[PATCH] mbox: add option to save all series revisions

[Andrea Righi <arighi@nvidia.com>]

There is currently no way to save a patch series together with all its
previous revisions in one mbox. Sometimes, having the full history in a
single mailbox can be helpful to see how the series evolved.

Add -a/--all-revisions to b4 mbox to find older and newer revisions and
save their discussion threads in revision order. Reuse the revision
search from review tracking and document the 12-month limit on backward
discovery.

Assisted-by: LLM [analysis, codegen, tests]
Signed-off-by: Andrea Righi <arighi@nvidia.com>
---
 docs/maintainer/mbox.rst | 12 +++++++
 src/b4/command.py        |  6 ++++
 src/b4/man/b4.1          |  4 +++
 src/b4/man/b4.1.rst      |  4 +++
 src/b4/mbox.py           | 35 +++++++++++++++++--
 src/tests/test_mbox.py   | 73 ++++++++++++++++++++++++++++++++++++++++
 6 files changed, 132 insertions(+), 2 deletions(-)

diff --git a/docs/maintainer/mbox.rst b/docs/maintainer/mbox.rst
index c2c5e4b..9289810 100644
--- a/docs/maintainer/mbox.rst
+++ b/docs/maintainer/mbox.rst
@@ -101,6 +101,18 @@ Option flags
   later revision is also available and will add these results to the
   retrieved thread.
 
+``-a, --all-revisions``
+  Check for both older and newer revisions of the selected patch series
+  and save the messages grouped by revision, oldest first. Messages
+  within each revision are ordered by date. For example::
+
+      b4 mbox -a -n review.mbox <msgid>
+
+  Revision discovery depends on the available public-inbox archive and
+  may not find revisions whose subject and author changed without a
+  matching ``change-id``. The older-revision search looks back up to
+  12 months from the newest revision.
+
 ``-n WANTNAME, --mbox-name WANTNAME``
   By default, the resulting mailbox file will use the message-id as the
   basis for its filename. This option lets you override this behaviour.
diff --git a/src/b4/command.py b/src/b4/command.py
index c65a5e4..0ef4c6c 100644
--- a/src/b4/command.py
+++ b/src/b4/command.py
@@ -355,6 +355,12 @@ def setup_parser() -> argparse.ArgumentParser:
     # b4 mbox
     sp_mbox = subparsers.add_parser('mbox', help='Download a thread as an mbox file')
     cmd_mbox_common_opts(sp_mbox)
+    sp_mbox.add_argument(
+        '-a',
+        '--all-revisions',
+        action='store_true',
+        help='Find older and newer revisions and save them in revision order',
+    )
     sp_mbox.add_argument(
         '-f',
         '--filter-dupes',
diff --git a/src/b4/man/b4.1 b/src/b4/man/b4.1
index b93e42c..22ad531 100644
--- a/src/b4/man/b4.1
+++ b/src/b4/man/b4.1
@@ -113,6 +113,10 @@ folder.
 Stitch multiple unthreaded messages into a single series and save as
 an mbox file. Useful when a submitter sends patches without proper
 threading.
+.TP
+.B \fBb4 mbox \-a \-n review.mbox [msgid]\fP
+Find older and newer revisions and save their discussion threads in
+revision order to review.mbox.
 .UNINDENT
 .SS b4 am
 .sp
diff --git a/src/b4/man/b4.1.rst b/src/b4/man/b4.1.rst
index 429d3ad..ded7d6e 100644
--- a/src/b4/man/b4.1.rst
+++ b/src/b4/man/b4.1.rst
@@ -83,6 +83,10 @@ For options and their descriptions, see ``b4 mbox --help``.
   an mbox file. Useful when a submitter sends patches without proper
   threading.
 
+``b4 mbox -a -n review.mbox [msgid]``
+  Find older and newer revisions and save their discussion threads in
+  revision order to review.mbox.
+
 
