Git development
 help / color / mirror / Atom feed
From: Maciej Ciemborowicz <maciej.ciemborowicz@gmail.com>
To: git@vger.kernel.org
Cc: "Karthik Nayak" <karthik.188@gmail.com>,
	"Junio C Hamano" <gitster@pobox.com>,
	"Patrick Steinhardt" <ps@pks.im>,
	"Phil Hord" <phil.hord@gmail.com>,
	"Elijah Newren" <newren@gmail.com>,
	"Ævar Arnfjörð Bjarmason" <avarab@gmail.com>,
	"D . Ben Knoble" <ben.knoble@gmail.com>,
	"Maciej Ciemborowicz" <maciej.ciemborowicz@gmail.com>
Subject: [PATCH v6 0/1] refs: report old values to transaction hooks
Date: Fri, 25 Sep 2026 00:33:09 +0200	[thread overview]
Message-ID: <cover.1790269745.git.maciej.ciemborowicz@gmail.com> (raw)
In-Reply-To: <cover.1790196627.git.maciej.ciemborowicz@gmail.com>

This follows up on the reference-transaction bug report at [1].

The reference-transaction hook currently reports an all-zero old object ID
when a caller queues an unconditional update. As a result, batched branch,
tag, and remote-ref deletions report zero for both the old and new values.

Earlier versions changed refs_delete_refs() and its callers to pass expected
old OIDs. As Patrick pointed out, that makes previously unconditional
deletions conditional and adds complexity at the wrong layer.

Resolve the old value in the common transaction hook layer instead. Keep it
separate from the caller-supplied old_oid and old_target so it cannot
constrain the update. Resolve it before the "preparing" hook and refresh it
after the backend locks the references, allowing later phases to report the
value protected by those locks. Do this work only when a
reference-transaction hook exists.

Changes since v5:

 * Replace the three-patch caller-specific approach with one transaction
   layer change.
 * Leave refs_delete_refs(), ref_transaction_delete(), and all command call
   sites unchanged.
 * Preserve unconditional deletion semantics and test a concurrent update
   from the "preparing" hook.
 * Report observed old values for all unconditional transactions, including
   symbolic ref targets.
 * Document that the unlocked value reported in "preparing" may differ from
   later phases if the reference changes before it is locked.

The full test suite passes (1060 files, 34660 tests). The focused
reference-transaction tests also pass with SHA-1 and SHA-256 using both the
files and reftable backends.

[1] https://lore.kernel.org/git/CACQ=SRGTTdQ+dHXhN6F52dBv5KxZBRfk_Em2fvmEmGJDoB6oTg@mail.gmail.com/

Maciej Ciemborowicz (1):
  refs: report old values to transaction hooks

 Documentation/githooks.adoc      | 17 +++++----
 refs.c                           | 54 ++++++++++++++++++++++++---
 refs/refs-internal.h             |  8 ++++
 t/t1416-ref-transaction-hooks.sh | 64 +++++++++++++++++++++++++++++++-
 4 files changed, 128 insertions(+), 15 deletions(-)

Range-diff against v5:
1:  9b76cc2c40 < -:  ---------- refs: allow callers to supply old OIDs for batch deletion
2:  6a8401c448 < -:  ---------- branch, tag: retain old OIDs in batched deletions
3:  541da44c37 < -:  ---------- fetch, remote: retain old OIDs when pruning refs
-:  ---------- > 1:  2af3eeadd1 refs: report old values to transaction hooks
-- 
2.39.3 (Apple Git-146)

  parent reply	other threads:[~2026-09-24 22:33 UTC|newest]

