULL;\n@@ -341,7 +332,7 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t}\n \t}\n \n-\treturn create_commit(repo, result->tree, pickme, replayed_base, mode);\n+\treturn create_commit(repo, result->tree, pickme, replayed_base, reverse);\n }\n \n void replay_result_release(struct replay_result *result)\n@@ -381,13 +372,13 @@ int replay_revisions(struct rev_info *revs,\n \tchar *revert;\n \tconst char *ref;\n \tstruct object_id old_oid;\n-\tenum replay_mode mode = REPLAY_MODE_PICK;\n+\tbool reverse;\n \tint ret;\n \n \tadvance = xstrdup_or_null(opts->advance);\n \trevert = xstrdup_or_null(opts->revert);\n-\tif (revert)\n-\t\tmode = REPLAY_MODE_REVERT;\n+\treverse = !!revert;\n+\n \tset_up_replay_mode(revs->repo, &revs->cmdline, opts->onto,\n \t\t\t &detached_head, &advance, &revert, &onto, &update_refs);\n \n@@ -430,8 +421,8 @@ int replay_revisions(struct rev_info *revs,\n \t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n \n \t\tlast_commit = pick_regular_commit(revs->repo, commit, replayed_commits,\n-\t\t\t\t\t\t mode == REPLAY_MODE_REVERT ? last_commit : onto,\n-\t\t\t\t\t\t &merge_opt, &result, mode, opts->empty);\n+\t\t\t\t\t\t reverse ? last_commit : onto,\n+\t\t\t\t\t\t &merge_opt, &result, reverse, opts->empty);\n \t\tif (!last_commit)\n \t\t\tbreak;\n \n\n-- \n2.53.0.1323.g189a785ab5\n\n"},{"id":"545150","messageId":"20260610-toon-git-replay-drop-merges-v2-2-5714a71c6d83@iotcl.com","threadId":"65771","inReplyTo":"20260610-toon-git-replay-drop-merges-v2-0-5714a71c6d83@iotcl.com","subject":"[PATCH v2 2/3] replay: add helper to put entry into mapped_commits","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-10T14:49:13Z","receivedAt":"2026-06-10T14:49:33Z","isPatch":true,"body":"The function replay_revisions() in replay.c is rather lengthy. Extract\nthe logic to put a commit entry into mapped_commits into a helper\nfunction put_mapped_commit().\n\nWhile at it, rename mapped_commit() to get_mapped_commit() to pair with\nthis new function.\n\nSigned-off-by: Toon Claes \n---\n replay.c | 31 ++++++++++++++++++++-----------\n 1 file changed, 20 insertions(+), 11 deletions(-)\n\ndiff --git a/replay.c b/replay.c\nindex 1f8e5b083b..7921d7dba3 100644\n--- a/replay.c\n+++ b/replay.c\n@@ -243,9 +243,9 @@ static void set_up_replay_mode(struct repository *repo,\n \tstrset_clear(&rinfo.positive_refs);\n }\n \n-static struct commit *mapped_commit(kh_oid_map_t *replayed_commits,\n-\t\t\t\t struct commit *commit,\n-\t\t\t\t struct commit *fallback)\n+static struct commit *get_mapped_commit(kh_oid_map_t *replayed_commits,\n+\t\t\t\t\tstruct commit *commit,\n+\t\t\t\t\tstruct commit *fallback)\n {\n \tkhint_t pos;\n \tif (!commit)\n@@ -256,6 +256,21 @@ static struct commit *mapped_commit(kh_oid_map_t *replayed_commits,\n \treturn kh_value(replayed_commits, pos);\n }\n \n+static void put_mapped_commit(kh_oid_map_t *replayed_commits,\n+\t\t\t struct commit *commit,\n+\t\t\t struct commit *new_commit)\n+{\n+\tkhint_t pos;\n+\tint ret;\n+\n+\tpos = kh_put_oid_map(replayed_commits, commit->object.oid, &ret);\n+\tif (ret == 0)\n+\t\tBUG(\"Duplicate rewritten commit: %s\\n\",\n+\t\t oid_to_hex(&commit->object.oid));\n+\n+\tkh_value(replayed_commits, pos) = new_commit;\n+}\n+\n static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t struct commit *pickme,\n \t\t\t\t\t kh_oid_map_t *replayed_commits,\n@@ -276,7 +291,7 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\tbase_tree = lookup_tree(repo, repo->hash_algo->empty_tree);\n \t}\n \n-\treplayed_base = mapped_commit(replayed_commits, base, onto);\n+\treplayed_base = get_mapped_commit(replayed_commits, base, onto);\n \treplayed_base_tree = repo_get_commit_tree(repo, replayed_base);\n \tpickme_tree = repo_get_commit_tree(repo, pickme);\n \n@@ -414,8 +429,6 @@ int replay_revisions(struct rev_info *revs,\n \treplayed_commits = kh_init_oid_map();\n \twhile ((commit = get_revision(revs))) {\n \t\tconst struct name_decoration *decoration;\n-\t\tkhint_t pos;\n-\t\tint hr;\n \n \t\tif (commit->parents && commit->parents->next)\n \t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n@@ -427,11 +440,7 @@ int replay_revisions(struct rev_info *revs,\n \t\t\tbreak;\n \n \t\t/* Record commit -> last_commit mapping */\n-\t\tpos = kh_put_oid_map(replayed_commits, commit->object.oid, &hr);\n-\t\tif (hr == 0)\n-\t\t\tBUG(\"Duplicate rewritten commit: %s\\n\",\n-\t\t\t oid_to_hex(&commit->object.oid));\n-\t\tkh_value(replayed_commits, pos) = last_commit;\n+\t\tput_mapped_commit(replayed_commits, commit, last_commit);\n \n \t\t/* Update any necessary branches */\n \t\tif (ref)\n\n-- \n2.53.0.1323.g189a785ab5\n\n"},{"id":"545151","messageId":"20260610-toon-git-replay-drop-merges-v2-3-5714a71c6d83@iotcl.com","threadId":"65771","inReplyTo":"20260610-toon-git-replay-drop-merges-v2-0-5714a71c6d83@iotcl.com","subject":"[PATCH v2 3/3] replay: offer an option to linearize the commit topology","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-10T14:49:14Z","receivedAt":"2026-06-10T14:50:16Z","isPatch":true,"body":"From: Johannes Schindelin \n\nOne of the stated goals of git-replay(1) is to allow implementing the\ngit-rebase(1) functionality on the server side.\n\nThe default mode of git-rebase(1) is to act as if `--no-rebase-merges`\nwas given. This mode drops merge commits instead of replaying them, and\nlinearizes the commit history into a sequence of the\nregular (single-parent) commits.\n\nAdd option `--linearize` to git-replay(1) to do the same.\n\nCo-authored-by: Toon Claes \nSigned-off-by: Johannes Schindelin \nSigned-off-by: Toon Claes \n---\n Documentation/git-replay.adoc | 5 +++++\n builtin/replay.c | 4 ++++\n replay.c | 30 +++++++++++++++++++++++-------\n replay.h | 5 +++++\n t/t3650-replay-basics.sh | 26 ++++++++++++++++++++++++++\n 5 files changed, 63 insertions(+), 7 deletions(-)\n\ndiff --git a/Documentation/git-replay.adoc b/Documentation/git-replay.adoc\nindex a32f72aead..41c96c7061 100644\n--- a/Documentation/git-replay.adoc\n+++ b/Documentation/git-replay.adoc\n@@ -88,6 +88,11 @@ incompatible with `--contained` (which is a modifier for `--onto` only).\n +\n The default mode can be configured via the `replay.refAction` configuration variable.\n \n+--linearize::\n+\tIn this mode, `git replay` imitates `git rebase --no-rebase-merges`,\n+\ti.e. it cherry-picks only non-merge commits, each one on top of the\n+\tprevious one.\n+\n ::\n \tRange of commits to replay; see \"Specifying Ranges\" in\n \tlinkgit:git-rev-parse[1]. In `--advance=` or\ndiff --git a/builtin/replay.c b/builtin/replay.c\nindex 39e3a86f6c..fedfe46dc6 100644\n--- a/builtin/replay.c\n+++ b/builtin/replay.c\n@@ -111,6 +111,8 @@ int cmd_replay(int argc,\n \t\t\t N_(\"mode\"),\n \t\t\t N_(\"control ref update behavior (update|print)\"),\n \t\t\t PARSE_OPT_NONEG),\n+\t\tOPT_BOOL(0, \"linearize\", &opts.linearize,\n+\t\t\t N_(\"ignore merge commits instead of replaying them\")),\n \t\tOPT_END()\n \t};\n \n@@ -132,6 +134,8 @@ int cmd_replay(int argc,\n \t\t\t\t opts.contained, \"--contained\");\n \tdie_for_incompatible_opt2(!!opts.ref, \"--ref\",\n \t\t\t\t !!opts.contained, \"--contained\");\n+\tdie_for_incompatible_opt2(!!opts.revert, \"--revert\",\n+\t\t\t\t opts.linearize, \"--linearize\");\n \n \t/* Parse ref action mode from command line or config */\n \tref_mode = get_ref_action_mode(repo, ref_action);\ndiff --git a/replay.c b/replay.c\nindex 7921d7dba3..81033fb889 100644\n--- a/replay.c\n+++ b/replay.c\n@@ -277,12 +277,16 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t struct commit *onto,\n \t\t\t\t\t struct merge_options *merge_opt,\n \t\t\t\t\t struct merge_result *result,\n+\t\t\t\t\t struct commit *replayed_base,\n \t\t\t\t\t bool reverse,\n \t\t\t\t\t enum replay_empty_commit_action empty)\n {\n-\tstruct commit *base, *replayed_base;\n+\tstruct commit *base;\n \tstruct tree *pickme_tree, *base_tree, *replayed_base_tree;\n \n+\tif (replayed_base && reverse)\n+\t\tBUG(\"Linearizing commits is not supported when replaying in reverse\");\n+\n \tif (pickme->parents) {\n \t\tbase = pickme->parents->item;\n \t\tbase_tree = repo_get_commit_tree(repo, base);\n@@ -291,7 +295,8 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\tbase_tree = lookup_tree(repo, repo->hash_algo->empty_tree);\n \t}\n \n-\treplayed_base = get_mapped_commit(replayed_commits, base, onto);\n+\tif (!replayed_base)\n+\t\treplayed_base = get_mapped_commit(replayed_commits, base, onto);\n \treplayed_base_tree = repo_get_commit_tree(repo, replayed_base);\n \tpickme_tree = repo_get_commit_tree(repo, pickme);\n \n@@ -430,12 +435,23 @@ int replay_revisions(struct rev_info *revs,\n \twhile ((commit = get_revision(revs))) {\n \t\tconst struct name_decoration *decoration;\n \n-\t\tif (commit->parents && commit->parents->next)\n-\t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n+\t\tif (commit->parents && commit->parents->next) {\n+\t\t\tif (!opts->linearize)\n+\t\t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n+\t\t\t/*\n+\t\t\t * When linearizing, a merge commit itself is not picked,\n+\t\t\t * but refs that point to it might need updating.\n+\t\t\t */\n+\t\t} else {\n+\t\t\tstruct commit *to_pick = reverse ? last_commit : onto;\n+\t\t\tlast_commit =\n+\t\t\t\tpick_regular_commit(revs->repo, commit,\n+\t\t\t\t\t\t replayed_commits, to_pick,\n+\t\t\t\t\t\t &merge_opt, &result,\n+\t\t\t\t\t\t opts->linearize ? last_commit : NULL,\n+\t\t\t\t\t\t reverse, opts->empty);\n+\t\t}\n \n-\t\tlast_commit = pick_regular_commit(revs->repo, commit, replayed_commits,\n-\t\t\t\t\t\t reverse ? last_commit : onto,\n-\t\t\t\t\t\t &merge_opt, &result, reverse, opts->empty);\n \t\tif (!last_commit)\n \t\t\tbreak;\n \ndiff --git a/replay.h b/replay.h\nindex 1851a07705..07e6fdcca3 100644\n--- a/replay.h\n+++ b/replay.h\n@@ -62,6 +62,11 @@ struct replay_revisions_options {\n \t * Defaults to REPLAY_EMPTY_COMMIT_DROP.\n \t */\n \tenum replay_empty_commit_action empty;\n+\n+\t/*\n+\t * Whether to linearize the commits (i.e. drop merge commits).\n+\t */\n+\tint linearize;\n };\n \n /* This struct is used as an out-parameter by `replay_revisions()`. */\ndiff --git a/t/t3650-replay-basics.sh b/t/t3650-replay-basics.sh\nindex 3353bc4a4d..64e0731188 100755\n--- a/t/t3650-replay-basics.sh\n+++ b/t/t3650-replay-basics.sh\n@@ -565,4 +565,30 @@ test_expect_success '--onto with --ref rejects multiple revision ranges' '\n \ttest_grep \"cannot be used with multiple revision ranges\" err\n '\n \n+test_expect_success 'replay merge commit fails' '\n+\techo \"fatal: replaying merge commits is not supported yet!\" >expect &&\n+\ttest_must_fail git replay --ref-action=print --onto main I..P 2>actual &&\n+\ttest_cmp expect actual\n+'\n+\n+test_expect_success 'replay to rebase merge commit with --linearize' '\n+\tgit replay --ref-action=print --linearize --onto main I..topic-with-merge >result &&\n+\n+\ttest_line_count = 1 result &&\n+\n+\tgit log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n+\ttest_write_lines O N J M L B A >expect &&\n+\ttest_cmp expect actual\n+'\n+\n+test_expect_success 'replay to rebase merge commit with --linearize down to root commit' '\n+\tgit replay --ref-action=print --linearize --onto main A..topic-with-merge >result &&\n+\n+\ttest_line_count = 1 result &&\n+\n+\tgit log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n+\ttest_write_lines O N J I M L B A >expect &&\n+\ttest_cmp expect actual\n+'\n+\n test_done\n\n-- \n2.53.0.1323.g189a785ab5\n\n"},{"id":"545172","messageId":"xmqqjys6wcpo.fsf@gitster.g","threadId":"65771","inReplyTo":"20260610-toon-git-replay-drop-merges-v2-3-5714a71c6d83@iotcl.com","subject":"Re: [PATCH v2 3/3] replay: offer an option to linearize the commit topology","fromName":"Junio C Hamano","fromEmail":"gitster@pobox.com","sentAt":"2026-06-10T17:02:27Z","receivedAt":"2026-06-10T17:02:29Z","isPatch":true,"body":"Toon Claes writes:\n\n> From: Johannes Schindelin \n>\n> One of the stated goals of git-replay(1) is to allow implementing the\n> git-rebase(1) functionality on the server side.\n>\n> The default mode of git-rebase(1) is to act as if `--no-rebase-merges`\n> was given. This mode drops merge commits instead of replaying them, and\n> linearizes the commit history into a sequence of the\n> regular (single-parent) commits.\n>\n> Add option `--linearize` to git-replay(1) to do the same.\n>\n> Co-authored-by: Toon Claes \n> Signed-off-by: Johannes Schindelin \n> Signed-off-by: Toon Claes \n> ---\n> Documentation/git-replay.adoc | 5 +++++\n> builtin/replay.c | 4 ++++\n> replay.c | 30 +++++++++++++++++++++++-------\n> replay.h | 5 +++++\n> t/t3650-replay-basics.sh | 26 ++++++++++++++++++++++++++\n> 5 files changed, 63 insertions(+), 7 deletions(-)\n>\n> @@ -430,12 +435,23 @@ int replay_revisions(struct rev_info *revs,\n> \twhile ((commit = get_revision(revs))) {\n> \t\tconst struct name_decoration *decoration;\n> \n> -\t\tif (commit->parents && commit->parents->next)\n> -\t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n> +\t\tif (commit->parents && commit->parents->next) {\n> +\t\t\tif (!opts->linearize)\n> +\t\t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n> +\t\t\t/*\n> +\t\t\t * When linearizing, a merge commit itself is not picked,\n> +\t\t\t * but refs that point to it might need updating.\n> +\t\t\t */\n\nIn the review response during the previous iteration, I commented\nthat (1) the original excluded only merges, but (2) your version\nexcluded both merges and the root commits the same way. Your\nresponse was:\n\n The way it was written in v1 was maybe a bit too smart and hard to\n follow. I agree with your suggestion and will adopt this (with some\n tweaks) in the next version.\n\nwhich I took as saying \"it may be confusing, but it correctly\nexpresses what we want to do\", meaning \"yes, roots and merges should\nbe handled the same way\". But the above no longer treats roots the\nsame way as merges. I think that is intended, but just wanted to\ndouble check.