 b4 am
 ~~~~~
diff --git a/src/b4/mbox.py b/src/b4/mbox.py
index e3c1df2..05da1ed 100644
--- a/src/b4/mbox.py
+++ b/src/b4/mbox.py
@@ -1053,6 +1053,16 @@ def _run_shazam_merge(
     sys.exit(0)
 
 
+def revision_sort_key(msg: EmailMessage) -> Tuple[int, int, int]:
+    lsub = b4.LoreSubject(msg.get('Subject', ''))
+    parsed_date = email.utils.parsedate_tz(msg.get('Date', ''))
+    if parsed_date is None:
+        timestamp = 0
+    else:
+        timestamp = email.utils.mktime_tz(parsed_date)
+    return lsub.revision, timestamp, lsub.counter
+
+
 def main(cmdargs: argparse.Namespace) -> None:
     # We force some settings
     if cmdargs.subcmd == 'shazam':
@@ -1070,7 +1080,8 @@ def main(cmdargs: argparse.Namespace) -> None:
     else:
         cmdargs.mergebase = False
 
-    if cmdargs.checknewer:
+    all_revisions = cmdargs.subcmd == 'mbox' and cmdargs.all_revisions
+    if cmdargs.checknewer or all_revisions:
         # Force nocache mode
         cmdargs.nocache = True
 
@@ -1086,9 +1097,29 @@ def main(cmdargs: argparse.Namespace) -> None:
     if not msgs or not msgid:
         sys.exit(1)
 
-    if len(msgs) and cmdargs.checknewer and b4.can_network:
+    if (cmdargs.checknewer or all_revisions) and b4.can_network:
         msgs = get_extra_series(msgs, direction=1, nocache=cmdargs.nocache)
 
+    if all_revisions:
+        if b4.can_network:
+            subjects = (b4.LoreSubject(msg.get('Subject', '')) for msg in msgs)
+            latest = max(
+                (
+                    sub.revision
+                    for sub in subjects
+                    if not sub.reply and sub.counter <= 1
+                ),
+                default=1,
+            )
+            if latest > 1:
+                msgs = get_extra_series(
+                    msgs,
+                    direction=-1,
+                    wantvers=list(range(1, latest)),
+                    nocache=cmdargs.nocache,
+                )
+        msgs.sort(key=revision_sort_key)
+
     if cmdargs.subcmd in ('am', 'shazam'):
         make_am(msgs, cmdargs, msgid)
         return
diff --git a/src/tests/test_mbox.py b/src/tests/test_mbox.py
index e1df42b..8401e22 100644
--- a/src/tests/test_mbox.py
+++ b/src/tests/test_mbox.py
@@ -1,7 +1,9 @@
 import io
+import mailbox
 import os
 import sys
 from email.message import EmailMessage
+from pathlib import Path
 from typing import Any, Dict, List
 from unittest.mock import patch as mock_patch
 
@@ -127,6 +129,77 @@ def _make_msg(
     return msg
 
 
+def test_mbox_all_revisions_from_middle_version(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    monkeypatch.setattr(b4, 'can_network', True)
+    author = 'Author <author@example.com>'
+    v1 = _make_msg(
+        '[PATCH] foo: fix bar',
+        author,
+        'Mon, 21 Sep 2026 10:00:00 +0000',
+        msgid='<v1@example.com>',
+    )
+    v1_reply = _make_msg(
+        'Re: [PATCH] foo: fix bar',
+        author,
+        'Mon, 21 Sep 2026 11:00:00 +0000',
+        msgid='<v1-reply@example.com>',
+    )
+    v2 = _make_msg(
+        '[PATCH v2] foo: fix bar',
+        author,
+        'Tue, 22 Sep 2026 10:00:00 +0000',
+        msgid='<v2@example.com>',
+    )
+    v3 = _make_msg(
+        '[PATCH v3] foo: fix bar',
+        author,
+        'Wed, 23 Sep 2026 10:00:00 +0000',
+        msgid='<v3@example.com>',
+    )
+    v3_reply = _make_msg(
+        'Re: [PATCH v3] foo: fix bar',
+        author,
+        'Wed, 23 Sep 2026 11:00:00 +0000',
+        msgid='<v3-reply@example.com>',
+    )
+
+    parser = b4.command.setup_parser()
+    cmdargs = parser.parse_args(
+        [
+            '--no-stdin',
+            'mbox',
+            '-a',
+            '-o',
+            str(tmp_path),
+            '-n',
+            'all.mbox',
+            'v2@example.com',
+        ]
+    )
+    with (
+        mock_patch('b4.retrieve_messages', return_value=('v2@example.com', [v2])),
+        mock_patch(
+            'b4.get_pi_search_results',
+            side_effect=[
+                [v3_reply, v3],
+                [v1_reply, v1],
+            ],
+        ),
+    ):
+        b4.mbox.main(cmdargs)
+
+    output = mailbox.mbox(tmp_path / 'all.mbox')
+    assert [msg['Message-Id'] for msg in output] == [
+        '<v1@example.com>',
+        '<v1-reply@example.com>',
+        '<v2@example.com>',
+        '<v3@example.com>',
+        '<v3-reply@example.com>',
+    ]
+
+
 def test_get_extra_series_rejects_prerequisite_change_id() -> None:
     """Series listing a change-id as prerequisite must not be
     treated as newer revisions of that series."""
-- 
2.55.0