Thread overview: 52+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-09-19 13:34 [BUG] reference-transaction reports zero OIDs for branch and tag deletion Maciej Ciemborowicz
2026-09-19 14:59 ` D. Ben Knoble
2026-09-19 15:42   ` Maciej Ciemborowicz
2026-09-19 20:11     ` [PATCH 0/3] refs: report old OIDs for batched deletions Maciej Ciemborowicz
2026-09-19 20:11       ` [PATCH 1/3] refs: allow callers to supply old OIDs for batch deletion Maciej Ciemborowicz
2026-09-19 20:41         ` Karthik Nayak
2026-09-20 10:38           ` Maciej Ciemborowicz
2026-09-20 10:54           ` [PATCH v2 0/3] refs: report old OIDs for batched deletions Maciej Ciemborowicz
2026-09-20 10:54             ` [PATCH v2 1/3] refs: allow callers to supply old OIDs for batch deletion Maciej Ciemborowicz
2026-09-21 13:12               ` Karthik Nayak
2026-09-21 23:55               ` Junio C Hamano
2026-09-20 10:54             ` [PATCH v2 2/3] branch, tag: retain old OIDs in batched deletions Maciej Ciemborowicz
2026-09-21 13:19               ` Karthik Nayak
2026-09-20 10:54             ` [PATCH v2 3/3] fetch, remote: retain old OIDs when pruning refs Maciej Ciemborowicz
2026-09-21 13:56               ` Karthik Nayak
2026-09-21 13:57             ` [PATCH v2 0/3] refs: report old OIDs for batched deletions Karthik Nayak
2026-09-21 20:01               ` Maciej Ciemborowicz
2026-09-22 12:26             ` [PATCH v3 " Maciej Ciemborowicz
2026-09-22 12:26               ` [PATCH v3 1/3] refs: allow callers to supply old OIDs for batch deletion Maciej Ciemborowicz
2026-09-22 18:55                 ` Junio C Hamano
2026-09-22 19:21                   ` Maciej Ciemborowicz
2026-09-22 23:31                     ` Junio C Hamano
2026-09-22 12:26               ` [PATCH v3 2/3] branch, tag: retain old OIDs in batched deletions Maciej Ciemborowicz
2026-09-22 12:26               ` [PATCH v3 3/3] fetch, remote: retain old OIDs when pruning refs Maciej Ciemborowicz
2026-09-22 19:16                 ` Junio C Hamano
2026-09-22 22:29               ` [PATCH v4 0/3] refs: report old OIDs for batched deletions Maciej Ciemborowicz
2026-09-22 22:31                 ` [PATCH v4 1/3] refs: allow callers to supply old OIDs for batch deletion Maciej Ciemborowicz
2026-09-22 22:31                 ` [PATCH v4 2/3] branch, tag: retain old OIDs in batched deletions Maciej Ciemborowicz
2026-09-22 22:31                 ` [PATCH v4 3/3] fetch, remote: retain old OIDs when pruning refs Maciej Ciemborowicz
2026-09-23 20:03                   ` Junio C Hamano
2026-09-23 21:02                     ` Maciej Ciemborowicz
2026-09-23 21:04                 ` [PATCH v5 0/3] refs: report old OIDs for batched deletions Maciej Ciemborowicz
2026-09-23 21:04                   ` [PATCH v5 1/3] refs: allow callers to supply old OIDs for batch deletion Maciej Ciemborowicz
2026-09-24 10:04                     ` Karthik Nayak
2026-09-24 16:34                       ` Junio C Hamano
2026-09-24 19:43                       ` Maciej Ciemborowicz
2026-09-24 11:07                     ` Patrick Steinhardt
2026-09-24 16:45                       ` Junio C Hamano
2026-09-24 20:13                         ` Maciej Ciemborowicz
2026-09-28  6:44                           ` Patrick Steinhardt
2026-09-28  6:43                         ` Patrick Steinhardt
2026-09-24 19:56                       ` Maciej Ciemborowicz
2026-09-23 21:04                   ` [PATCH v5 2/3] branch, tag: retain old OIDs in batched deletions Maciej Ciemborowicz
2026-09-24 11:08                     ` Patrick Steinhardt
2026-09-23 21:04                   ` [PATCH v5 3/3] fetch, remote: retain old OIDs when pruning refs Maciej Ciemborowicz
2026-09-23 21:55                   ` [PATCH v5 0/3] refs: report old OIDs for batched deletions Junio C Hamano
2026-09-24 22:33                   ` Maciej Ciemborowicz [this message]
2026-09-24 22:33                     ` [PATCH v6 1/1] refs: report old values to transaction hooks Maciej Ciemborowicz
2026-09-30  3:11                       ` Maciej Ciemborowicz
2026-10-01 17:37                         ` Maciej Ciemborowicz
2026-09-19 20:11       ` [PATCH 2/3] branch, tag: retain old OIDs in batched deletions Maciej Ciemborowicz
2026-09-19 20:11       ` [PATCH 3/3] fetch, remote: retain old OIDs when pruning refs Maciej Ciemborowicz

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=cover.1790269745.git.maciej.ciemborowicz@gmail.com \
    --to=maciej.ciemborowicz@gmail.com \
    --cc=avarab@gmail.com \
    --cc=ben.knoble@gmail.com \
    --cc=git@vger.kernel.org \
    --cc=gitster@pobox.com \
    --cc=karthik.188@gmail.com \
    --cc=newren@gmail.com \
    --cc=phil.hord@gmail.com \
    --cc=ps@pks.im \
    /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