\n\n> diff --git a/replay.h b/replay.h\n> index 1851a07705..07e6fdcca3 100644\n> --- a/replay.h\n> +++ b/replay.h\n> @@ -62,6 +62,11 @@ struct replay_revisions_options {\n> \t * Defaults to REPLAY_EMPTY_COMMIT_DROP.\n> \t */\n> \tenum replay_empty_commit_action empty;\n> +\n> +\t/*\n> +\t * Whether to linearize the commits (i.e. drop merge commits).\n> +\t */\n> +\tint linearize;\n> };\n\nOK.\n\n> diff --git a/t/t3650-replay-basics.sh b/t/t3650-replay-basics.sh\n> index 3353bc4a4d..64e0731188 100755\n> --- a/t/t3650-replay-basics.sh\n> +++ b/t/t3650-replay-basics.sh\n> @@ -565,4 +565,30 @@ test_expect_success '--onto with --ref rejects multiple revision ranges' '\n> \ttest_grep \"cannot be used with multiple revision ranges\" err\n> '\n> \n> +test_expect_success 'replay merge commit fails' '\n> +\techo \"fatal: replaying merge commits is not supported yet!\" >expect &&\n> +\ttest_must_fail git replay --ref-action=print --onto main I..P 2>actual &&\n> +\ttest_cmp expect actual\n> +'\n> +\n> +test_expect_success 'replay to rebase merge commit with --linearize' '\n> +\tgit replay --ref-action=print --linearize --onto main I..topic-with-merge >result &&\n> +\n> +\ttest_line_count = 1 result &&\n> +\n> +\tgit log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n> +\ttest_write_lines O N J M L B A >expect &&\n> +\ttest_cmp expect actual\n> +'\n> +\n> +test_expect_success 'replay to rebase merge commit with --linearize down to root commit' '\n> +\tgit replay --ref-action=print --linearize --onto main A..topic-with-merge >result &&\n\nAs with other test pieces, this \"git replay\" command line is overly\nlong and hides the important bit which is that the range being\nreplayed is *not* actually down to the root, which is A (it excludes\nA). Intended?\n\n> +\n> +\ttest_line_count = 1 result &&\n> +\n> +\tgit log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n> +\ttest_write_lines O N J I M L B A >expect &&\n> +\ttest_cmp expect actual\n> +'\n> +\n> test_done\n\nThanks.\n"},{"id":"545286","messageId":"airORumUxyTsN7Bz@denethor","threadId":"65771","inReplyTo":"20260610-toon-git-replay-drop-merges-v2-1-5714a71c6d83@iotcl.com","subject":"Re: [PATCH v2 1/3] replay: refactor enum replay_mode into a bool","fromName":"Justin Tobler","fromEmail":"jltobler@gmail.com","sentAt":"2026-06-11T15:09:13Z","receivedAt":"2026-06-11T15:09:18Z","isPatch":true,"body":"On 26/06/10 04:49PM, Toon Claes wrote:\n> In 2760ee4983 (replay: add --revert mode to reverse commit changes,\n> 2026-03-26) the enum `replay_mode` was introduced. This has two possible\n> values:\n> \n> - The value `REPLAY_MODE_REVERT` is used when option `--revert` is\n> passed to git-replay(1). When using this value the commits are\n> processed in reverse order and the inverse of the changes are\n> applied.\n> \n> - The value `REPLAY_MODE_PICK` is used when either option `--onto` or\n> `--advance` is used. In both cases the commits are processed in\n> normal order, and the changes are applied as-is.\n> \n> Since there are only two possible values of this enum, simplify the code\n> by converting the enum into a bool. This avoids adding code paths that\n> check for invalid values of the enum, and shortens code where the value\n> is checked with a ternary operator.\n\nNaive question: Do we expect there to only be two replay modes in the\nforseeable future? I suppose if other modes were added in the future\nthis change would essentially be reverted.\n\n> Signed-off-by: Toon Claes \n> ---\n> replay.c | 59 +++++++++++++++++++++++++----------------------------------\n> 1 file changed, 25 insertions(+), 34 deletions(-)\n> \n> diff --git a/replay.c b/replay.c\n> index 4ef8abb607..1f8e5b083b 100644\n> --- a/replay.c\n> +++ b/replay.c\n> @@ -18,11 +18,6 @@\n> */\n> #define the_repository DO_NOT_USE_THE_REPOSITORY\n> \n> -enum replay_mode {\n> -\tREPLAY_MODE_PICK,\n> -\tREPLAY_MODE_REVERT,\n> -};\n> -\n> static const char *short_commit_name(struct repository *repo,\n> \t\t\t\t struct commit *commit)\n> {\n> @@ -81,7 +76,7 @@ static struct commit *create_commit(struct repository *repo,\n> \t\t\t\t struct tree *tree,\n> \t\t\t\t struct commit *based_on,\n> \t\t\t\t struct commit *parent,\n> -\t\t\t\t enum replay_mode mode)\n> +\t\t\t\t bool reverse)\n> {\n> \tstruct object_id ret;\n> \tstruct object *obj = NULL;\n> @@ -98,15 +93,13 @@ static struct commit *create_commit(struct repository *repo,\n> \n> \tcommit_list_insert(parent, &parents);\n> \textra = read_commit_extra_headers(based_on, exclude_gpgsig);\n> -\tif (mode == REPLAY_MODE_REVERT) {\n> +\tif (reverse) {\n> \t\tgenerate_revert_message(&msg, based_on, repo);\n> \t\t/* For revert, use current user as author (NULL = use default) */\n> -\t} else if (mode == REPLAY_MODE_PICK) {\n> +\t} else {\n> \t\tfind_commit_subject(message, &orig_message);\n> \t\tstrbuf_addstr(&msg, orig_message);\n> \t\tauthor = get_author(message);\n> -\t} else {\n> -\t\tBUG(\"unexpected replay mode %d\", mode);\n> \t}\n> \treset_ident_date();\n> \tif (commit_tree_extended(msg.buf, msg.len, &tree->object.oid, parents,\n> @@ -269,7 +262,7 @@ static struct commit *pick_regular_commit(struct repository *repo,\n> \t\t\t\t\t struct commit *onto,\n> \t\t\t\t\t struct merge_options *merge_opt,\n> \t\t\t\t\t struct merge_result *result,\n> -\t\t\t\t\t enum replay_mode mode,\n> +\t\t\t\t\t bool reverse,\n> \t\t\t\t\t enum replay_empty_commit_action empty)\n> {\n> \tstruct commit *base, *replayed_base;\n> @@ -287,7 +280,21 @@ static struct commit *pick_regular_commit(struct repository *repo,\n> \treplayed_base_tree = repo_get_commit_tree(repo, replayed_base);\n> \tpickme_tree = repo_get_commit_tree(repo, pickme);\n> \n> -\tif (mode == REPLAY_MODE_PICK) {\n> +\tif (reverse) {\n> +\t\t/* Revert: swap base and pickme to reverse the diff */\n> +\t\tconst char *pickme_name = short_commit_name(repo, pickme);\n> +\t\tmerge_opt->branch1 = short_commit_name(repo, replayed_base);\n> +\t\tmerge_opt->branch2 = xstrfmt(\"parent of %s\", pickme_name);\n> +\t\tmerge_opt->ancestor = pickme_name;\n> +\n> +\t\tmerge_incore_nonrecursive(merge_opt,\n> +\t\t\t\t\t pickme_tree,\n> +\t\t\t\t\t replayed_base_tree,\n> +\t\t\t\t\t base_tree,\n> +\t\t\t\t\t result);\n> +\n> +\t\tfree((char *)merge_opt->branch2);\n> +\t} else {\n> \t\t/* Cherry-pick: normal order */\n> \t\tmerge_opt->branch1 = short_commit_name(repo, replayed_base);\n> \t\tmerge_opt->branch2 = short_commit_name(repo, pickme);\n> @@ -303,22 +310,6 @@ static struct commit *pick_regular_commit(struct repository *repo,\n> \t\t\t\t\t result);\n> \n> \t\tfree((char *)merge_opt->ancestor);\n> -\t} else if (mode == REPLAY_MODE_REVERT) {\n> -\t\t/* Revert: swap base and pickme to reverse the diff */\n> -\t\tconst char *pickme_name = short_commit_name(repo, pickme);\n> -\t\tmerge_opt->branch1 = short_commit_name(repo, replayed_base);\n> -\t\tmerge_opt->branch2 = xstrfmt(\"parent of %s\", pickme_name);\n> -\t\tmerge_opt->ancestor = pickme_name;\n> -\n> -\t\tmerge_incore_nonrecursive(merge_opt,\n> -\t\t\t\t\t pickme_tree,\n> -\t\t\t\t\t replayed_base_tree,\n> -\t\t\t\t\t base_tree,\n> -\t\t\t\t\t result);\n> -\n> -\t\tfree((char *)merge_opt->branch2);\n> -\t} else {\n> -\t\tBUG(\"unexpected replay mode %d\", mode);\n> \t}\n> \tmerge_opt->ancestor = NULL;\n> \tmerge_opt->branch2 = NULL;\n> @@ -341,7 +332,7 @@ static struct commit *pick_regular_commit(struct repository *repo,\n> \t\t}\n> \t}\n> \n> -\treturn create_commit(repo, result->tree, pickme, replayed_base, mode);\n> +\treturn create_commit(repo, result->tree, pickme, replayed_base, reverse);\n> }\n> \n> void replay_result_release(struct replay_result *result)\n> @@ -381,13 +372,13 @@ int replay_revisions(struct rev_info *revs,\n> \tchar *revert;\n> \tconst char *ref;\n> \tstruct object_id old_oid;\n> -\tenum replay_mode mode = REPLAY_MODE_PICK;\n> +\tbool reverse;\n> \tint ret;\n> \n> \tadvance = xstrdup_or_null(opts->advance);\n> \trevert = xstrdup_or_null(opts->revert);\n> -\tif (revert)\n> -\t\tmode = REPLAY_MODE_REVERT;\n> +\treverse = !!revert;\n> +\n> \tset_up_replay_mode(revs->repo, &revs->cmdline, opts->onto,\n> \t\t\t &detached_head, &advance, &revert, &onto, &update_refs);\n> \n> @@ -430,8 +421,8 @@ int replay_revisions(struct rev_info *revs,\n> \t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n> \n> \t\tlast_commit = pick_regular_commit(revs->repo, commit, replayed_commits,\n> -\t\t\t\t\t\t mode == REPLAY_MODE_REVERT ? last_commit : onto,\n> -\t\t\t\t\t\t &merge_opt, &result, mode, opts->empty);\n> +\t\t\t\t\t\t reverse ? last_commit : onto,\n> +\t\t\t\t\t\t &merge_opt, &result, reverse, opts->empty);\n> \t\tif (!last_commit)\n> \t\t\tbreak;\n\nThe patch itself looks trivially correct.\n\n-Justin\n"},{"id":"545352","messageId":"87h5n8yxur.fsf@emacs.iotcl.com","threadId":"65771","inReplyTo":"airORumUxyTsN7Bz@denethor","subject":"Re: [PATCH v2 1/3] replay: refactor enum replay_mode into a bool","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-12T08:19:40Z","receivedAt":"2026-06-12T08:20:10Z","isPatch":true,"body":"Justin Tobler writes:\n\n> Naive question: Do we expect there to only be two replay modes in the\n> forseeable future? I suppose if other modes were added in the future\n> this change would essentially be reverted.\n\nThe enum was mainly used to determine \"direction\": PICK to apply commits\nforward, and REVERT to apply them in opposite order. But it's a bit\ntwofold, because REVERT also applies the inverse change. At the moment\n--onto and --advance use PICK and --revert uses REVERT. There could be\nadded more options in the future, but I don't expect any of them to add\na new mode. And if there is ever a new mode needed, I think it's better\nto re-add the enum then, or maybe a second bool makes sense then, who\nknows...\n\n-- \nCheers,\nToon\n"},{"id":"545484","messageId":"CABPp-BGRi2obnqRGEY9pSMyvRbNGs8AdVUpZmr0C6vZSgHb=cg@mail.gmail.com","threadId":"65771","inReplyTo":"20260610-toon-git-replay-drop-merges-v2-3-5714a71c6d83@iotcl.com","subject":"Re: [PATCH v2 3/3] replay: offer an option to linearize the commit topology","fromName":"Elijah Newren","fromEmail":"newren@gmail.com","sentAt":"2026-06-14T06:56:04Z","receivedAt":"2026-06-14T06:56:16Z","isPatch":true,"body":"Hi,\n\nOn Wed, Jun 10, 2026 at 7:51 AM Toon Claes wrote:\n>\n> From: Johannes Schindelin \n>\n> One of the stated goals of git-replay(1) is to allow implementing the\n> git-rebase(1) functionality on the server side.\n>\n> The default mode of git-rebase(1) is to act as if `--no-rebase-merges`\n> was given. This mode drops merge commits instead of replaying them, and\n> linearizes the commit history into a sequence of the\n> regular (single-parent) commits.\n>\n> Add option `--linearize` to git-replay(1) to do the same.\n\nI think this version is nicer overall than the one from my\nreplay-upstream branch; sorry for repeatedly getting distracted from\nthat, but this does look nice.\n\nA few small comments:\n\n> Co-authored-by: Toon Claes \n> Signed-off-by: Johannes Schindelin \n> Signed-off-by: Toon Claes \n> ---\n> Documentation/git-replay.adoc | 5 +++++\n> builtin/replay.c | 4 ++++\n> replay.c | 30 +++++++++++++++++++++++-------\n> replay.h | 5 +++++\n> t/t3650-replay-basics.sh | 26 ++++++++++++++++++++++++++\n> 5 files changed, 63 insertions(+), 7 deletions(-)\n>\n> diff --git a/Documentation/git-replay.adoc b/Documentation/git-replay.adoc\n> index a32f72aead..41c96c7061 100644\n> --- a/Documentation/git-replay.adoc\n> +++ b/Documentation/git-replay.adoc\n> @@ -88,6 +88,11 @@ incompatible with `--contained` (which is a modifier for `--onto` only).\n> +\n> The default mode can be configured via the `replay.refAction` configuration variable.\n>\n> +--linearize::\n> + In this mode, `git replay` imitates `git rebase --no-rebase-merges`,\n> + i.e. it cherry-picks only non-merge commits, each one on top of the\n> + previous one.\n\nThe SYNOPSIS block at the top of the file is missing this new flag.\n\nThe replay_usage[] variable in cmd_replay is also missing this new flag.\n\n> ::\n> Range of commits to replay; see \"Specifying Ranges\" in\n> linkgit:git-rev-parse[1]. In `--advance=` or\n> diff --git a/builtin/replay.c b/builtin/replay.c\n> index 39e3a86f6c..fedfe46dc6 100644\n> --- a/builtin/replay.c\n> +++ b/builtin/replay.c\n> @@ -111,6 +111,8 @@ int cmd_replay(int argc,\n> N_(\"mode\"),\n> N_(\"control ref update behavior (update|print)\"),\n> PARSE_OPT_NONEG),\n> + OPT_BOOL(0, \"linearize\", &opts.linearize,\n> + N_(\"ignore merge commits instead of replaying them\")),\n\n\"ignore\" feels a bit ambiguous to me. Can we use \"drop\" instead,\nmatching your commit message?\n\n> OPT_END()\n> };\n>\n> @@ -132,6 +134,8 @@ int cmd_replay(int argc,\n> opts.contained, \"--contained\");\n> die_for_incompatible_opt2(!!opts.ref, \"--ref\",\n> !!opts.contained, \"--contained\");\n> + die_for_incompatible_opt2(!!opts.revert, \"--revert\",\n> + opts.linearize, \"--linearize\");\n\nSensible; should the docs mention this incompatibility? (I'm not sure\nmyself; just throwing it out as food for thought.)\n\n>\n> /* Parse ref action mode from command line or config */\n> ref_mode = get_ref_action_mode(repo, ref_action);\n> diff --git a/replay.c b/replay.c\n> index 7921d7dba3..81033fb889 100644\n> --- a/replay.c\n> +++ b/replay.c\n> @@ -277,12 +277,16 @@ static struct commit *pick_regular_commit(struct repository *repo,\n> struct commit *onto,\n> struct merge_options *merge_opt,\n> struct merge_result *result,\n> + struct commit *replayed_base,\n> bool reverse,\n> enum replay_empty_commit_action empty)\n> {\n> - struct commit *base, *replayed_base;\n> + struct commit *base;\n> struct tree *pickme_tree, *base_tree, *replayed_base_tree;\n>\n> + if (replayed_base && reverse)\n> + BUG(\"Linearizing commits is not supported when replaying in reverse\");\n> +\n\nThis is dead code given the die_for_incompatible_opt2 check above,\nright? Just extra defense in depth?\n\n> if (pickme->parents) {\n> base = pickme->parents->item;\n> base_tree = repo_get_commit_tree(repo, base);\n> @@ -291,7 +295,8 @@ static struct commit *pick_regular_commit(struct repository *repo,\n> base_tree = lookup_tree(repo, repo->hash_algo->empty_tree);\n> }\n>\n> - replayed_base = get_mapped_commit(replayed_commits, base, onto);\n> + if (!replayed_base)\n> + replayed_base = get_mapped_commit(replayed_commits, base, onto);\n> replayed_base_tree = repo_get_commit_tree(repo, replayed_base);\n> pickme_tree = repo_get_commit_tree(repo, pickme);\n>\n> @@ -430,12 +435,23 @@ int replay_revisions(struct rev_info *revs,\n> while ((commit = get_revision(revs))) {\n> const struct name_decoration *decoration;\n>\n> - if (commit->parents && commit->parents->next)\n> - die(_(\"replaying merge commits is not supported yet!\"));\n> + if (commit->parents && commit->parents->next) {\n> + if (!opts->linearize)\n> + die(_(\"replaying merge commits is not supported yet!\"));\n> + /*\n> + * When linearizing, a merge commit itself is not picked,\n> + * but refs that point to it might need updating.\n> + */\n\nIs it worth pointing out that last_commit is intentionally not updated\nby this code path? That is implied by your comment, but it takes a\nbit of reasoning to get there, and I think it might help future\nreaders to just explicitly state it.\n\n> + } else {\n> + struct commit *to_pick = reverse ? last_commit : onto;\n> + last_commit =\n> + pick_regular_commit(revs->repo, commit,\n> + replayed_commits, to_pick,\n> + &merge_opt, &result,\n> + opts->linearize ? last_commit : NULL,\n> + reverse, opts->empty);\n> + }\n>\n> - last_commit = pick_regular_commit(revs->repo, commit, replayed_commits,\n> - reverse ? last_commit : onto,\n> - &merge_opt, &result, reverse, opts->empty);\n> if (!last_commit)\n> break;\n>\n> diff --git a/replay.h b/replay.h\n> index 1851a07705..07e6fdcca3 100644\n> --- a/replay.h\n> +++ b/replay.h\n> @@ -62,6 +62,11 @@ struct replay_revisions_options {\n> * Defaults to REPLAY_EMPTY_COMMIT_DROP.\n> */\n> enum replay_empty_commit_action empty;\n> +\n> + /*\n> + * Whether to linearize the commits (i.e. drop merge commits).\n> + */\n> + int linearize;\n> };\n>\n> /* This struct is used as an out-parameter by `replay_revisions()`. */\n> diff --git a/t/t3650-replay-basics.sh b/t/t3650-replay-basics.sh\n> index 3353bc4a4d..64e0731188 100755\n> --- a/t/t3650-replay-basics.sh\n> +++ b/t/t3650-replay-basics.sh\n> @@ -565,4 +565,30 @@ test_expect_success '--onto with --ref rejects multiple revision ranges' '\n> test_grep \"cannot be used with multiple revision ranges\" err\n> '\n>\n> +test_expect_success 'replay merge commit fails' '\n> + echo \"fatal: replaying merge commits is not supported yet!\" >expect &&\n> + test_must_fail git replay --ref-action=print --onto main I..P 2>actual &&\n> + test_cmp expect actual\n> +'\n> +\n> +test_expect_success 'replay to rebase merge commit with --linearize' '\n> + git replay --ref-action=print --linearize --onto main I..topic-with-merge >result &&\n> +\n> + test_line_count = 1 result &&\n> +\n> + git log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n> + test_write_lines O N J M L B A >expect &&\n> + test_cmp expect actual\n> +'\n> +\n> +test_expect_success 'replay to rebase merge commit with --linearize down to root commit' '\n> + git replay --ref-action=print --linearize --onto main A..topic-with-merge >result &&\n\nYou'd need to drop \"A..\" to have it go down to the root commit, as\nJunio mentioned elsewhere.\n\n> +\n> + test_line_count = 1 result &&\n> +\n> + git log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n> + test_write_lines O N J I M L B A >expect &&\n> + test_cmp expect actual\n> +'\n> +\n> test_done\n\nShould there also be a testcase combining --linearize and --advance?\n\nShould there be a test with the incompatibility of --revert &\n--linearize? I think we have a few other tests for incompatible\noptions.\n\nOne additional testing idea, borrowed from an older variant of\nthis patch I had sitting in a local branch (dscho's original\nlinearize patch, adapted): in addition to checking specific commit\nsubjects, it's worth verifying that the linearized chain produces\nthe *same patches* as the original. Something along the lines of:\n\n test_expect_success '--linearize preserves patches' '\n test_when_finished \"git update-ref -d refs/heads/merge_I_L\" &&\n test_tick &&\n git checkout -b merge_I_L I &&\n git merge --no-edit L &&\n\n git replay --linearize --onto A B..merge_I_L &&\n\n # range-diff ignores merges, so the original\n # {I, L, merge} reduces to {I, L} on the LHS,\n # and the replayed chain on the RHS should match.\n git range-diff B..merge_I_L@{1} B..merge_I_L >out &&\n ! test_grep -v \"=\" out &&\n\n git log --oneline A..merge_I_L >out &&\n test_line_count = 2 out\n '\n\nThe range-diff check is nice because it asserts patch equivalence\nrather than tying the test to a particular replay ordering, which\nmakes the test less brittle if the rev-walk order ever changes.\nFeel free to take, adapt, or ignore.\n\nAnyway, thanks for working on this; looking good.\n\nElijah\n"},{"id":"545628","messageId":"874ij3ynab.fsf@emacs.iotcl.com","threadId":"65771","inReplyTo":"CABPp-BGRi2obnqRGEY9pSMyvRbNGs8AdVUpZmr0C6vZSgHb=cg@mail.gmail.com","subject":"Re: [PATCH v2 3/3] replay: offer an option to linearize the commit topology","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-16T07:09:16Z","receivedAt":"2026-06-16T07:09:24Z","isPatch":true,"body":"Elijah Newren writes:\n\n> Hi,\n>\n> On Wed, Jun 10, 2026 at 7:51 AM Toon Claes wrote:\n>>\n>> From: Johannes Schindelin \n>>\n>> One of the stated goals of git-replay(1) is to allow implementing the\n>> git-rebase(1) functionality on the server side.\n>>\n>> The default mode of git-rebase(1) is to act as if `--no-rebase-merges`\n>> was given. This mode drops merge commits instead of replaying them, and\n>> linearizes the commit history into a sequence of the\n>> regular (single-parent) commits.\n>>\n>> Add option `--linearize` to git-replay(1) to do the same.\n>\n> I think this version is nicer overall than the one from my\n> replay-upstream branch; sorry for repeatedly getting distracted from\n> that, but this does look nice.\n\nDscho gets most of the credit here. And don't worry about being\ndistracted, we know how things go around here. I appreciate the review!\n\n>\n> A few small comments:\n>\n>> Co-authored-by: Toon Claes \n>> Signed-off-by: Johannes Schindelin \n>> Signed-off-by: Toon Claes \n>> ---\n>> Documentation/git-replay.adoc | 5 +++++\n>> builtin/replay.c | 4 ++++\n>> replay.c | 30 +++++++++++++++++++++++-------\n>> replay.h | 5 +++++\n>> t/t3650-replay-basics.sh | 26 ++++++++++++++++++++++++++\n>> 5 files changed, 63 insertions(+), 7 deletions(-)\n>>\n>> diff --git a/Documentation/git-replay.adoc b/Documentation/git-replay.adoc\n>> index a32f72aead..41c96c7061 100644\n>> --- a/Documentation/git-replay.adoc\n>> +++ b/Documentation/git-replay.adoc\n>> @@ -88,6 +88,11 @@ incompatible with `--contained` (which is a modifier for `--onto` only).\n>> +\n>> The default mode can be configured via the `replay.refAction` configuration variable.\n>>\n>> +--linearize::\n>> + In this mode, `git replay` imitates `git rebase --no-rebase-merges`,\n>> + i.e. it cherry-picks only non-merge commits, each one on top of the\n>> + previous one.\n>\n> The SYNOPSIS block at the top of the file is missing this new flag.\n>\n> The replay_usage[] variable in cmd_replay is also missing this new flag.\n>\n>> ::\n>> Range of commits to replay; see \"Specifying Ranges\" in\n>> linkgit:git-rev-parse[1]. In `--advance=` or\n>> diff --git a/builtin/replay.c b/builtin/replay.c\n>> index 39e3a86f6c..fedfe46dc6 100644\n>> --- a/builtin/replay.c\n>> +++ b/builtin/replay.c\n>> @@ -111,6 +111,8 @@ int cmd_replay(int argc,\n>> N_(\"mode\"),\n>> N_(\"control ref update behavior (update|print)\"),\n>> PARSE_OPT_NONEG),\n>> + OPT_BOOL(0, \"linearize\", &opts.linearize,\n>> + N_(\"ignore merge commits instead of replaying them\")),\n>\n> \"ignore\" feels a bit ambiguous to me. Can we use \"drop\" instead,\n> matching your commit message?\n\nAgreed, I don't like it too. \"drop\" sounds better.\n\n>> OPT_END()\n>> };\n>>\n>> @@ -132,6 +134,8 @@ int cmd_replay(int argc,\n>> opts.contained, \"--contained\");\n>> die_for_incompatible_opt2(!!opts.ref, \"--ref\",\n>> !!opts.contained, \"--contained\");\n>> + die_for_incompatible_opt2(!!opts.revert, \"--revert\",\n>> + opts.linearize, \"--linearize\");\n>\n> Sensible; should the docs mention this incompatibility? (I'm not sure\n> myself; just throwing it out as food for thought.)\n\nLet's add it.\n\n>>\n>> /* Parse ref action mode from command line or config */\n>> ref_mode = get_ref_action_mode(repo, ref_action);\n>> diff --git a/replay.c b/replay.c\n>> index 7921d7dba3..81033fb889 100644\n>> --- a/replay.c\n>> +++ b/replay.c\n>> @@ -277,12 +277,16 @@ static struct commit *pick_regular_commit(struct repository *repo,\n>> struct commit *onto,\n>> struct merge_options *merge_opt,\n>> struct merge_result *result,\n>> + struct commit *replayed_base,\n>> bool reverse,\n>> enum replay_empty_commit_action empty)\n>> {\n>> - struct commit *base, *replayed_base;\n>> + struct commit *base;\n>> struct tree *pickme_tree, *base_tree, *replayed_base_tree;\n>>\n>> + if (replayed_base && reverse)\n>> + BUG(\"Linearizing commits is not supported when replaying in reverse\");\n>> +\n>\n> This is dead code given the die_for_incompatible_opt2 check above,\n> right? Just extra defense in depth?\n\nWe also have another defense-in-depth for --onto/--advance/--revert:\n\n BUG(\"expected one of onto_name, *advance_name, or *revert_name\");\n\nI don't mind having it for --linearize too.\n\n>> if (pickme->parents) {\n>> base = pickme->parents->item;\n>> base_tree = repo_get_commit_tree(repo, base);\n>> @@ -291,7 +295,8 @@ static struct commit *pick_regular_commit(struct repository *repo,\n>> base_tree = lookup_tree(repo, repo->hash_algo->empty_tree);\n>> }\n>>\n>> - replayed_base = get_mapped_commit(replayed_commits, base, onto);\n>> + if (!replayed_base)\n>> + replayed_base = get_mapped_commit(replayed_commits, base, onto);\n>> replayed_base_tree = repo_get_commit_tree(repo, replayed_base);\n>> pickme_tree = repo_get_commit_tree(repo, pickme);\n>>\n>> @@ -430,12 +435,23 @@ int replay_revisions(struct rev_info *revs,\n>> while ((commit = get_revision(revs))) {\n>> const struct name_decoration *decoration;\n>>\n>> - if (commit->parents && commit->parents->next)\n>> - die(_(\"replaying merge commits is not supported yet!\"));\n>> + if (commit->parents && commit->parents->next) {\n>> + if (!opts->linearize)\n>> + die(_(\"replaying merge commits is not supported yet!\"));\n>> + /*\n>> + * When linearizing, a merge commit itself is not picked,\n>> + * but refs that point to it might need updating.\n>> + */\n>\n> Is it worth pointing out that last_commit is intentionally not updated\n> by this code path? That is implied by your comment, but it takes a\n> bit of reasoning to get there, and I think it might help future\n> readers to just explicitly state it.\n\nAh yes, I didn't realize that, but you make a good point. I'll rephrase\nthe comment a bit.\n\n>> + } else {\n>> + struct commit *to_pick = reverse ? last_commit : onto;\n>> + last_commit =\n>> + pick_regular_commit(revs->repo, commit,\n>> + replayed_commits, to_pick,\n>> + &merge_opt, &result,\n>> + opts->linearize ? last_commit : NULL,\n>> + reverse, opts->empty);\n>> + }\n>>\n>> - last_commit = pick_regular_commit(revs->repo, commit, replayed_commits,\n>> - reverse ? last_commit : onto,\n>> - &merge_opt, &result, reverse, opts->empty);\n>> if (!last_commit)\n>> break;\n>>\n>> diff --git a/replay.h b/replay.h\n>> index 1851a07705..07e6fdcca3 100644\n>> --- a/replay.h\n>> +++ b/replay.h\n>> @@ -62,6 +62,11 @@ struct replay_revisions_options {\n>> * Defaults to REPLAY_EMPTY_COMMIT_DROP.\n>> */\n>> enum replay_empty_commit_action empty;\n>> +\n>> + /*\n>> + * Whether to linearize the commits (i.e. drop merge commits).\n>> + */\n>> + int linearize;\n>> };\n>>\n>> /* This struct is used as an out-parameter by `replay_revisions()`. */\n>> diff --git a/t/t3650-replay-basics.sh b/t/t3650-replay-basics.sh\n>> index 3353bc4a4d..64e0731188 100755\n>> --- a/t/t3650-replay-basics.sh\n>> +++ b/t/t3650-replay-basics.sh\n>> @@ -565,4 +565,30 @@ test_expect_success '--onto with --ref rejects multiple revision ranges' '\n>> test_grep \"cannot be used with multiple revision ranges\" err\n>> '\n>>\n>> +test_expect_success 'replay merge commit fails' '\n>> + echo \"fatal: replaying merge commits is not supported yet!\" >expect &&\n>> + test_must_fail git replay --ref-action=print --onto main I..P 2>actual &&\n>> + test_cmp expect actual\n>> +'\n>> +\n>> +test_expect_success 'replay to rebase merge commit with --linearize' '\n>> + git replay --ref-action=print --linearize --onto main I..topic-with-merge >result &&\n>> +\n>> + test_line_count = 1 result &&\n>> +\n>> + git log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n>> + test_write_lines O N J M L B A >expect &&\n>> + test_cmp expect actual\n>> +'\n>> +\n>> +test_expect_success 'replay to rebase merge commit with --linearize down to root commit' '\n>> + git replay --ref-action=print --linearize --onto main A..topic-with-merge >result &&\n>\n> You'd need to drop \"A..\" to have it go down to the root commit, as\n> Junio mentioned elsewhere.\n\nYes, thanks for double confirmation.\n\n>> +\n>> + test_line_count = 1 result &&\n>> +\n>> + git log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n>> + test_write_lines O N J I M L B A >expect &&\n>> + test_cmp expect actual\n>> +'\n>> +\n>> test_done\n>\n> Should there also be a testcase combining --linearize and --advance?\n\nSure.\n\n> Should there be a test with the incompatibility of --revert &\n> --linearize? I think we have a few other tests for incompatible\n> options.\n\nI was already about to add that.