Git development
 help / color / mirror / Atom feed
From: Taylor Blau <ttaylorr@openai.com>
To: git@vger.kernel.org
Cc: Junio C Hamano <gitster@pobox.com>, Jeff King <peff@peff.net>,
	Ted Nyman <tnyman@openai.com>, Elijah Newren <newren@github.com>
Subject: [PATCH 3/4] repack: retain cruft packs in MIDXs after incremental repacks
Date: Tue, 29 Sep 2026 20:28:53 -0500	[thread overview]
Message-ID: <1774fed77be11b37ce9eb4b7806f5f14539503fb.1790731662.git.me@ttaylorr.com> (raw)
In-Reply-To: <cover.1790731662.git.me@ttaylorr.com>

An incremental repack can write a commit and tree into a new pack while
leaving objects they reach in an existing cruft pack. For example, a
commit can make a previously unreachable blob reachable again. Since
'repack' will invoke 'pack-objects' with '--incremental', it will not
copy the blob out of its cruft pack.

When the 'repack.midxMustContainCruft' configuration is set to "false",
writing the first MIDX after such a repack may omit that cruft pack. The
new pack bypasses the `!names.nr` fallback, and there are no previous
MIDX packs for `midx_has_unknown_packs()` to check. Selecting the new
commit for bitmap coverage then fails because its reachable objects are
not all in the MIDX.

The omission dates all the way back to 5ee86c273bf (repack: exclude
cruft pack(s) from the MIDX where possible, 2025-06-23). It relies on
geometric repacking to copy once-cruft objects with
'--stdin-packs=follow'. However, an ordinary incremental repack makes no
such guarantee. Require the MIDX to include cruft packs in that case,
even when a new pack was written.

Exercise this with the existing fixture that makes a cruft commit
reachable again and adds a new (unpacked) commit on top, and ensure that
the incremental repack is able to successfully write a reachability
bitmap.

Signed-off-by: Taylor Blau <ttaylorr@openai.com>
---
 builtin/repack.c        |  6 ++++++
 t/t7704-repack-cruft.sh | 11 +++++++++++
 2 files changed, 17 insertions(+)

diff --git a/builtin/repack.c b/builtin/repack.c
index c4360382c1f..b7596d488da 100644
--- a/builtin/repack.c
+++ b/builtin/repack.c
@@ -539,6 +539,12 @@ int cmd_repack(int argc,
 			strvec_push(&cmd.args, "--stdin-packs=follow");
 		strvec_push(&cmd.args, "--unpacked");
 	} else {
+		/*
+		 * Incremental repacks do not copy already-packed objects,
+		 * so cruft packs may be required to form a reachability
+		 * closure for the MIDX.
+		 */
+		midx_must_contain_cruft = 1;
 		strvec_push(&cmd.args, "--unpacked");
 		strvec_push(&cmd.args, "--incremental");
 	}
diff --git a/t/t7704-repack-cruft.sh b/t/t7704-repack-cruft.sh
index b49f22878f7..f7f83e70ffe 100755
--- a/t/t7704-repack-cruft.sh
+++ b/t/t7704-repack-cruft.sh
@@ -787,6 +787,17 @@ test_expect_success 'geometric repack rescues descendants of loose trees' '
 	)
 '
 
+test_expect_success 'incremental repack includes cruft for MIDX bitmaps' '
+	setup_cruft_exclude_tests incremental-cruft &&
+	(
+		cd incremental-cruft &&
+
+		GIT_TEST_MULTI_PACK_INDEX=0 \
+		git repack -d --write-midx --write-bitmap-index &&
+		git rev-list --test-bitmap HEAD
+	)
+'
+
 test_expect_success 'repack --write-midx includes cruft when instructed' '
 	setup_cruft_exclude_tests exclude-cruft-when-instructed &&
 	(
-- 
2.56.0.4.gbee41d2fc68


  parent reply	other threads:[~2026-09-30  1:28 UTC|newest]

Thread overview: 41+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-09-30  1:28 [PATCH 0/4] repack: various corner cases for cruft-less MIDXs Taylor Blau
2026-09-30  1:28 ` [PATCH 1/4] pack-objects: introduce `stdin_packs_context` struct Taylor Blau
2026-09-30 17:42   ` Junio C Hamano
2026-10-01  3:13     ` Taylor Blau
2026-09-30  1:28 ` [PATCH 2/4] pack-objects: ensure tree/tag closure with '--stdin-packs=follow' Taylor Blau
2026-09-30 17:51   ` Junio C Hamano
2026-09-30 18:16   ` Derrick Stolee
2026-10-01  3:14     ` Taylor Blau
2026-10-01 23:22     ` Elijah Newren
2026-10-02  0:51       ` Taylor Blau
2026-10-02 23:02         ` Jeff King
2026-09-30 20:31   ` Jeff King
2026-10-01  3:18     ` Taylor Blau
2026-09-30  1:28 ` Taylor Blau [this message]
2026-09-30 20:45   ` [PATCH 3/4] repack: retain cruft packs in MIDXs after incremental repacks Jeff King
2026-10-01  3:21     ` Taylor Blau
2026-09-30  1:28 ` [PATCH 4/4] repack: retain cruft packs in MIDXs containing kept packs Taylor Blau
2026-09-30 20:53   ` Jeff King
2026-10-01  3:35     ` Taylor Blau
2026-09-30 20:55 ` [PATCH 0/4] repack: various corner cases for cruft-less MIDXs Jeff King
2026-10-01  3:37   ` Taylor Blau
2026-10-01  4:11 ` [PATCH v2 0/8] " Taylor Blau
2026-10-01  4:11   ` [PATCH v2 1/8] pack-objects: introduce `stdin_packs_context` struct Taylor Blau
2026-10-01  4:11   ` [PATCH v2 2/8] pack-objects: ensure tree/tag closure with '--stdin-packs=follow' Taylor Blau
2026-10-02 23:13     ` Jeff King
2026-10-03  0:55       ` Taylor Blau
2026-10-03  1:06         ` Jeff King
2026-10-01  4:11   ` [PATCH v2 3/8] repack: retain cruft packs in MIDXs after incremental repacks Taylor Blau
2026-10-01  4:11   ` [PATCH v2 4/8] repack: use a sorted list for explicitly kept packs Taylor Blau
2026-10-02 23:16     ` Jeff King
2026-10-01  4:11   ` [PATCH v2 5/8] repack: follow kept packs when omitting cruft from the MIDX Taylor Blau
2026-10-02 23:25     ` Jeff King
2026-10-03  0:50       ` Taylor Blau
2026-10-01  4:11   ` [PATCH v2 6/8] repack: track the preferred pack explicitly in MIDX write steps Taylor Blau
2026-10-02 23:28     ` Jeff King
2026-10-03  1:00       ` Taylor Blau
2026-10-03  1:07         ` Jeff King
2026-10-01  4:12   ` [PATCH v2 7/8] repack: defer allocating the append plan's write step Taylor Blau
2026-10-01  4:12   ` [PATCH v2 8/8] repack: include required packs in incremental MIDX writes Taylor Blau
2026-10-02 23:41     ` Jeff King
2026-10-03  1:01       ` Taylor Blau

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1774fed77be11b37ce9eb4b7806f5f14539503fb.1790731662.git.me@ttaylorr.com \
    --to=ttaylorr@openai.com \
    --cc=git@vger.kernel.org \
    --cc=gitster@pobox.com \
    --cc=newren@github.com \
    --cc=peff@peff.net \
    --cc=tnyman@openai.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox