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

[PATCH v8 0/2] rerere: wait for MERGE_RR.lock, but not in auto maintenance

From
Thomas Bachem via GitGitGadget <gitgitgadget@gmail.com>
Date
Oct 13, 2026, 12:07 UTC
Message-ID
<pull.2214.v8.git.1791893274.gitgitgadget@gmail.com>
In-Reply-To
<pull.2214.git.1788337897490.gitgitgadget@gmail.com>

A rebase dies at a conflict when a background "git rerere gc" holds MERGE_RR.lock at that moment. With the first patch, rerere waits for the lock instead of dying at once. With the second, the "git rerere gc" started by auto maintenance does nothing while the lock is held.

For v8 I took Junio's suggestions on patch 1 and Patrick's on patch 2. Changes since v7:

 * The rerere.lockTimeout entry no longer says that the command fails when
   the time is up (Junio).
 * The lock tests grep for the "Unable to create" message about
   MERGE_RR.lock (Junio).
 * The wait for the background unlocker is now in test_when_finished
   (Junio).
 * The rebase --abort test aborts its rebase in test_when_finished, so a
   failed check cannot break later tests.
 * The test "merge fails once rerere.lockTimeout is up" never waits, so it
   is now called "merge fails on a lock it cannot take".
 * In patch 2, rerere_gc() takes RERERE_NOWAIT again instead of its own enum
   rerere_gc_flags (Patrick).
 * The comment about waiting in setup_rerere() now says "by default", and as
   in patch 1 it sits directly above the call that takes the lock.
 * A comment in t7900 says why the tests with a held lock create rr-cache.
Thomas Bachem (2):
  rerere: wait for MERGE_RR.lock before giving up
  rerere: add "gc --skip-locked" for auto maintenance
 Documentation/config/rerere.adoc |  9 ++++
 builtin/gc.c                     |  4 +-
 builtin/rerere.c                 | 10 +++-
 rerere.c                         | 37 ++++++++++++---
 rerere.h                         |  4 +-
 t/t4200-rerere.sh                | 79 ++++++++++++++++++++++++++++++++
 t/t7900-maintenance.sh           | 26 ++++++++++-
 7 files changed, 157 insertions(+), 12 deletions(-)
base-commit: 34f06850c16c7f7ac822b1adc71354f11b0f2ca3
Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-2214%2Fthomasbachem%2Frerere-gc-lock-v8
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-2214/thomasbachem/rerere-gc-lock-v8
Pull-Request: https://github.com/gitgitgadget/git/pull/2214
Range-diff vs v7:
 1:  3dc3d02f12 ! 1:  06f6c77439 rerere: wait for MERGE_RR.lock before giving up
     @@ Metadata
       ## Commit message ##
          rerere: wait for MERGE_RR.lock before giving up
      
     -    setup_rerere() takes MERGE_RR.lock with LOCK_DIE_ON_ERROR, so of two
     -    processes that want it at the same time the second one dies. That
     +    setup_rerere() takes MERGE_RR.lock with LOCK_DIE_ON_ERROR, so when two
     +    processes want it at the same time, the second one dies. That
          used to be rare. Since 452b12c2e0 (builtin/maintenance: use
          "geometric" strategy by default, 2026-02-24) the auto maintenance
          after a commit runs "git rerere gc" whenever rr-cache has enough
          stale entries, and the gc holds the lock while it prunes.
      
     -    A rebase whose next pick conflicts while that happens dies inside
     -    repo_rerere(), before the sequencer has written the state that
     +    If a rebase's next pick conflicts while the gc holds the lock, the
     +    rebase dies inside repo_rerere(), before the sequencer has written what
          "git rebase --continue" needs. Every later "git rebase --continue"
          then fails with "you have staged changes in your working tree".
      
     @@ Documentation/config/rerere.adoc: rerere.enabled::
      +	lock when another process holds it, typically a background
      +	`git rerere gc`.  Value 0 means not to wait at all; -1 means
      +	to wait indefinitely.  Default is 1000 (i.e., wait for 1
     -+	second).  When the time is up, the command fails as it does
     -+	for any other lock it cannot take.
     ++	second).
      