\n\n> One additional testing idea, borrowed from an older variant of\n> this patch I had sitting in a local branch (dscho's original\n> linearize patch, adapted): in addition to checking specific commit\n> subjects, it's worth verifying that the linearized chain produces\n> the *same patches* as the original. Something along the lines of:\n>\n> test_expect_success '--linearize preserves patches' '\n> test_when_finished \"git update-ref -d refs/heads/merge_I_L\" &&\n> test_tick &&\n> git checkout -b merge_I_L I &&\n> git merge --no-edit L &&\n>\n> git replay --linearize --onto A B..merge_I_L &&\n>\n> # range-diff ignores merges, so the original\n> # {I, L, merge} reduces to {I, L} on the LHS,\n> # and the replayed chain on the RHS should match.\n> git range-diff B..merge_I_L@{1} B..merge_I_L >out &&\n> ! test_grep -v \"=\" out &&\n>\n> git log --oneline A..merge_I_L >out &&\n> test_line_count = 2 out\n> '\n>\n> The range-diff check is nice because it asserts patch equivalence\n> rather than tying the test to a particular replay ordering, which\n> makes the test less brittle if the rev-walk order ever changes.\n> Feel free to take, adapt, or ignore.\n\nInteresting idea and I like it. Lemme add it.\n\n> Anyway, thanks for working on this; looking good.\n\nThanks!\n\n-- \nCheers,\nToon\n"},{"id":"545632","messageId":"871pe6zxpx.fsf@emacs.iotcl.com","threadId":"65771","inReplyTo":"xmqqjys6wcpo.fsf@gitster.g","subject":"Re: [PATCH v2 3/3] replay: offer an option to linearize the commit topology","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-16T08:38:34Z","receivedAt":"2026-06-16T08:38:41Z","isPatch":true,"body":"Junio C Hamano writes:\n\n> In the review response during the previous iteration, I commented\n> that (1) the original excluded only merges, but (2) your version\n> excluded both merges and the root commits the same way. Your\n> response was:\n>\n> The way it was written in v1 was maybe a bit too smart and hard to\n> follow. I agree with your suggestion and will adopt this (with some\n> tweaks) in the next version.\n>\n> which I took as saying \"it may be confusing, but it correctly\n> expresses what we want to do\", meaning \"yes, roots and merges should\n> be handled the same way\". But the above no longer treats roots the\n> same way as merges. I think that is intended, but just wanted to\n> double check.\n\nGreat callout. I was running the \"replay down to root\" test with v1 vs\nv2, but as you pointed out, the test I wrote doesn't actually replay\ndown to root. Now I've fixed the test and reran the test against both\nversions and verified what you're saying.\n\nSo to answer your question, yes this change is intentional and shout out\nto you for requesting to add this test (properly) so we actually catch\nthis. \n\n>> +test_expect_success 'replay to rebase merge commit with --linearize down to root commit' '\n>> +\tgit replay --ref-action=print --linearize --onto main A..topic-with-merge >result &&\n>\n> As with other test pieces, this \"git replay\" command line is overly\n> long and hides the important bit which is that the range being\n> replayed is *not* actually down to the root, which is A (it excludes\n> A). Intended?\n\nNo, not intended. And while at it, I'll split up the command on two\nlines. \n\n-- \nCheers,\nToon\n"},{"id":"545638","messageId":"20260616-toon-git-replay-drop-merges-v3-0-153e9eb99ce1@iotcl.com","threadId":"65771","inReplyTo":"20260610-toon-git-replay-drop-merges-v2-0-5714a71c6d83@iotcl.com","subject":"[PATCH v3 0/3] Teach git-replay(1) to linearize merge commits","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-16T09:26:34Z","receivedAt":"2026-06-16T09:27:03Z","isPatch":true,"body":"As an alternative to dscho's patch series to replay merges[1], add\noption to git-replay(1) to linearize merges. This mimics what\ngit-rebase(1) does too with --no-rebase-merges (the default).\n\nThe first two patches do some refactoring. The third patch implements\nthe actual change. This patch was kindly provided by Dscho, which I've\ntweaked to be upstreamed.\n\nThe --linearize option is only added to git-replay(1) and not to\ngit-history(1) because in my opinion it doesn't make much sense to do\nso, but I'm happy to hear if anyone disagrees.\n\nThis series might conflict with Kristoffer's series to make\ndocumentation changes[2], but should be trivial to resolve. And I don't\nthink there's a conflict with Patrick's series on adding \"drop\" to\ngit-history(1)[3].\n\ndscho's series to replay merges[1] needs a bit of rework to fit on top\nof this, but I'm happy to help figuring that out. We've been discussing\nto either name the option --flatten or --linearize, but I've decided on\n\"linearize\" because the documentation of git-rebase(1) also mentions\n\"linearize\".\n\n[1]: \n[2]: \n[3]: <20260603-b4-pks-history-drop-v2-0-742cb5b5176d@pks.im>\n\nSigned-off-by: Toon Claes \n---\nChanges in v3:\n- Add --linearize to Documentation SYNOPSIS, and mention it's\n incompatible with --revert.\n- Small language change in help message for --linearize.\n- Rephrase comment to include last_commit isn't modified when\n linearizing merges.\n- Remove test that was added in earlier versions, but actually is\n a duplicate of 'replaying merge commits is not supported yet'.\n- Add test to verify --revert and --linearize are incompatible.\n- Properly test that replaying down to root with --linearize works.\n- Add test for --linearize with --advance.\n- Add test that uses git-range-diff(1) to verify the patches created by\n --linearize are correct.\n- Link to v2: https://patch.msgid.link/20260610-toon-git-replay-drop-merges-v2-0-5714a71c6d83@iotcl.com\n\nChanges in v2:\n- Restructured the conditions to detect merge commits and added a line\n of comment why the loop continues.\n- Rewrote tests to use the history from the setup step and added a few\n test cases.\n- Re-added Johannes's Signed-off-by trailer. Johannes gave me the\n patches with this trailer, and if I understand correctly, I can keep\n it. Please let me know if that wrong.\n- Link to v1: https://patch.msgid.link/20260608-toon-git-replay-drop-merges-v1-0-e3ee71fce7b4@iotcl.com\n\n---\nJohannes Schindelin (1):\n replay: offer an option to linearize the commit topology\n\nToon Claes (2):\n replay: refactor enum replay_mode into a bool\n replay: add helper to put entry into mapped_commits\n\n Documentation/git-replay.adoc | 8 ++-\n builtin/replay.c | 6 ++-\n replay.c | 116 ++++++++++++++++++++++++------------------\n replay.h | 5 ++\n t/t3650-replay-basics.sh | 68 ++++++++++++++++++++++++-\n 5 files changed, 151 insertions(+), 52 deletions(-)\n\nRange-diff versus v2:\n\n1: 2075988ef1 = 1: 542b1c9267 replay: refactor enum replay_mode into a bool\n2: 93ff03be65 = 2: 62f6df8375 replay: add helper to put entry into mapped_commits\n3: ef56010c96 ! 3: 768646ee24 replay: offer an option to linearize the commit topology\n @@ Commit message\n Signed-off-by: Toon Claes \n \n ## Documentation/git-replay.adoc ##\n +@@ Documentation/git-replay.adoc: SYNOPSIS\n + --------\n + [verse]\n + (EXPERIMENTAL!) 'git replay' ([--contained] --onto= | --advance= | --revert=)\n +-\t\t\t [--ref=] [--ref-action=] \n ++\t\t\t [--ref=] [--ref-action=] [--linearize] \n + \n + DESCRIPTION\n + -----------\n @@ Documentation/git-replay.adoc: incompatible with `--contained` (which is a modifier for `--onto` only).\n +\n The default mode can be configured via the `replay.refAction` configuration variable.\n @@ Documentation/git-replay.adoc: incompatible with `--contained` (which is a modif\n +\tIn this mode, `git replay` imitates `git rebase --no-rebase-merges`,\n +\ti.e. it cherry-picks only non-merge commits, each one on top of the\n +\tprevious one.\n ++\tThis option is incompatible with `--revert`.\n +\n ::\n \tRange of commits to replay; see \"Specifying Ranges\" in\n \tlinkgit:git-rev-parse[1]. In `--advance=` or\n \n ## builtin/replay.c ##\n +@@ builtin/replay.c: int cmd_replay(int argc,\n + \tconst char *const replay_usage[] = {\n + \t\tN_(\"(EXPERIMENTAL!) git replay \"\n + \t\t \"([--contained] --onto= | --advance= | --revert=)\\n\"\n +-\t\t \"[--ref=] [--ref-action=] \"),\n ++\t\t \"[--ref=] [--ref-action=] [--linearize] \"),\n + \t\tNULL\n + \t};\n + \tstruct option replay_options[] = {\n @@ builtin/replay.c: int cmd_replay(int argc,\n \t\t\t N_(\"mode\"),\n \t\t\t N_(\"control ref update behavior (update|print)\"),\n \t\t\t PARSE_OPT_NONEG),\n +\t\tOPT_BOOL(0, \"linearize\", &opts.linearize,\n -+\t\t\t N_(\"ignore merge commits instead of replaying them\")),\n ++\t\t\t N_(\"drop merge commits, replaying only non-merge commits\")),\n \t\tOPT_END()\n \t};\n \n @@ replay.c: int replay_revisions(struct rev_info *revs,\n +\t\t\tif (!opts->linearize)\n +\t\t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n +\t\t\t/*\n -+\t\t\t * When linearizing, a merge commit itself is not picked,\n -+\t\t\t * but refs that point to it might need updating.\n ++\t\t\t * Drop the merge commit: do not pick it and leave\n ++\t\t\t * last_commit unchanged, so its children (and any ref\n ++\t\t\t * pointing at it) are reparented onto the previous\n ++\t\t\t * non-merge commit, which the ref-update loop below uses.\n +\t\t\t */\n +\t\t} else {\n +\t\t\tstruct commit *to_pick = reverse ? last_commit : onto;\n @@ replay.h: struct replay_revisions_options {\n /* This struct is used as an out-parameter by `replay_revisions()`. */\n \n ## t/t3650-replay-basics.sh ##\n -@@ t/t3650-replay-basics.sh: test_expect_success '--onto with --ref rejects multiple revision ranges' '\n - \ttest_grep \"cannot be used with multiple revision ranges\" err\n +@@ t/t3650-replay-basics.sh: test_expect_success 'setup' '\n + \ttest_merge P O --no-ff &&\n + \tgit switch main &&\n + \n ++\tgit switch --orphan unrelated &&\n ++\ttest_commit unrelated-root &&\n ++\n + \tgit switch -c conflict B &&\n +-\ttest_commit C.conflict C.t conflict\n ++\ttest_commit C.conflict C.t conflict &&\n ++\tgit branch -D unrelated\n '\n \n -+test_expect_success 'replay merge commit fails' '\n -+\techo \"fatal: replaying merge commits is not supported yet!\" >expect &&\n -+\ttest_must_fail git replay --ref-action=print --onto main I..P 2>actual &&\n -+\ttest_cmp expect actual\n + test_expect_success 'setup bare' '\n +@@ t/t3650-replay-basics.sh: test_expect_success '--advance and --contained cannot be used together' '\n + \ttest_grep \"cannot be used together\" actual\n + '\n + \n ++test_expect_success '--revert and --linearize cannot be used together' '\n ++\ttest_must_fail git replay --revert=main --linearize \\\n ++\t\ttopic1..topic2 2>actual &&\n ++\ttest_grep \"cannot be used together\" actual\n +'\n +\n + test_expect_success 'cannot advance target ... ordering would be ill-defined' '\n + \techo \"fatal: ${SQ}--advance${SQ} cannot be used with multiple revision ranges because the ordering would be ill-defined\" >expect &&\n + \ttest_must_fail git replay --advance=main main topic1 topic2 2>actual &&\n +@@ t/t3650-replay-basics.sh: test_expect_success '--onto with --ref rejects multiple revision ranges' '\n + \ttest_grep \"cannot be used with multiple revision ranges\" err\n + '\n + \n +test_expect_success 'replay to rebase merge commit with --linearize' '\n -+\tgit replay --ref-action=print --linearize --onto main I..topic-with-merge >result &&\n ++\tgit replay --ref-action=print --linearize \\\n ++\t\t--onto main I..topic-with-merge >result &&\n +\n +\ttest_line_count = 1 result &&\n +\n @@ t/t3650-replay-basics.sh: test_expect_success '--onto with --ref rejects multipl\n +\ttest_cmp expect actual\n +'\n +\n -+test_expect_success 'replay to rebase merge commit with --linearize down to root commit' '\n -+\tgit replay --ref-action=print --linearize --onto main A..topic-with-merge >result &&\n ++test_expect_success 'replay to rebase merge commit with --linearize down to the root commit' '\n ++\tgit replay --ref-action=print --linearize \\\n ++\t\t--onto unrelated-root topic-with-merge >result &&\n +\n +\ttest_line_count = 1 result &&\n +\n +\tgit log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n -+\ttest_write_lines O N J I M L B A >expect &&\n ++\ttest_write_lines O N J I B A unrelated-root >expect &&\n +\ttest_cmp expect actual\n +'\n ++\n ++test_expect_success 'replay to cherry-pick merge commit with --linearize' '\n ++\tgit replay --ref-action=print --linearize \\\n ++\t\t--advance main I..topic-with-merge >result &&\n ++\n ++\ttest_line_count = 1 result &&\n ++\n ++\tgit log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n ++\ttest_write_lines O N J M L B A >expect &&\n ++\ttest_cmp expect actual &&\n ++\n ++\tprintf \"update refs/heads/main \" >expect &&\n ++\tprintf \"%s \" $(cut -f 3 -d \" \" result) >>expect &&\n ++\tgit rev-parse main >>expect &&\n ++\ttest_cmp expect result\n ++'\n ++\n ++test_expect_success 'replay --linearize produces the same patches' '\n ++\tgit replay --ref-action=print --linearize \\\n ++\t\t--onto main I..topic-with-merge >result &&\n ++\n ++\ttest_line_count = 1 result &&\n ++\ttip=$(cut -f 3 -d \" \" result) &&\n ++\n ++\t# range-diff does not care about the dropped merge,\n ++\t# so the original commits (I..topic-with-merge)\n ++\t# and the replayed chain (main..tip) must produce identical patches.\n ++\tgit range-diff I..topic-with-merge main..$tip >out &&\n ++\ttest_file_not_empty out &&\n ++\t! grep -v \"=\" out &&\n ++\n ++\tgit log --oneline main..$tip >out &&\n ++\ttest_line_count = 3 out\n ++'\n +\n test_done\n\n\n---\nbase-commit: 9ac3f193c05c2237e2b14ebaa1149e9fc8a1abe0\nchange-id: 20260604-toon-git-replay-drop-merges-807fa008d395\n\n"},{"id":"545639","messageId":"20260616-toon-git-replay-drop-merges-v3-1-153e9eb99ce1@iotcl.com","threadId":"65771","inReplyTo":"20260616-toon-git-replay-drop-merges-v3-0-153e9eb99ce1@iotcl.com","subject":"[PATCH v3 1/3] replay: refactor enum replay_mode into a bool","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-16T09:26:35Z","receivedAt":"2026-06-16T09:27:13Z","isPatch":true,"body":"In 2760ee4983 (replay: add --revert mode to reverse commit changes,\n2026-03-26) the enum `replay_mode` was introduced. This has two possible\nvalues:\n\n - The value `REPLAY_MODE_REVERT` is used when option `--revert` is\n passed to git-replay(1). When using this value the commits are\n processed in reverse order and the inverse of the changes are\n applied.\n\n - The value `REPLAY_MODE_PICK` is used when either option `--onto` or\n `--advance` is used. In both cases the commits are processed in\n normal order, and the changes are applied as-is.\n\nSince there are only two possible values of this enum, simplify the code\nby converting the enum into a bool. This avoids adding code paths that\ncheck for invalid values of the enum, and shortens code where the value\nis checked with a ternary operator.\n\nSigned-off-by: Toon Claes \n---\n replay.c | 59 +++++++++++++++++++++++++----------------------------------\n 1 file changed, 25 insertions(+), 34 deletions(-)\n\ndiff --git a/replay.c b/replay.c\nindex 4ef8abb607..1f8e5b083b 100644\n--- a/replay.c\n+++ b/replay.c\n@@ -18,11 +18,6 @@\n */\n #define the_repository DO_NOT_USE_THE_REPOSITORY\n \n-enum replay_mode {\n-\tREPLAY_MODE_PICK,\n-\tREPLAY_MODE_REVERT,\n-};\n-\n static const char *short_commit_name(struct repository *repo,\n \t\t\t\t struct commit *commit)\n {\n@@ -81,7 +76,7 @@ static struct commit *create_commit(struct repository *repo,\n \t\t\t\t struct tree *tree,\n \t\t\t\t struct commit *based_on,\n \t\t\t\t struct commit *parent,\n-\t\t\t\t enum replay_mode mode)\n+\t\t\t\t bool reverse)\n {\n \tstruct object_id ret;\n \tstruct object *obj = NULL;\n@@ -98,15 +93,13 @@ static struct commit *create_commit(struct repository *repo,\n \n \tcommit_list_insert(parent, &parents);\n \textra = read_commit_extra_headers(based_on, exclude_gpgsig);\n-\tif (mode == REPLAY_MODE_REVERT) {\n+\tif (reverse) {\n \t\tgenerate_revert_message(&msg, based_on, repo);\n \t\t/* For revert, use current user as author (NULL = use default) */\n-\t} else if (mode == REPLAY_MODE_PICK) {\n+\t} else {\n \t\tfind_commit_subject(message, &orig_message);\n \t\tstrbuf_addstr(&msg, orig_message);\n \t\tauthor = get_author(message);\n-\t} else {\n-\t\tBUG(\"unexpected replay mode %d\", mode);\n \t}\n \treset_ident_date();\n \tif (commit_tree_extended(msg.buf, msg.len, &tree->object.oid, parents,\n@@ -269,7 +262,7 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t struct commit *onto,\n \t\t\t\t\t struct merge_options *merge_opt,\n \t\t\t\t\t struct merge_result *result,\n-\t\t\t\t\t enum replay_mode mode,\n+\t\t\t\t\t bool reverse,\n \t\t\t\t\t enum replay_empty_commit_action empty)\n {\n \tstruct commit *base, *replayed_base;\n@@ -287,7 +280,21 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \treplayed_base_tree = repo_get_commit_tree(repo, replayed_base);\n \tpickme_tree = repo_get_commit_tree(repo, pickme);\n \n-\tif (mode == REPLAY_MODE_PICK) {\n+\tif (reverse) {\n+\t\t/* Revert: swap base and pickme to reverse the diff */\n+\t\tconst char *pickme_name = short_commit_name(repo, pickme);\n+\t\tmerge_opt->branch1 = short_commit_name(repo, replayed_base);\n+\t\tmerge_opt->branch2 = xstrfmt(\"parent of %s\", pickme_name);\n+\t\tmerge_opt->ancestor = pickme_name;\n+\n+\t\tmerge_incore_nonrecursive(merge_opt,\n+\t\t\t\t\t pickme_tree,\n+\t\t\t\t\t replayed_base_tree,\n+\t\t\t\t\t base_tree,\n+\t\t\t\t\t result);\n+\n+\t\tfree((char *)merge_opt->branch2);\n+\t} else {\n \t\t/* Cherry-pick: normal order */\n \t\tmerge_opt->branch1 = short_commit_name(repo, replayed_base);\n \t\tmerge_opt->branch2 = short_commit_name(repo, pickme);\n@@ -303,22 +310,6 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t result);\n \n \t\tfree((char *)merge_opt->ancestor);\n-\t} else if (mode == REPLAY_MODE_REVERT) {\n-\t\t/* Revert: swap base and pickme to reverse the diff */\n-\t\tconst char *pickme_name = short_commit_name(repo, pickme);\n-\t\tmerge_opt->branch1 = short_commit_name(repo, replayed_base);\n-\t\tmerge_opt->branch2 = xstrfmt(\"parent of %s\", pickme_name);\n-\t\tmerge_opt->ancestor = pickme_name;\n-\n-\t\tmerge_incore_nonrecursive(merge_opt,\n-\t\t\t\t\t pickme_tree,\n-\t\t\t\t\t replayed_base_tree,\n-\t\t\t\t\t base_tree,\n-\t\t\t\t\t result);\n-\n-\t\tfree((char *)merge_opt->branch2);\n-\t} else {\n-\t\tBUG(\"unexpected replay mode %d\", mode);\n \t}\n \tmerge_opt->ancestor = NULL;\n \tmerge_opt->branch2 = NULL;\n@@ -341,7 +332,7 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t}\n \t}\n \n-\treturn create_commit(repo, result->tree, pickme, replayed_base, mode);\n+\treturn create_commit(repo, result->tree, pickme, replayed_base, reverse);\n }\n \n void replay_result_release(struct replay_result *result)\n@@ -381,13 +372,13 @@ int replay_revisions(struct rev_info *revs,\n \tchar *revert;\n \tconst char *ref;\n \tstruct object_id old_oid;\n-\tenum replay_mode mode = REPLAY_MODE_PICK;\n+\tbool reverse;\n \tint ret;\n \n \tadvance = xstrdup_or_null(opts->advance);\n \trevert = xstrdup_or_null(opts->revert);\n-\tif (revert)\n-\t\tmode = REPLAY_MODE_REVERT;\n+\treverse = !!revert;\n+\n \tset_up_replay_mode(revs->repo, &revs->cmdline, opts->onto,\n \t\t\t &detached_head, &advance, &revert, &onto, &update_refs);\n \n@@ -430,8 +421,8 @@ int replay_revisions(struct rev_info *revs,\n \t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n \n \t\tlast_commit = pick_regular_commit(revs->repo, commit, replayed_commits,\n-\t\t\t\t\t\t mode == REPLAY_MODE_REVERT ? last_commit : onto,\n-\t\t\t\t\t\t &merge_opt, &result, mode, opts->empty);\n+\t\t\t\t\t\t reverse ? last_commit : onto,\n+\t\t\t\t\t\t &merge_opt, &result, reverse, opts->empty);\n \t\tif (!last_commit)\n \t\t\tbreak;\n \n\n-- \n2.53.0.1323.g189a785ab5\n\n"},{"id":"545640","messageId":"20260616-toon-git-replay-drop-merges-v3-2-153e9eb99ce1@iotcl.com","threadId":"65771","inReplyTo":"20260616-toon-git-replay-drop-merges-v3-0-153e9eb99ce1@iotcl.com","subject":"[PATCH v3 2/3] replay: add helper to put entry into mapped_commits","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-16T09:26:36Z","receivedAt":"2026-06-16T09:27:22Z","isPatch":true,"body":"The function replay_revisions() in replay.c is rather lengthy. Extract\nthe logic to put a commit entry into mapped_commits into a helper\nfunction put_mapped_commit().\n\nWhile at it, rename mapped_commit() to get_mapped_commit() to pair with\nthis new function.\n\nSigned-off-by: Toon Claes \n---\n replay.c | 31 ++++++++++++++++++++-----------\n 1 file changed, 20 insertions(+), 11 deletions(-)\n\ndiff --git a/replay.c b/replay.c\nindex 1f8e5b083b..7921d7dba3 100644\n--- a/replay.c\n+++ b/replay.c\n@@ -243,9 +243,9 @@ static void set_up_replay_mode(struct repository *repo,\n \tstrset_clear(&rinfo.positive_refs);\n }\n \n-static struct commit *mapped_commit(kh_oid_map_t *replayed_commits,\n-\t\t\t\t struct commit *commit,\n-\t\t\t\t struct commit *fallback)\n+static struct commit *get_mapped_commit(kh_oid_map_t *replayed_commits,\n+\t\t\t\t\tstruct commit *commit,\n+\t\t\t\t\tstruct commit *fallback)\n {\n \tkhint_t pos;\n \tif (!commit)\n@@ -256,6 +256,21 @@ static struct commit *mapped_commit(kh_oid_map_t *replayed_commits,\n \treturn kh_value(replayed_commits, pos);\n }\n \n+static void put_mapped_commit(kh_oid_map_t *replayed_commits,\n+\t\t\t struct commit *commit,\n+\t\t\t struct commit *new_commit)\n+{\n+\tkhint_t pos;\n+\tint ret;\n+\n+\tpos = kh_put_oid_map(replayed_commits, commit->object.oid, &ret);\n+\tif (ret == 0)\n+\t\tBUG(\"Duplicate rewritten commit: %s\\n\",\n+\t\t oid_to_hex(&commit->object.oid));\n+\n+\tkh_value(replayed_commits, pos) = new_commit;\n+}\n+\n static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t struct commit *pickme,\n \t\t\t\t\t kh_oid_map_t *replayed_commits,\n@@ -276,7 +291,7 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\tbase_tree = lookup_tree(repo, repo->hash_algo->empty_tree);\n \t}\n \n-\treplayed_base = mapped_commit(replayed_commits, base, onto);\n+\treplayed_base = get_mapped_commit(replayed_commits, base, onto);\n \treplayed_base_tree = repo_get_commit_tree(repo, replayed_base);\n \tpickme_tree = repo_get_commit_tree(repo, pickme);\n \n@@ -414,8 +429,6 @@ int replay_revisions(struct rev_info *revs,\n \treplayed_commits = kh_init_oid_map();\n \twhile ((commit = get_revision(revs))) {\n \t\tconst struct name_decoration *decoration;\n-\t\tkhint_t pos;\n-\t\tint hr;\n \n \t\tif (commit->parents && commit->parents->next)\n \t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n@@ -427,11 +440,7 @@ int replay_revisions(struct rev_info *revs,\n \t\t\tbreak;\n \n \t\t/* Record commit -> last_commit mapping */\n-\t\tpos = kh_put_oid_map(replayed_commits, commit->object.oid, &hr);\n-\t\tif (hr == 0)\n-\t\t\tBUG(\"Duplicate rewritten commit: %s\\n\",\n-\t\t\t oid_to_hex(&commit->object.oid));\n-\t\tkh_value(replayed_commits, pos) = last_commit;\n+\t\tput_mapped_commit(replayed_commits, commit, last_commit);\n \n \t\t/* Update any necessary branches */\n \t\tif (ref)\n\n-- \n2.53.0.1323.g189a785ab5\n\n"},{"id":"545641","messageId":"20260616-toon-git-replay-drop-merges-v3-3-153e9eb99ce1@iotcl.com","threadId":"65771","inReplyTo":"20260616-toon-git-replay-drop-merges-v3-0-153e9eb99ce1@iotcl.com","subject":"[PATCH v3 3/3] replay: offer an option to linearize the commit topology","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-16T09:26:37Z","receivedAt":"2026-06-16T09:27:27Z","isPatch":true,"body":"From: Johannes Schindelin \n\nOne of the stated goals of git-replay(1) is to allow implementing the\ngit-rebase(1) functionality on the server side.\n\nThe default mode of git-rebase(1) is to act as if `--no-rebase-merges`\nwas given. This mode drops merge commits instead of replaying them, and\nlinearizes the commit history into a sequence of the\nregular (single-parent) commits.\n\nAdd option `--linearize` to git-replay(1) to do the same.\n\nCo-authored-by: Toon Claes \nSigned-off-by: Johannes Schindelin \nSigned-off-by: Toon Claes \n---\n Documentation/git-replay.adoc | 8 ++++-\n builtin/replay.c | 6 +++-\n replay.c | 32 +++++++++++++++-----\n replay.h | 5 ++++\n t/t3650-replay-basics.sh | 68 ++++++++++++++++++++++++++++++++++++++++++-\n 5 files changed, 109 insertions(+), 10 deletions(-)\n\ndiff --git a/Documentation/git-replay.adoc b/Documentation/git-replay.adoc\nindex a32f72aead..ef56ee0f1b 100644\n--- a/Documentation/git-replay.adoc\n+++ b/Documentation/git-replay.adoc\n@@ -10,7 +10,7 @@ SYNOPSIS\n --------\n [verse]\n (EXPERIMENTAL!) 'git replay' ([--contained] --onto= | --advance= | --revert=)\n-\t\t\t [--ref=] [--ref-action=] \n+\t\t\t [--ref=] [--ref-action=] [--linearize] \n \n DESCRIPTION\n -----------\n@@ -88,6 +88,12 @@ incompatible with `--contained` (which is a modifier for `--onto` only).\n +\n The default mode can be configured via the `replay.refAction` configuration variable.\n \n+--linearize::\n+\tIn this mode, `git replay` imitates `git rebase --no-rebase-merges`,\n+\ti.e. it cherry-picks only non-merge commits, each one on top of the\n+\tprevious one.\n+\tThis option is incompatible with `--revert`.\n+\n ::\n \tRange of commits to replay; see \"Specifying Ranges\" in\n \tlinkgit:git-rev-parse[1]. In `--advance=` or\ndiff --git a/builtin/replay.c b/builtin/replay.c\nindex 39e3a86f6c..62962c73c7 100644\n--- a/builtin/replay.c\n+++ b/builtin/replay.c\n@@ -85,7 +85,7 @@ int cmd_replay(int argc,\n \tconst char *const replay_usage[] = {\n \t\tN_(\"(EXPERIMENTAL!) git replay \"\n \t\t \"([--contained] --onto= | --advance= | --revert=)\\n\"\n-\t\t \"[--ref=] [--ref-action=] \"),\n+\t\t \"[--ref=] [--ref-action=] [--linearize] \"),\n \t\tNULL\n \t};\n \tstruct option replay_options[] = {\n@@ -111,6 +111,8 @@ int cmd_replay(int argc,\n \t\t\t N_(\"mode\"),\n \t\t\t N_(\"control ref update behavior (update|print)\"),\n \t\t\t PARSE_OPT_NONEG),\n+\t\tOPT_BOOL(0, \"linearize\", &opts.linearize,\n+\t\t\t N_(\"drop merge commits, replaying only non-merge commits\")),\n \t\tOPT_END()\n \t};\n \n@@ -132,6 +134,8 @@ int cmd_replay(int argc,\n \t\t\t\t opts.contained, \"--contained\");\n \tdie_for_incompatible_opt2(!!opts.ref, \"--ref\",\n \t\t\t\t !!opts.contained, \"--contained\");\n+\tdie_for_incompatible_opt2(!!opts.revert, \"--revert\",\n+\t\t\t\t opts.linearize, \"--linearize\");\n \n \t/* Parse ref action mode from command line or config */\n \tref_mode = get_ref_action_mode(repo, ref_action);\ndiff --git a/replay.c b/replay.c\nindex 7921d7dba3..5539daff00 100644\n--- a/replay.c\n+++ b/replay.c\n@@ -277,12 +277,16 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t struct commit *onto,\n \t\t\t\t\t struct merge_options *merge_opt,\n \t\t\t\t\t struct merge_result *result,\n+\t\t\t\t\t struct commit *replayed_base,\n \t\t\t\t\t bool reverse,\n \t\t\t\t\t enum replay_empty_commit_action empty)\n {\n-\tstruct commit *base, *replayed_base;\n+\tstruct commit *base;\n \tstruct tree *pickme_tree, *base_tree, *replayed_base_tree;\n \n+\tif (replayed_base && reverse)\n+\t\tBUG(\"Linearizing commits is not supported when replaying in reverse\");\n+\n \tif (pickme->parents) {\n \t\tbase = pickme->parents->item;\n \t\tbase_tree = repo_get_commit_tree(repo, base);\n@@ -291,7 +295,8 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\tbase_tree = lookup_tree(repo, repo->hash_algo->empty_tree);\n \t}\n \n-\treplayed_base = get_mapped_commit(replayed_commits, base, onto);\n+\tif (!replayed_base)\n+\t\treplayed_base = get_mapped_commit(replayed_commits, base, onto);\n \treplayed_base_tree = repo_get_commit_tree(repo, replayed_base);\n \tpickme_tree = repo_get_commit_tree(repo, pickme);\n \n@@ -430,12 +435,25 @@ int replay_revisions(struct rev_info *revs,\n \twhile ((commit = get_revision(revs))) {\n \t\tconst struct name_decoration *decoration;\n \n-\t\tif (commit->parents && commit->parents->next)\n-\t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n+\t\tif (commit->parents && commit->parents->next) {\n+\t\t\tif (!opts->linearize)\n+\t\t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n+\t\t\t/*\n+\t\t\t * Drop the merge commit: do not pick it and leave\n+\t\t\t * last_commit unchanged, so its children (and any ref\n+\t\t\t * pointing at it) are reparented onto the previous\n+\t\t\t * non-merge commit, which the ref-update loop below uses.