       ## rerere.c ##
      @@ rerere.c: static int rerere_enabled = -1;
     @@ t/t4200-rerere.sh: test_expect_success 'old records rest in peace' '
      +test_expect_success 'a held lock is waited out within rerere.lockTimeout' '
      +	git reset --hard &&
      +	rm -rf $rr &&
     -+	test_when_finished "rm -f .git/MERGE_RR.lock" &&
     ++	test_when_finished "wait && rm -f .git/MERGE_RR.lock" &&
      +	>.git/MERGE_RR.lock &&
      +	{
      +		( sleep 1 && rm -f .git/MERGE_RR.lock ) &
      +	} &&
      +	test_must_fail git -c rerere.lockTimeout=5000 merge first 2>err &&
     -+	wait &&
     -+	test_grep ! "MERGE_RR" err &&
     ++	test_grep ! "Unable to create.*MERGE_RR\.lock" err &&
      +	test_grep "^=======\$" $rr/preimage
      +'
      +
     -+test_expect_success 'merge fails once rerere.lockTimeout is up' '
     ++test_expect_success 'merge fails on a lock it cannot take' '
      +	git reset --hard &&
      +	rm -rf $rr &&
      +	test_when_finished "rm -f .git/MERGE_RR.lock" &&
      +	>.git/MERGE_RR.lock &&
      +	test_must_fail git -c rerere.lockTimeout=0 merge first 2>err &&
     -+	test_grep "Unable to create" err &&
     ++	test_grep "Unable to create.*MERGE_RR\.lock" err &&
      +	test_grep "^=======\$" a1 &&
      +	test_path_is_missing $rr/preimage
      +'
     @@ t/t4200-rerere.sh: test_expect_success 'old records rest in peace' '
      +	test_when_finished "rm -f .git/MERGE_RR.lock" &&
      +	>.git/MERGE_RR.lock &&
      +	test_must_fail git -c rerere.lockTimeout=0 rerere 2>err &&
     -+	test_grep "Unable to create" err &&
     ++	test_grep "Unable to create.*MERGE_RR\.lock" err &&
      +	test_must_fail git -c rerere.lockTimeout=0 rerere forget a1 2>err &&
     -+	test_grep "Unable to create" err &&
     ++	test_grep "Unable to create.*MERGE_RR\.lock" err &&
      +	test_must_fail git -c rerere.lockTimeout=0 rerere clear 2>err &&
     -+	test_grep "Unable to create" err &&
     ++	test_grep "Unable to create.*MERGE_RR\.lock" err &&
      +	test_must_fail git -c rerere.lockTimeout=0 rerere gc 2>err &&
     -+	test_grep "Unable to create" err
     ++	test_grep "Unable to create.*MERGE_RR\.lock" err
      +'
      +
      +test_expect_success 'rebase --abort fails on a lock it cannot take' '
     @@ t/t4200-rerere.sh: test_expect_success 'old records rest in peace' '
      +	git checkout -b lock-held-abort third &&
      +	test_when_finished "git checkout third && git branch -D lock-held-abort" &&
      +	test_must_fail git rebase first &&
     ++	test_when_finished "test_might_fail git rebase --abort" &&
      +	test_when_finished "rm -f .git/MERGE_RR.lock" &&
      +	>.git/MERGE_RR.lock &&
      +	test_must_fail git -c rerere.lockTimeout=0 rebase --abort 2>err &&
     -+	test_grep "Unable to create" err &&
     ++	test_grep "Unable to create.*MERGE_RR\.lock" err &&
      +	test_path_is_dir .git/rebase-merge &&
      +	rm .git/MERGE_RR.lock &&
      +	git rebase --abort &&
 2:  0002881196 ! 2:  4202361858 rerere: add "gc --skip-locked" for auto maintenance
     @@ Commit message
      
       ## Documentation/config/rerere.adoc ##
      @@ Documentation/config/rerere.adoc: rerere.lockTimeout::
     + 	lock when another process holds it, typically a background
       	`git rerere gc`.  Value 0 means not to wait at all; -1 means
       	to wait indefinitely.  Default is 1000 (i.e., wait for 1
     - 	second).  When the time is up, the command fails as it does
     --	for any other lock it cannot take.
     -+	for any other lock it cannot take.  A `git rerere gc` run by
     -+	`git maintenance run --auto` or `git gc --auto` does not wait
     -+	and does nothing while the lock is held.
     +-	second).
     ++	second).  A `git rerere gc` run by `git maintenance run --auto`
     ++	or `git gc --auto` does not wait and does nothing while the
     ++	lock is held.
      