\n+\t\t\t */\n+\t\t} else {\n+\t\t\tstruct commit *to_pick = reverse ? last_commit : onto;\n+\t\t\tlast_commit =\n+\t\t\t\tpick_regular_commit(revs->repo, commit,\n+\t\t\t\t\t\t replayed_commits, to_pick,\n+\t\t\t\t\t\t &merge_opt, &result,\n+\t\t\t\t\t\t opts->linearize ? last_commit : NULL,\n+\t\t\t\t\t\t reverse, opts->empty);\n+\t\t}\n \n-\t\tlast_commit = pick_regular_commit(revs->repo, commit, replayed_commits,\n-\t\t\t\t\t\t reverse ? last_commit : onto,\n-\t\t\t\t\t\t &merge_opt, &result, reverse, opts->empty);\n \t\tif (!last_commit)\n \t\t\tbreak;\n \ndiff --git a/replay.h b/replay.h\nindex 1851a07705..07e6fdcca3 100644\n--- a/replay.h\n+++ b/replay.h\n@@ -62,6 +62,11 @@ struct replay_revisions_options {\n \t * Defaults to REPLAY_EMPTY_COMMIT_DROP.\n \t */\n \tenum replay_empty_commit_action empty;\n+\n+\t/*\n+\t * Whether to linearize the commits (i.e. drop merge commits).\n+\t */\n+\tint linearize;\n };\n \n /* This struct is used as an out-parameter by `replay_revisions()`. */\ndiff --git a/t/t3650-replay-basics.sh b/t/t3650-replay-basics.sh\nindex 3353bc4a4d..1874d06769 100755\n--- a/t/t3650-replay-basics.sh\n+++ b/t/t3650-replay-basics.sh\n@@ -52,8 +52,12 @@ test_expect_success 'setup' '\n \ttest_merge P O --no-ff &&\n \tgit switch main &&\n \n+\tgit switch --orphan unrelated &&\n+\ttest_commit unrelated-root &&\n+\n \tgit switch -c conflict B &&\n-\ttest_commit C.conflict C.t conflict\n+\ttest_commit C.conflict C.t conflict &&\n+\tgit branch -D unrelated\n '\n \n test_expect_success 'setup bare' '\n@@ -97,6 +101,12 @@ test_expect_success '--advance and --contained cannot be used together' '\n \ttest_grep \"cannot be used together\" actual\n '\n \n+test_expect_success '--revert and --linearize cannot be used together' '\n+\ttest_must_fail git replay --revert=main --linearize \\\n+\t\ttopic1..topic2 2>actual &&\n+\ttest_grep \"cannot be used together\" actual\n+'\n+\n test_expect_success 'cannot advance target ... ordering would be ill-defined' '\n \techo \"fatal: ${SQ}--advance${SQ} cannot be used with multiple revision ranges because the ordering would be ill-defined\" >expect &&\n \ttest_must_fail git replay --advance=main main topic1 topic2 2>actual &&\n@@ -565,4 +575,60 @@ test_expect_success '--onto with --ref rejects multiple revision ranges' '\n \ttest_grep \"cannot be used with multiple revision ranges\" err\n '\n \n+test_expect_success 'replay to rebase merge commit with --linearize' '\n+\tgit replay --ref-action=print --linearize \\\n+\t\t--onto main I..topic-with-merge >result &&\n+\n+\ttest_line_count = 1 result &&\n+\n+\tgit log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n+\ttest_write_lines O N J M L B A >expect &&\n+\ttest_cmp expect actual\n+'\n+\n+test_expect_success 'replay to rebase merge commit with --linearize down to the root commit' '\n+\tgit replay --ref-action=print --linearize \\\n+\t\t--onto unrelated-root topic-with-merge >result &&\n+\n+\ttest_line_count = 1 result &&\n+\n+\tgit log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n+\ttest_write_lines O N J I B A unrelated-root >expect &&\n+\ttest_cmp expect actual\n+'\n+\n+test_expect_success 'replay to cherry-pick merge commit with --linearize' '\n+\tgit replay --ref-action=print --linearize \\\n+\t\t--advance main I..topic-with-merge >result &&\n+\n+\ttest_line_count = 1 result &&\n+\n+\tgit log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n+\ttest_write_lines O N J M L B A >expect &&\n+\ttest_cmp expect actual &&\n+\n+\tprintf \"update refs/heads/main \" >expect &&\n+\tprintf \"%s \" $(cut -f 3 -d \" \" result) >>expect &&\n+\tgit rev-parse main >>expect &&\n+\ttest_cmp expect result\n+'\n+\n+test_expect_success 'replay --linearize produces the same patches' '\n+\tgit replay --ref-action=print --linearize \\\n+\t\t--onto main I..topic-with-merge >result &&\n+\n+\ttest_line_count = 1 result &&\n+\ttip=$(cut -f 3 -d \" \" result) &&\n+\n+\t# range-diff does not care about the dropped merge,\n+\t# so the original commits (I..topic-with-merge)\n+\t# and the replayed chain (main..tip) must produce identical patches.\n+\tgit range-diff I..topic-with-merge main..$tip >out &&\n+\ttest_file_not_empty out &&\n+\t! grep -v \"=\" out &&\n+\n+\tgit log --oneline main..$tip >out &&\n+\ttest_line_count = 3 out\n+'\n+\n test_done\n\n-- \n2.53.0.1323.g189a785ab5\n\n"},{"id":"546171","messageId":"20260622-toon-git-replay-drop-merges-v4-0-ff257f534319@iotcl.com","threadId":"65771","inReplyTo":"20260616-toon-git-replay-drop-merges-v3-0-153e9eb99ce1@iotcl.com","subject":"[PATCH v4 0/3] Teach git-replay(1) to linearize merge commits","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-22T12:41:54Z","receivedAt":"2026-06-22T12:42:05Z","isPatch":true,"body":"As an alternative to dscho's patch series to replay merges[1], add\noption to git-replay(1) to linearize merges. This mimics what\ngit-rebase(1) does too with --no-rebase-merges (the default).\n\nThe first two patches do some refactoring. The third patch implements\nthe actual change. This patch was kindly provided by Dscho, which I've\ntweaked to be upstreamed.\n\nThe --linearize option is only added to git-replay(1) and not to\ngit-history(1) because in my opinion it doesn't make much sense to do\nso, but I'm happy to hear if anyone disagrees.\n\nThis series might conflict with Kristoffer's series to make\ndocumentation changes[2], but should be trivial to resolve. And I don't\nthink there's a conflict with Patrick's series on adding \"drop\" to\ngit-history(1)[3].\n\ndscho's series to replay merges[1] needs a bit of rework to fit on top\nof this, but I'm happy to help figuring that out. We've been discussing\nto either name the option --flatten or --linearize, but I've decided on\n\"linearize\" because the documentation of git-rebase(1) also mentions\n\"linearize\".\n\n[1]: \n[2]: \n[3]: <20260603-b4-pks-history-drop-v2-0-742cb5b5176d@pks.im>\n\n---\nChanges in v4:\n- Use test_grep instead of a bare grep in the range-diff test, to\n prepare for mm/test-grep-lint.\n- Link to v3: https://patch.msgid.link/20260616-toon-git-replay-drop-merges-v3-0-153e9eb99ce1@iotcl.com\n\nChanges in v3:\n- Add --linearize to Documentation SYNOPSIS, and mention it's\n incompatible with --revert.\n- Small language change in help message for --linearize.\n- Rephrase comment to include last_commit isn't modified when\n linearizing merges.\n- Remove test that was added in earlier versions, but actually is\n a duplicate of 'replaying merge commits is not supported yet'.\n- Add test to verify --revert and --linearize are incompatible.\n- Properly test that replaying down to root with --linearize works.\n- Add test for --linearize with --advance.\n- Add test that uses git-range-diff(1) to verify the patches created by\n --linearize are correct.\n- Link to v2: https://patch.msgid.link/20260610-toon-git-replay-drop-merges-v2-0-5714a71c6d83@iotcl.com\n\nChanges in v2:\n- Restructured the conditions to detect merge commits and added a line\n of comment why the loop continues.\n- Rewrote tests to use the history from the setup step and added a few\n test cases.\n- Re-added Johannes's Signed-off-by trailer. Johannes gave me the\n patches with this trailer, and if I understand correctly, I can keep\n it. Please let me know if that wrong.\n- Link to v1: https://patch.msgid.link/20260608-toon-git-replay-drop-merges-v1-0-e3ee71fce7b4@iotcl.com\n\n---\nJohannes Schindelin (1):\n replay: offer an option to linearize the commit topology\n\nToon Claes (2):\n replay: refactor enum replay_mode into a bool\n replay: add helper to put entry into mapped_commits\n\n Documentation/git-replay.adoc | 8 ++-\n builtin/replay.c | 6 ++-\n replay.c | 116 ++++++++++++++++++++++++------------------\n replay.h | 5 ++\n t/t3650-replay-basics.sh | 68 ++++++++++++++++++++++++-\n 5 files changed, 151 insertions(+), 52 deletions(-)\n\nRange-diff versus v3:\n\n1: 759fa1b52c = 1: 0f0e50c67f replay: refactor enum replay_mode into a bool\n2: 68dd5ad77c = 2: 919a6495ee replay: add helper to put entry into mapped_commits\n3: f99aeb3887 ! 3: bb03e78210 replay: offer an option to linearize the commit topology\n @@ t/t3650-replay-basics.sh: test_expect_success '--onto with --ref rejects multipl\n +\t# and the replayed chain (main..tip) must produce identical patches.\n +\tgit range-diff I..topic-with-merge main..$tip >out &&\n +\ttest_file_not_empty out &&\n -+\t! grep -v \"=\" out &&\n ++\ttest_grep ! -v \"=\" out &&\n +\n +\tgit log --oneline main..$tip >out &&\n +\ttest_line_count = 3 out\n\n\n---\nbase-commit: 9ac3f193c05c2237e2b14ebaa1149e9fc8a1abe0\nchange-id: 20260604-toon-git-replay-drop-merges-807fa008d395\n\n"},{"id":"546172","messageId":"20260622-toon-git-replay-drop-merges-v4-1-ff257f534319@iotcl.com","threadId":"65771","inReplyTo":"20260622-toon-git-replay-drop-merges-v4-0-ff257f534319@iotcl.com","subject":"[PATCH v4 1/3] replay: refactor enum replay_mode into a bool","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-22T12:41:55Z","receivedAt":"2026-06-22T12:42:17Z","isPatch":true,"body":"In 2760ee4983 (replay: add --revert mode to reverse commit changes,\n2026-03-26) the enum `replay_mode` was introduced. This has two possible\nvalues:\n\n - The value `REPLAY_MODE_REVERT` is used when option `--revert` is\n passed to git-replay(1). When using this value the commits are\n processed in reverse order and the inverse of the changes are\n applied.\n\n - The value `REPLAY_MODE_PICK` is used when either option `--onto` or\n `--advance` is used. In both cases the commits are processed in\n normal order, and the changes are applied as-is.\n\nSince there are only two possible values of this enum, simplify the code\nby converting the enum into a bool. This avoids adding code paths that\ncheck for invalid values of the enum, and shortens code where the value\nis checked with a ternary operator.\n\nSigned-off-by: Toon Claes \n---\n replay.c | 59 +++++++++++++++++++++++++----------------------------------\n 1 file changed, 25 insertions(+), 34 deletions(-)\n\ndiff --git a/replay.c b/replay.c\nindex 4ef8abb607..1f8e5b083b 100644\n--- a/replay.c\n+++ b/replay.c\n@@ -18,11 +18,6 @@\n */\n #define the_repository DO_NOT_USE_THE_REPOSITORY\n \n-enum replay_mode {\n-\tREPLAY_MODE_PICK,\n-\tREPLAY_MODE_REVERT,\n-};\n-\n static const char *short_commit_name(struct repository *repo,\n \t\t\t\t struct commit *commit)\n {\n@@ -81,7 +76,7 @@ static struct commit *create_commit(struct repository *repo,\n \t\t\t\t struct tree *tree,\n \t\t\t\t struct commit *based_on,\n \t\t\t\t struct commit *parent,\n-\t\t\t\t enum replay_mode mode)\n+\t\t\t\t bool reverse)\n {\n \tstruct object_id ret;\n \tstruct object *obj = NULL;\n@@ -98,15 +93,13 @@ static struct commit *create_commit(struct repository *repo,\n \n \tcommit_list_insert(parent, &parents);\n \textra = read_commit_extra_headers(based_on, exclude_gpgsig);\n-\tif (mode == REPLAY_MODE_REVERT) {\n+\tif (reverse) {\n \t\tgenerate_revert_message(&msg, based_on, repo);\n \t\t/* For revert, use current user as author (NULL = use default) */\n-\t} else if (mode == REPLAY_MODE_PICK) {\n+\t} else {\n \t\tfind_commit_subject(message, &orig_message);\n \t\tstrbuf_addstr(&msg, orig_message);\n \t\tauthor = get_author(message);\n-\t} else {\n-\t\tBUG(\"unexpected replay mode %d\", mode);\n \t}\n \treset_ident_date();\n \tif (commit_tree_extended(msg.buf, msg.len, &tree->object.oid, parents,\n@@ -269,7 +262,7 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t struct commit *onto,\n \t\t\t\t\t struct merge_options *merge_opt,\n \t\t\t\t\t struct merge_result *result,\n-\t\t\t\t\t enum replay_mode mode,\n+\t\t\t\t\t bool reverse,\n \t\t\t\t\t enum replay_empty_commit_action empty)\n {\n \tstruct commit *base, *replayed_base;\n@@ -287,7 +280,21 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \treplayed_base_tree = repo_get_commit_tree(repo, replayed_base);\n \tpickme_tree = repo_get_commit_tree(repo, pickme);\n \n-\tif (mode == REPLAY_MODE_PICK) {\n+\tif (reverse) {\n+\t\t/* Revert: swap base and pickme to reverse the diff */\n+\t\tconst char *pickme_name = short_commit_name(repo, pickme);\n+\t\tmerge_opt->branch1 = short_commit_name(repo, replayed_base);\n+\t\tmerge_opt->branch2 = xstrfmt(\"parent of %s\", pickme_name);\n+\t\tmerge_opt->ancestor = pickme_name;\n+\n+\t\tmerge_incore_nonrecursive(merge_opt,\n+\t\t\t\t\t pickme_tree,\n+\t\t\t\t\t replayed_base_tree,\n+\t\t\t\t\t base_tree,\n+\t\t\t\t\t result);\n+\n+\t\tfree((char *)merge_opt->branch2);\n+\t} else {\n \t\t/* Cherry-pick: normal order */\n \t\tmerge_opt->branch1 = short_commit_name(repo, replayed_base);\n \t\tmerge_opt->branch2 = short_commit_name(repo, pickme);\n@@ -303,22 +310,6 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t result);\n \n \t\tfree((char *)merge_opt->ancestor);\n-\t} else if (mode == REPLAY_MODE_REVERT) {\n-\t\t/* Revert: swap base and pickme to reverse the diff */\n-\t\tconst char *pickme_name = short_commit_name(repo, pickme);\n-\t\tmerge_opt->branch1 = short_commit_name(repo, replayed_base);\n-\t\tmerge_opt->branch2 = xstrfmt(\"parent of %s\", pickme_name);\n-\t\tmerge_opt->ancestor = pickme_name;\n-\n-\t\tmerge_incore_nonrecursive(merge_opt,\n-\t\t\t\t\t pickme_tree,\n-\t\t\t\t\t replayed_base_tree,\n-\t\t\t\t\t base_tree,\n-\t\t\t\t\t result);\n-\n-\t\tfree((char *)merge_opt->branch2);\n-\t} else {\n-\t\tBUG(\"unexpected replay mode %d\", mode);\n \t}\n \tmerge_opt->ancestor = NULL;\n \tmerge_opt->branch2 = NULL;\n@@ -341,7 +332,7 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t}\n \t}\n \n-\treturn create_commit(repo, result->tree, pickme, replayed_base, mode);\n+\treturn create_commit(repo, result->tree, pickme, replayed_base, reverse);\n }\n \n void replay_result_release(struct replay_result *result)\n@@ -381,13 +372,13 @@ int replay_revisions(struct rev_info *revs,\n \tchar *revert;\n \tconst char *ref;\n \tstruct object_id old_oid;\n-\tenum replay_mode mode = REPLAY_MODE_PICK;\n+\tbool reverse;\n \tint ret;\n \n \tadvance = xstrdup_or_null(opts->advance);\n \trevert = xstrdup_or_null(opts->revert);\n-\tif (revert)\n-\t\tmode = REPLAY_MODE_REVERT;\n+\treverse = !!revert;\n+\n \tset_up_replay_mode(revs->repo, &revs->cmdline, opts->onto,\n \t\t\t &detached_head, &advance, &revert, &onto, &update_refs);\n \n@@ -430,8 +421,8 @@ int replay_revisions(struct rev_info *revs,\n \t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n \n \t\tlast_commit = pick_regular_commit(revs->repo, commit, replayed_commits,\n-\t\t\t\t\t\t mode == REPLAY_MODE_REVERT ? last_commit : onto,\n-\t\t\t\t\t\t &merge_opt, &result, mode, opts->empty);\n+\t\t\t\t\t\t reverse ? last_commit : onto,\n+\t\t\t\t\t\t &merge_opt, &result, reverse, opts->empty);\n \t\tif (!last_commit)\n \t\t\tbreak;\n \n\n-- \n2.53.0.1323.g189a785ab5\n\n"},{"id":"546173","messageId":"20260622-toon-git-replay-drop-merges-v4-2-ff257f534319@iotcl.com","threadId":"65771","inReplyTo":"20260622-toon-git-replay-drop-merges-v4-0-ff257f534319@iotcl.com","subject":"[PATCH v4 2/3] replay: add helper to put entry into mapped_commits","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-22T12:41:56Z","receivedAt":"2026-06-22T12:42:21Z","isPatch":true,"body":"The function replay_revisions() in replay.c is rather lengthy. Extract\nthe logic to put a commit entry into mapped_commits into a helper\nfunction put_mapped_commit().\n\nWhile at it, rename mapped_commit() to get_mapped_commit() to pair with\nthis new function.\n\nSigned-off-by: Toon Claes \n---\n replay.c | 31 ++++++++++++++++++++-----------\n 1 file changed, 20 insertions(+), 11 deletions(-)\n\ndiff --git a/replay.c b/replay.c\nindex 1f8e5b083b..7921d7dba3 100644\n--- a/replay.c\n+++ b/replay.c\n@@ -243,9 +243,9 @@ static void set_up_replay_mode(struct repository *repo,\n \tstrset_clear(&rinfo.positive_refs);\n }\n \n-static struct commit *mapped_commit(kh_oid_map_t *replayed_commits,\n-\t\t\t\t struct commit *commit,\n-\t\t\t\t struct commit *fallback)\n+static struct commit *get_mapped_commit(kh_oid_map_t *replayed_commits,\n+\t\t\t\t\tstruct commit *commit,\n+\t\t\t\t\tstruct commit *fallback)\n {\n \tkhint_t pos;\n \tif (!commit)\n@@ -256,6 +256,21 @@ static struct commit *mapped_commit(kh_oid_map_t *replayed_commits,\n \treturn kh_value(replayed_commits, pos);\n }\n \n+static void put_mapped_commit(kh_oid_map_t *replayed_commits,\n+\t\t\t struct commit *commit,\n+\t\t\t struct commit *new_commit)\n+{\n+\tkhint_t pos;\n+\tint ret;\n+\n+\tpos = kh_put_oid_map(replayed_commits, commit->object.oid, &ret);\n+\tif (ret == 0)\n+\t\tBUG(\"Duplicate rewritten commit: %s\\n\",\n+\t\t oid_to_hex(&commit->object.oid));\n+\n+\tkh_value(replayed_commits, pos) = new_commit;\n+}\n+\n static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t struct commit *pickme,\n \t\t\t\t\t kh_oid_map_t *replayed_commits,\n@@ -276,7 +291,7 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\tbase_tree = lookup_tree(repo, repo->hash_algo->empty_tree);\n \t}\n \n-\treplayed_base = mapped_commit(replayed_commits, base, onto);\n+\treplayed_base = get_mapped_commit(replayed_commits, base, onto);\n \treplayed_base_tree = repo_get_commit_tree(repo, replayed_base);\n \tpickme_tree = repo_get_commit_tree(repo, pickme);\n \n@@ -414,8 +429,6 @@ int replay_revisions(struct rev_info *revs,\n \treplayed_commits = kh_init_oid_map();\n \twhile ((commit = get_revision(revs))) {\n \t\tconst struct name_decoration *decoration;\n-\t\tkhint_t pos;\n-\t\tint hr;\n \n \t\tif (commit->parents && commit->parents->next)\n \t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n@@ -427,11 +440,7 @@ int replay_revisions(struct rev_info *revs,\n \t\t\tbreak;\n \n \t\t/* Record commit -> last_commit mapping */\n-\t\tpos = kh_put_oid_map(replayed_commits, commit->object.oid, &hr);\n-\t\tif (hr == 0)\n-\t\t\tBUG(\"Duplicate rewritten commit: %s\\n\",\n-\t\t\t oid_to_hex(&commit->object.oid));\n-\t\tkh_value(replayed_commits, pos) = last_commit;\n+\t\tput_mapped_commit(replayed_commits, commit, last_commit);\n \n \t\t/* Update any necessary branches */\n \t\tif (ref)\n\n-- \n2.53.0.1323.g189a785ab5\n\n"},{"id":"546174","messageId":"20260622-toon-git-replay-drop-merges-v4-3-ff257f534319@iotcl.com","threadId":"65771","inReplyTo":"20260622-toon-git-replay-drop-merges-v4-0-ff257f534319@iotcl.com","subject":"[PATCH v4 3/3] replay: offer an option to linearize the commit topology","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-22T12:41:57Z","receivedAt":"2026-06-22T12:42:25Z","isPatch":true,"body":"From: Johannes Schindelin \n\nOne of the stated goals of git-replay(1) is to allow implementing the\ngit-rebase(1) functionality on the server side.\n\nThe default mode of git-rebase(1) is to act as if `--no-rebase-merges`\nwas given. This mode drops merge commits instead of replaying them, and\nlinearizes the commit history into a sequence of the\nregular (single-parent) commits.\n\nAdd option `--linearize` to git-replay(1) to do the same.\n\nCo-authored-by: Toon Claes \nSigned-off-by: Johannes Schindelin \nSigned-off-by: Toon Claes \n---\n Documentation/git-replay.adoc | 8 ++++-\n builtin/replay.c | 6 +++-\n replay.c | 32 +++++++++++++++-----\n replay.h | 5 ++++\n t/t3650-replay-basics.sh | 68 ++++++++++++++++++++++++++++++++++++++++++-\n 5 files changed, 109 insertions(+), 10 deletions(-)\n\ndiff --git a/Documentation/git-replay.adoc b/Documentation/git-replay.adoc\nindex a32f72aead..ef56ee0f1b 100644\n--- a/Documentation/git-replay.adoc\n+++ b/Documentation/git-replay.adoc\n@@ -10,7 +10,7 @@ SYNOPSIS\n --------\n [verse]\n (EXPERIMENTAL!) 'git replay' ([--contained] --onto= | --advance= | --revert=)\n-\t\t\t [--ref=] [--ref-action=] \n+\t\t\t [--ref=] [--ref-action=] [--linearize] \n \n DESCRIPTION\n -----------\n@@ -88,6 +88,12 @@ incompatible with `--contained` (which is a modifier for `--onto` only).\n +\n The default mode can be configured via the `replay.refAction` configuration variable.\n \n+--linearize::\n+\tIn this mode, `git replay` imitates `git rebase --no-rebase-merges`,\n+\ti.e. it cherry-picks only non-merge commits, each one on top of the\n+\tprevious one.\n+\tThis option is incompatible with `--revert`.\n+\n ::\n \tRange of commits to replay; see \"Specifying Ranges\" in\n \tlinkgit:git-rev-parse[1]. In `--advance=` or\ndiff --git a/builtin/replay.c b/builtin/replay.c\nindex 39e3a86f6c..62962c73c7 100644\n--- a/builtin/replay.c\n+++ b/builtin/replay.c\n@@ -85,7 +85,7 @@ int cmd_replay(int argc,\n \tconst char *const replay_usage[] = {\n \t\tN_(\"(EXPERIMENTAL!) git replay \"\n \t\t \"([--contained] --onto= | --advance= | --revert=)\\n\"\n-\t\t \"[--ref=] [--ref-action=] \"),\n+\t\t \"[--ref=] [--ref-action=] [--linearize] \"),\n \t\tNULL\n \t};\n \tstruct option replay_options[] = {\n@@ -111,6 +111,8 @@ int cmd_replay(int argc,\n \t\t\t N_(\"mode\"),\n \t\t\t N_(\"control ref update behavior (update|print)\"),\n \t\t\t PARSE_OPT_NONEG),\n+\t\tOPT_BOOL(0, \"linearize\", &opts.linearize,\n+\t\t\t N_(\"drop merge commits, replaying only non-merge commits\")),\n \t\tOPT_END()\n \t};\n \n@@ -132,6 +134,8 @@ int cmd_replay(int argc,\n \t\t\t\t opts.contained, \"--contained\");\n \tdie_for_incompatible_opt2(!!opts.ref, \"--ref\",\n \t\t\t\t !!opts.contained, \"--contained\");\n+\tdie_for_incompatible_opt2(!!opts.revert, \"--revert\",\n+\t\t\t\t opts.linearize, \"--linearize\");\n \n \t/* Parse ref action mode from command line or config */\n \tref_mode = get_ref_action_mode(repo, ref_action);\ndiff --git a/replay.c b/replay.c\nindex 7921d7dba3..5539daff00 100644\n--- a/replay.c\n+++ b/replay.c\n@@ -277,12 +277,16 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\t\t\t\t struct commit *onto,\n \t\t\t\t\t struct merge_options *merge_opt,\n \t\t\t\t\t struct merge_result *result,\n+\t\t\t\t\t struct commit *replayed_base,\n \t\t\t\t\t bool reverse,\n \t\t\t\t\t enum replay_empty_commit_action empty)\n {\n-\tstruct commit *base, *replayed_base;\n+\tstruct commit *base;\n \tstruct tree *pickme_tree, *base_tree, *replayed_base_tree;\n \n+\tif (replayed_base && reverse)\n+\t\tBUG(\"Linearizing commits is not supported when replaying in reverse\");\n+\n \tif (pickme->parents) {\n \t\tbase = pickme->parents->item;\n \t\tbase_tree = repo_get_commit_tree(repo, base);\n@@ -291,7 +295,8 @@ static struct commit *pick_regular_commit(struct repository *repo,\n \t\tbase_tree = lookup_tree(repo, repo->hash_algo->empty_tree);\n \t}\n \n-\treplayed_base = get_mapped_commit(replayed_commits, base, onto);\n+\tif (!replayed_base)\n+\t\treplayed_base = get_mapped_commit(replayed_commits, base, onto);\n \treplayed_base_tree = repo_get_commit_tree(repo, replayed_base);\n \tpickme_tree = repo_get_commit_tree(repo, pickme);\n \n@@ -430,12 +435,25 @@ int replay_revisions(struct rev_info *revs,\n \twhile ((commit = get_revision(revs))) {\n \t\tconst struct name_decoration *decoration;\n \n-\t\tif (commit->parents && commit->parents->next)\n-\t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n+\t\tif (commit->parents && commit->parents->next) {\n+\t\t\tif (!opts->linearize)\n+\t\t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n+\t\t\t/*\n+\t\t\t * Drop the merge commit: do not pick it and leave\n+\t\t\t * last_commit unchanged, so its children (and any ref\n+\t\t\t * pointing at it) are reparented onto the previous\n+\t\t\t * non-merge commit, which the ref-update loop below uses.\n+\t\t\t */\n+\t\t} else {\n+\t\t\tstruct commit *to_pick = reverse ? last_commit : onto;\n+\t\t\tlast_commit =\n+\t\t\t\tpick_regular_commit(revs->repo, commit,\n+\t\t\t\t\t\t replayed_commits, to_pick,\n+\t\t\t\t\t\t &merge_opt, &result,\n+\t\t\t\t\t\t opts->linearize ? last_commit : NULL,\n+\t\t\t\t\t\t reverse, opts->empty);\n+\t\t}\n \n-\t\tlast_commit = pick_regular_commit(revs->repo, commit, replayed_commits,\n-\t\t\t\t\t\t reverse ? last_commit : onto,\n-\t\t\t\t\t\t &merge_opt, &result, reverse, opts->empty);\n \t\tif (!last_commit)\n \t\t\tbreak;\n \ndiff --git a/replay.h b/replay.h\nindex 1851a07705..07e6fdcca3 100644\n--- a/replay.h\n+++ b/replay.h\n@@ -62,6 +62,11 @@ struct replay_revisions_options {\n \t * Defaults to REPLAY_EMPTY_COMMIT_DROP.\n \t */\n \tenum replay_empty_commit_action empty;\n+\n+\t/*\n+\t * Whether to linearize the commits (i.e. drop merge commits).\n+\t */\n+\tint linearize;\n };\n \n /* This struct is used as an out-parameter by `replay_revisions()`. */\ndiff --git a/t/t3650-replay-basics.sh b/t/t3650-replay-basics.sh\nindex 3353bc4a4d..b9ce6c4868 100755\n--- a/t/t3650-replay-basics.sh\n+++ b/t/t3650-replay-basics.sh\n@@ -52,8 +52,12 @@ test_expect_success 'setup' '\n \ttest_merge P O --no-ff &&\n \tgit switch main &&\n \n+\tgit switch --orphan unrelated &&\n+\ttest_commit unrelated-root &&\n+\n \tgit switch -c conflict B &&\n-\ttest_commit C.conflict C.t conflict\n+\ttest_commit C.conflict C.t conflict &&\n+\tgit branch -D unrelated\n '\n \n test_expect_success 'setup bare' '\n@@ -97,6 +101,12 @@ test_expect_success '--advance and --contained cannot be used together' '\n \ttest_grep \"cannot be used together\" actual\n '\n \n+test_expect_success '--revert and --linearize cannot be used together' '\n+\ttest_must_fail git replay --revert=main --linearize \\\n+\t\ttopic1..topic2 2>actual &&\n+\ttest_grep \"cannot be used together\" actual\n+'\n+\n test_expect_success 'cannot advance target ... ordering would be ill-defined' '\n \techo \"fatal: ${SQ}--advance${SQ} cannot be used with multiple revision ranges because the ordering would be ill-defined\" >expect &&\n \ttest_must_fail git replay --advance=main main topic1 topic2 2>actual &&\n@@ -565,4 +575,60 @@ test_expect_success '--onto with --ref rejects multiple revision ranges' '\n \ttest_grep \"cannot be used with multiple revision ranges\" err\n '\n \n+test_expect_success 'replay to rebase merge commit with --linearize' '\n+\tgit replay --ref-action=print --linearize \\\n+\t\t--onto main I..topic-with-merge >result &&\n+\n+\ttest_line_count = 1 result &&\n+\n+\tgit log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n+\ttest_write_lines O N J M L B A >expect &&\n+\ttest_cmp expect actual\n+'\n+\n+test_expect_success 'replay to rebase merge commit with --linearize down to the root commit' '\n+\tgit replay --ref-action=print --linearize \\\n+\t\t--onto unrelated-root topic-with-merge >result &&\n+\n+\ttest_line_count = 1 result &&\n+\n+\tgit log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n+\ttest_write_lines O N J I B A unrelated-root >expect &&\n+\ttest_cmp expect actual\n+'\n+\n+test_expect_success 'replay to cherry-pick merge commit with --linearize' '\n+\tgit replay --ref-action=print --linearize \\\n+\t\t--advance main I..topic-with-merge >result &&\n+\n+\ttest_line_count = 1 result &&\n+\n+\tgit log --format=%s $(cut -f 3 -d \" \" result) >actual &&\n+\ttest_write_lines O N J M L B A >expect &&\n+\ttest_cmp expect actual &&\n+\n+\tprintf \"update refs/heads/main \" >expect &&\n+\tprintf \"%s \" $(cut -f 3 -d \" \" result) >>expect &&\n+\tgit rev-parse main >>expect &&\n+\ttest_cmp expect result\n+'\n+\n+test_expect_success 'replay --linearize produces the same patches' '\n+\tgit replay --ref-action=print --linearize \\\n+\t\t--onto main I..topic-with-merge >result &&\n+\n+\ttest_line_count = 1 result &&\n+\ttip=$(cut -f 3 -d \" \" result) &&\n+\n+\t# range-diff does not care about the dropped merge,\n+\t# so the original commits (I..topic-with-merge)\n+\t# and the replayed chain (main..tip) must produce identical patches.\n+\tgit range-diff I..topic-with-merge main..$tip >out &&\n+\ttest_file_not_empty out &&\n+\ttest_grep ! -v \"=\" out &&\n+\n+\tgit log --oneline main..