       ## builtin/gc.c ##
      @@ builtin/gc.c: out:
     @@ builtin/rerere.c: int cmd_rerere(int argc,
       	} else if (!strcmp(argv[0], "gc"))
      -		rerere_gc(the_repository, &merge_rr);
      +		rerere_gc(the_repository, &merge_rr,
     -+			  skip_locked ? RERERE_GC_NOWAIT : 0);
     ++			  skip_locked ? RERERE_NOWAIT : 0);
       	else if (!strcmp(argv[0], "status")) {
       		if (setup_rerere(the_repository, &merge_rr,
       				 flags | RERERE_READONLY) < 0)
     @@ rerere.c: int setup_rerere(struct repository *r, struct string_list *merge_rr, i
      +		int lock_flags = LOCK_DIE_ON_ERROR;
      +		int timeout_ms = rerere_lock_timeout_ms;
      +
     - 		/*
     - 		 * Another process may hold the lock for a while, e.g.
     - 		 * "git rerere gc" while it prunes rr-cache, so wait for
     - 		 * it instead of dying right away.
     - 		 */
      +		if (flags & RERERE_NOWAIT) {
      +			lock_flags = 0;
      +			timeout_ms = 0;
      +		}
     ++
     + 		/*
     + 		 * Another process may hold the lock for a while, e.g.
     +-		 * "git rerere gc" while it prunes rr-cache, so wait for
     +-		 * it instead of dying right away.
     ++		 * "git rerere gc" while it prunes rr-cache, so by default
     ++		 * wait for it instead of dying right away.
     + 		 */
       		fd = repo_hold_lock_file_for_update_timeout(r, &write_lock,
       							    git_path_merge_rr(r),
      -							    LOCK_DIE_ON_ERROR,
     @@ rerere.c: out:
       }
       
      -void rerere_gc(struct repository *r, struct string_list *rr)
     -+void rerere_gc(struct repository *r, struct string_list *rr,
     -+	       enum rerere_gc_flags flags)
     ++void rerere_gc(struct repository *r, struct string_list *rr, int flags)
       {
       	struct string_list to_remove = STRING_LIST_INIT_DUP;
       	DIR *dir;
     @@ rerere.c: void rerere_gc(struct repository *r, struct string_list *rr)
       	struct strbuf buf = STRBUF_INIT;
       
      -	if (setup_rerere(r, rr, 0) < 0)
     -+	if (setup_rerere(r, rr,
     -+			 (flags & RERERE_GC_NOWAIT) ? RERERE_NOWAIT : 0) < 0)
     ++	if (setup_rerere(r, rr, flags) < 0)
       		return;
       
       	rerere_gc_cutoffs(r, &cutoff_resolve, &cutoff_noresolve);
     @@ rerere.h: const char *rerere_path(struct strbuf *buf, const struct rerere_id *,
       int rerere_remaining(struct repository *, struct string_list *);
       void rerere_clear(struct repository *, struct string_list *);
      -void rerere_gc(struct repository *, struct string_list *);
     -+
     -+enum rerere_gc_flags {
     -+	/* Skip the operation in case the MERGE_RR.lock is already taken. */
     -+	RERERE_GC_NOWAIT = (1 << 0),
     -+};
     -+
     -+void rerere_gc(struct repository *, struct string_list *,
     -+	       enum rerere_gc_flags flags);
     ++void rerere_gc(struct repository *, struct string_list *, int);
       
       /*
        * Check whether garbage collection for rerere entries is needed, which is
     @@ t/t7900-maintenance.sh: test_expect_success 'rerere-gc task with --auto honors m
       