$tip >out &&\n+\ttest_line_count = 3 out\n+'\n+\n test_done\n\n-- \n2.53.0.1323.g189a785ab5\n\n"},{"id":"546183","messageId":"ajk-YQxLWfspNWIm@pks.im","threadId":"65771","inReplyTo":"20260622-toon-git-replay-drop-merges-v4-1-ff257f534319@iotcl.com","subject":"Re: [PATCH v4 1/3] replay: refactor enum replay_mode into a bool","fromName":"Patrick Steinhardt","fromEmail":"ps@pks.im","sentAt":"2026-06-22T13:53:37Z","receivedAt":"2026-06-22T13:53:43Z","isPatch":true,"body":"On Mon, Jun 22, 2026 at 02:41:55PM +0200, Toon Claes wrote:\n> In 2760ee4983 (replay: add --revert mode to reverse commit changes,\n> 2026-03-26) the enum `replay_mode` was introduced. This has two possible\n> values:\n> \n> - The value `REPLAY_MODE_REVERT` is used when option `--revert` is\n> passed to git-replay(1). When using this value the commits are\n> processed in reverse order and the inverse of the changes are\n> applied.\n> \n> - The value `REPLAY_MODE_PICK` is used when either option `--onto` or\n> `--advance` is used. In both cases the commits are processed in\n> normal order, and the changes are applied as-is.\n> \n> Since there are only two possible values of this enum, simplify the code\n> by converting the enum into a bool. This avoids adding code paths that\n> check for invalid values of the enum, and shortens code where the value\n> is checked with a ternary operator.\n\nThat's fair, and the result is easier to write. But is it really easier\nto read? And what if we ever have to create a third mode going forward?\n\nI'm generally no fan of booleans as parameters as they basically give\nyou no information at all at the callsite, except if you're lucky and\nyou already have an aptly-named variable available that you can pass.\nWhich seems to be the case here, but I'm still not sure whether this\nchange really improves the code.\n\nPatrick\n"},{"id":"546184","messageId":"ajk-Zok5pAClw7N3@pks.im","threadId":"65771","inReplyTo":"20260622-toon-git-replay-drop-merges-v4-2-ff257f534319@iotcl.com","subject":"Re: [PATCH v4 2/3] replay: add helper to put entry into mapped_commits","fromName":"Patrick Steinhardt","fromEmail":"ps@pks.im","sentAt":"2026-06-22T13:53:42Z","receivedAt":"2026-06-22T13:53:47Z","isPatch":true,"body":"On Mon, Jun 22, 2026 at 02:41:56PM +0200, Toon Claes wrote:\n> diff --git a/replay.c b/replay.c\n> index 1f8e5b083b..7921d7dba3 100644\n> --- a/replay.c\n> +++ b/replay.c\n> @@ -256,6 +256,21 @@ static struct commit *mapped_commit(kh_oid_map_t *replayed_commits,\n> \treturn kh_value(replayed_commits, pos);\n> }\n> \n> +static void put_mapped_commit(kh_oid_map_t *replayed_commits,\n> +\t\t\t struct commit *commit,\n> +\t\t\t struct commit *new_commit)\n> +{\n> +\tkhint_t pos;\n> +\tint ret;\n> +\n> +\tpos = kh_put_oid_map(replayed_commits, commit->object.oid, &ret);\n> +\tif (ret == 0)\n> +\t\tBUG(\"Duplicate rewritten commit: %s\\n\",\n> +\t\t oid_to_hex(&commit->object.oid));\n> +\n> +\tkh_value(replayed_commits, pos) = new_commit;\n> +}\n\nThe khash map interfaces are quite awkward to use, so having a small\nwrapper feels sensible to me. It is one of those interfaces that really\nmake you wish for generics in C.\n\nPatrick\n"},{"id":"546185","messageId":"ajk-a4a3KSJ2u7Ju@pks.im","threadId":"65771","inReplyTo":"20260622-toon-git-replay-drop-merges-v4-3-ff257f534319@iotcl.com","subject":"Re: [PATCH v4 3/3] replay: offer an option to linearize the commit topology","fromName":"Patrick Steinhardt","fromEmail":"ps@pks.im","sentAt":"2026-06-22T13:53:47Z","receivedAt":"2026-06-22T13:53:52Z","isPatch":true,"body":"On Mon, Jun 22, 2026 at 02:41:57PM +0200, Toon Claes wrote:\n> From: Johannes Schindelin \n> \n> One of the stated goals of git-replay(1) is to allow implementing the\n> git-rebase(1) functionality on the server side.\n> \n> The default mode of git-rebase(1) is to act as if `--no-rebase-merges`\n> was given. This mode drops merge commits instead of replaying them, and\n> linearizes the commit history into a sequence of the\n> regular (single-parent) commits.\n> \n> Add option `--linearize` to git-replay(1) to do the same.\n\ngit-rebase(1) essentially knows about three different modes:\n\n - \"--no-rebase-merges\", which is the default and maps to your\n \"--linearize\".\n\n - \"--rebase-merges\", which by default doesn't rebase cousins by using\n \"--ancestry-path\" internally.\n\n - \"--rebase-merges=rebase-cousins\", which doesn't pass the above\n option.\n\nSo it's not a simple boolean there, which makes me wonder whether we\nshould mirror the same interface so that all of git-rebase(1)'s modes\ncan be represented, as well.\n\n> diff --git a/replay.c b/replay.c\n> index 7921d7dba3..5539daff00 100644\n> --- a/replay.c\n> +++ b/replay.c\n> @@ -277,12 +277,16 @@ static struct commit *pick_regular_commit(struct repository *repo,\n> \t\t\t\t\t struct commit *onto,\n> \t\t\t\t\t struct merge_options *merge_opt,\n> \t\t\t\t\t struct merge_result *result,\n> +\t\t\t\t\t struct commit *replayed_base,\n> \t\t\t\t\t bool reverse,\n> \t\t\t\t\t enum replay_empty_commit_action empty)\n> {\n> -\tstruct commit *base, *replayed_base;\n> +\tstruct commit *base;\n> \tstruct tree *pickme_tree, *base_tree, *replayed_base_tree;\n> \n> +\tif (replayed_base && reverse)\n> +\t\tBUG(\"Linearizing commits is not supported when replaying in reverse\");\n\nNit: Error messages should typically start with a lower-case letter.\n\n> @@ -430,12 +435,25 @@ int replay_revisions(struct rev_info *revs,\n> \twhile ((commit = get_revision(revs))) {\n> \t\tconst struct name_decoration *decoration;\n> \n> -\t\tif (commit->parents && commit->parents->next)\n> -\t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n> +\t\tif (commit->parents && commit->parents->next) {\n> +\t\t\tif (!opts->linearize)\n> +\t\t\t\tdie(_(\"replaying merge commits is not supported yet!\"));\n> +\t\t\t/*\n> +\t\t\t * Drop the merge commit: do not pick it and leave\n> +\t\t\t * last_commit unchanged, so its children (and any ref\n> +\t\t\t * pointing at it) are reparented onto the previous\n> +\t\t\t * non-merge commit, which the ref-update loop below uses.\n> +\t\t\t */\n\nOne could add a hint here that tells the user to pass the option. But I\nguess that might be somewhat weird, as we cannot assume that we're\ncalled by git-replay(1) here.\n\nIn any case, this here is the core of the change where we stop dying in\ncase \"--linearize\" was passed, and instead we simply skip the commit\naltogether. Makes sense.\n\nThanks!\n\nPatrick\n"},{"id":"546188","messageId":"xmqq7bnq37jm.fsf@gitster.g","threadId":"65771","inReplyTo":"ajk-YQxLWfspNWIm@pks.im","subject":"Re: [PATCH v4 1/3] replay: refactor enum replay_mode into a bool","fromName":"Junio C Hamano","fromEmail":"gitster@pobox.com","sentAt":"2026-06-22T15:43:09Z","receivedAt":"2026-06-22T15:43:12Z","isPatch":true,"body":"Patrick Steinhardt writes:\n\n> On Mon, Jun 22, 2026 at 02:41:55PM +0200, Toon Claes wrote:\n>> In 2760ee4983 (replay: add --revert mode to reverse commit changes,\n>> 2026-03-26) the enum `replay_mode` was introduced. This has two possible\n>> values:\n>> \n>> - The value `REPLAY_MODE_REVERT` is used when option `--revert` is\n>> passed to git-replay(1). When using this value the commits are\n>> processed in reverse order and the inverse of the changes are\n>> applied.\n>> \n>> - The value `REPLAY_MODE_PICK` is used when either option `--onto` or\n>> `--advance` is used. In both cases the commits are processed in\n>> normal order, and the changes are applied as-is.\n>> \n>> Since there are only two possible values of this enum, simplify the code\n>> by converting the enum into a bool. This avoids adding code paths that\n>> check for invalid values of the enum, and shortens code where the value\n>> is checked with a ternary operator.\n>\n> That's fair, and the result is easier to write. But is it really easier\n> to read? And what if we ever have to create a third mode going forward?\n>\n> I'm generally no fan of booleans as parameters as they basically give\n> you no information at all at the callsite, except if you're lucky and\n> you already have an aptly-named variable available that you can pass.\n> Which seems to be the case here, but I'm still not sure whether this\n> change really improves the code.\n\nI tend to agree with you on both counts. The \"what happens when\nsomebody else wants a third choice?\" is a quesiton I would ask the\nfirst thing as the maintainer of a project.\n\nEven if the boolean parameter is so obviously named, the callsite\ncan only say \"true\" or \"false\", unlike some other popular languages\nthat lets you say\n\n\tmy_function(use_revert_mode=true, verbose=false);\n\nand you cannot tell what effect the author wanted out of that \"true\"\nif all you can write were\n\n\tmy_function(true, false);\n\nOf course, we could go ultra verbose, like\n\n\tmy_function(true, /* use_revert_mode */\n\t\t false, /* verbose */);\n\nbut then we are often better off writing:\n\n\tmy_function(REPLAY_MODE_REVERT, REPLAY_QUIET);\n\nThanks.\n"},{"id":"546342","messageId":"87tsqrycke.fsf@emacs.iotcl.com","threadId":"65771","inReplyTo":"xmqq7bnq37jm.fsf@gitster.g","subject":"Re: [PATCH v4 1/3] replay: refactor enum replay_mode into a bool","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-24T19:15:45Z","receivedAt":"2026-06-24T19:15:54Z","isPatch":true,"body":"Junio C Hamano writes:\n\n> Patrick Steinhardt writes:\n>\n>> That's fair, and the result is easier to write. But is it really easier\n>> to read?\n\nYou're bringing up a very valid point there, and not directly in what\nyou're saying, but how it makes me reconsider.\n\nSo we're comparing:\n\n pick_regular_commit(revs->repo, commit, replayed_commits,\n mode == REPLAY_MODE_REVERT ? last_commit : onto,\n &merge_opt, &result, mode, opts->empty);\n\nwith:\n\n pick_regular_commit(revs->repo, commit, replayed_commits,\n reverse ? last_commit : onto,\n &merge_opt, &result, reverse, opts->empty);\n\nYou can argue which of both is easier to read, but the problem isn't\nreally whether it's a bool or an enum, but the ternary operator in this\nlengthy function call is. That is the problem I was trying to solve, and\nconverting enum to bool isn't really the solution.\n\n>> And what if we ever have to create a third mode going forward?\n\nPersonally I find this weak argument. As far as I know we most of the\ntime do not write code in a way so \"it will be ready to add X in the\nfuture\". In my personal experience, I'm always wrong in predicting what\nmight be added in the future. Although I must say this case is\ndifferent, because we're not adding something new, no this commit was\ndumbing down something existing. So I'll revisit this commit in the next\niteration.\n\n>> I'm generally no fan of booleans as parameters as they basically give\n>> you no information at all at the callsite, except if you're lucky and\n>> you already have an aptly-named variable available that you can pass.\n>> Which seems to be the case here, but I'm still not sure whether this\n>> change really improves the code.\n\nThat's also a very valid argument, which I didn't take in mind.\n\n> I tend to agree with you on both counts. The \"what happens when\n> somebody else wants a third choice?\" is a quesiton I would ask the\n> first thing as the maintainer of a project.\n>\n> Even if the boolean parameter is so obviously named, the callsite\n> can only say \"true\" or \"false\", unlike some other popular languages\n> that lets you say\n>\n> \tmy_function(use_revert_mode=true, verbose=false);\n>\n> and you cannot tell what effect the author wanted out of that \"true\"\n> if all you can write were\n>\n> \tmy_function(true, false);\n>\n> Of course, we could go ultra verbose, like\n>\n> \tmy_function(true, /* use_revert_mode */\n> \t\t false, /* verbose */);\n>\n> but then we are often better off writing:\n>\n> \tmy_function(REPLAY_MODE_REVERT, REPLAY_QUIET);\n\nThanks for bringing in this illustrative example. Point made, I'll\nrevisit.\n\n-- \nCheers,\nToon\n"},{"id":"546441","messageId":"87qzltyiao.fsf@emacs.iotcl.com","threadId":"65771","inReplyTo":"ajk-a4a3KSJ2u7Ju@pks.im","subject":"Re: [PATCH v4 3/3] replay: offer an option to linearize the commit topology","fromName":"Toon Claes","fromEmail":"toon@iotcl.com","sentAt":"2026-06-26T05:36:31Z","receivedAt":"2026-06-26T05:36:42Z","isPatch":true,"body":"Patrick Steinhardt writes:\n\n> git-rebase(1) essentially knows about three different modes:\n>\n> - \"--no-rebase-merges\", which is the default and maps to your\n> \"--linearize\".\n>\n> - \"--rebase-merges\", which by default doesn't rebase cousins by using\n> \"--ancestry-path\" internally.\n>\n> - \"--rebase-merges=rebase-cousins\", which doesn't pass the above\n> option.\n>\n> So it's not a simple boolean there, which makes me wonder whether we\n> should mirror the same interface so that all of git-rebase(1)'s modes\n> can be represented, as well.\n\nThat's a valid question, although I don't know a good answer to that.\n\nBasically you're asking for what the command line options will look\nlike? Allow me to think out loud.\n\nIn this series I'm adding --linearize to git-replay(1). As mentioned, I\ndon't think it makes sense to add it to git-history(1) as well. Without\nthis option, the process aborts when it encounters a merge.\n\nDscho sent a patch series to properly replay (2-way) merges. I think\nthis should become the default for both git-replay(1) and\ngit-history(1).\n\nBut then, do we want to have an option that brings back the current\nbehavior of aborting at merges? Maybe with --no-merges?\n\nThen there's the option of rebasing cousins left. That's something that\nisn't covered by Dscho's series yet. Maybe --replay-cousins?\n\nTo reiterate what the final design could look like:\n\n * : replay merges preserving topology.\n * \"--linearize\": flattens merges (only git-replay(1)).\n * \"--no-merges\": dies when the process tries to replay a merge.\n * \"--replay-cousins\": does what --rebase-merges=rebase-cousins does.\n\nNow, all these options are (I think) mutually exclusive, so we could\nconsider an option \"--replay-merges=\", but personally I find\n\"--