      +test_expect_success 'rerere-gc task with --auto succeeds while MERGE_RR is locked' '
      +	test_when_finished "rm -rf .git/rr-cache .git/MERGE_RR.lock" &&
     ++	# Without "rr-cache", rerere is disabled and never takes the lock.
      +	mkdir .git/rr-cache &&
      +	>.git/MERGE_RR.lock &&
      +	test_expect_rerere_gc git -c maintenance.rerere-gc.auto=-1 maintenance run --auto --task=rerere-gc
     @@ t/t7900-maintenance.sh: test_expect_success 'rerere-gc task with --auto honors m
      +	mkdir .git/rr-cache &&
      +	>.git/MERGE_RR.lock &&
      +	test_must_fail git -c rerere.lockTimeout=0 maintenance run --task=rerere-gc 2>err &&
     -+	test_grep "Unable to create" err
     ++	test_grep "Unable to create.*MERGE_RR\.lock" err
      +'
      +
       test_expect_success '--auto and --schedule incompatible' '
-- 
gitgitgadget
Previous: Patrick SteinhardtNext: Thomas Bachem via GitGitGadget
Message 60 of 66 in “rerere: keep a background gc from killing a rebase”
  1. rerere: keep a background gc from killing a rebaseThomas Bachem via GitGitGadget, Sep 2, 2026
  2. Phillip WoodSep 2, 2026
  3. Thomas BachemSep 2, 2026
  4. Phillip WoodSep 3, 2026
  5. Patrick SteinhardtSep 3, 2026
  6. Thomas BachemSep 3, 2026
  7. Patrick SteinhardtSep 3, 2026
  8. Thomas BachemSep 3, 2026
  9. Phillip WoodSep 3, 2026
  10. rerere: keep a background gc from killing a rebaseThomas Bachem via GitGitGadget, Sep 4, 2026
  11. Phillip WoodSep 4, 2026
  12. Thomas BachemSep 4, 2026
  13. Phillip WoodSep 7, 2026
  14. Junio C HamanoSep 4, 2026
  15. Thomas BachemSep 4, 2026
  16. rerere: keep a background gc from killing a rebaseThomas Bachem via GitGitGadget, Sep 4, 2026
  17. Junio C HamanoSep 4, 2026
  18. Thomas BachemSep 5, 2026
  19. Junio C HamanoSep 5, 2026
  20. Thomas BachemSep 6, 2026
  21. Patrick SteinhardtSep 7, 2026
  22. 0/2 rerere: wait for MERGE_RR.lock, and go on at a conflictThomas Bachem via GitGitGadget, Sep 14, 2026
  23. 1/2 rerere: wait for MERGE_RR.lock, and let the gc skip itThomas Bachem via GitGitGadget, Sep 14, 2026
  24. Patrick SteinhardtSep 28, 2026
  25. 2/2 rerere: go on at a conflict when the lock stays busyThomas Bachem via GitGitGadget, Sep 14, 2026
  26. Patrick SteinhardtSep 28, 2026
  27. 0/3 rerere: wait for MERGE_RR.lock, and go on at a conflictThomas Bachem via GitGitGadget, Sep 28, 2026
  28. 1/3 rerere: wait for MERGE_RR.lock before giving upThomas Bachem via GitGitGadget, Sep 28, 2026
  29. 2/3 rerere: add "gc --auto" that skips a held lockThomas Bachem via GitGitGadget, Sep 28, 2026
  30. Patrick SteinhardtSep 30, 2026
  31. Thomas BachemOct 1, 2026
  32. Patrick SteinhardtOct 1, 2026
  33. 3/3 rerere: go on at a conflict when the lock stays busyThomas Bachem via GitGitGadget, Sep 28, 2026
  34. 0/3 rerere: wait for MERGE_RR.lock, and go on at a conflictThomas Bachem via GitGitGadget, Oct 2, 2026
  35. 1/3 rerere: wait for MERGE_RR.lock before giving upThomas Bachem via GitGitGadget, Oct 2, 2026
  36. 2/3 rerere: add "gc --skip-locked" for auto maintenanceThomas Bachem via GitGitGadget, Oct 2, 2026
  37. Patrick SteinhardtOct 9, 2026
  38. Thomas BachemOct 10, 2026
  39. 3/3 rerere: go on at a conflict when the lock stays busyThomas Bachem via GitGitGadget, Oct 2, 2026
  40. Patrick SteinhardtOct 9, 2026
  41. Thomas BachemOct 10, 2026
  42. 0/2 rerere: wait for MERGE_RR.lock, but not in auto maintenanceThomas Bachem via GitGitGadget, Oct 10, 2026
  43. 1/2 rerere: wait for MERGE_RR.lock before giving upThomas Bachem via GitGitGadget, Oct 10, 2026
  44. Junio C HamanoOct 11, 2026
  45. Thomas BachemOct 11, 2026
  46. Junio C HamanoOct 12, 2026
  47. Okhuomon AjayiOct 14, 2026
  48. credential-store: add lockTimeout config optionOkhuomon Ajayi, Oct 14, 2026
  49. Junio C HamanoOct 15, 2026
  50. Okhuomon AjayiOct 16, 2026
  51. credential-store: add lockTimeout config optionOkhuomon Ajayi, Oct 16, 2026
  52. Okhuomon AjayiOct 20, 2026
  53. Okhuomon AjayiNov 12, 2026
  54. Junio C HamanoNov 12, 2026
  55. Okhuomon AjayiNov 15, 2026
  56. credential-store: add lockTimeout config optionOkhuomon Ajayi, Nov 16, 2026
  57. Junio C HamanoNov 16, 2026
  58. 2/2 rerere: add "gc --skip-locked" for auto maintenanceThomas Bachem via GitGitGadget, Oct 10, 2026
  59. Patrick SteinhardtOct 13, 2026
  60. 0/2 rerere: wait for MERGE_RR.lock, but not in auto maintenanceThomas Bachem via GitGitGadget, Oct 13, 2026
  61. 1/2 rerere: wait for MERGE_RR.lock before giving upThomas Bachem via GitGitGadget, Oct 13, 2026
  62. 2/2 rerere: add "gc --skip-locked" for auto maintenanceThomas Bachem via GitGitGadget, Oct 13, 2026
  63. Patrick SteinhardtOct 13, 2026
  64. Junio C HamanoOct 13, 2026
  65. Patrick SteinhardtOct 14, 2026
  66. Thomas BachemOct 14, 2026

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

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