logtail: a log rotated while the panel was down is read to its end (stage 2b)

At start, a saved offset whose fingerprint was not the current file's meant
"start on the new file": what Postfix wrote between the panel's stop and the
rotation was never read - delivery statuses, stored lines, refusals, inbound
journal rows, the deferred counter.

- tracker.resume marks the case (behind, prev); follow calls readRotated
  before the new file: it opens mail.log.1 (delaycompress leaves it plain),
  checks by fingerprint that it is the file the place belongs to, reads it
  from the saved offset to its end through handle (catching up: the
  once-protections apply) and flushes the stored lines.
- Not found, not that file, or the offset past its end (rotated twice):
  logged, with the saved place's time - from when nothing was read.
- The fingerprint gains the file's identity (device.inode, fileid_unix.go;
  empty elsewhere): two logs that begin alike are two files - the
  review-2026-08-followups row on collisions. A fingerprint saved before
  (hash alone) still matches by its hash (sameLog), so the first start after
  the upgrade resumes where it was.
- store.LogtailState reads back updated_at.

Tests: stop, append, rotate, append, start - both halves read, once each,
the earlier line not again; a mail.log.1 that is another file is not read
from the place; sameLog with an old and a new fingerprint.

CHANGELOG [Unreleased], audit plan line ticked, roadmap 23/62.
Evidence: GOOS=linux go vet ./internal/logtail clean; go test ./internal/logtail
ok on Windows but for TestFollowTailsAndRotates; the server run follows
before the push.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

CHANGELOG.md

@@ -5,6 +5,18 @@ Format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/); version
 
 ## [Unreleased]
 
+### Fixed
+
+- **A log rotated while the panel was down is read to its end.** On start,
+  a saved place in a file that had been rotated meant "start on the new
+  file", and what Postfix wrote between the panel's stop and the rotation -
+  deliveries, refusals, deferrals - was never read. The tailer now reads the
+  rotated file (`mail.log.1`) from the saved place to its end first; when
+  that file is not the one the place belongs to (rotated twice), it says so
+  in the log, with the time from which nothing was read. The log's
+  fingerprint now includes the file's inode, so two files that begin alike
+  are still told apart (audit 2026-10-12, stage 2b).
+
 ## [2.5.0] - 2026-10-14
 
 Stages 1 and 2 of the 2026-10-12 audit: mail the delivery log showed wrong,

docs/audit-2026-10-12.md

@@ -736,7 +736,7 @@ stage follows as a cut of its own, with its own reviews. Five of the six
 change one thing, the tailer and the state it saves, and are done in this
 order - each stands on the one before.
 
-- [ ] **A log rotated while the panel was down is read to its end** (owner,
+- [x] **A log rotated while the panel was down is read to its end** (owner,
    2026-10-13; the Not done list of from-header-binding had it as "deferrals
    are lost", and it is wider). At start, a saved offset whose fingerprint
    is not the current file's means "start on the new file"

docs/roadmap.md

@@ -42,7 +42,7 @@ in `git log` and [CHANGELOG.md](../CHANGELOG.md).
 | from-header-binding | Hold a message's `From:` header against the application that sends it | shipped `2.4.0` | 9/9 | [plans/from-header-binding.md](plans/from-header-binding.md) |
 | panel-redesign | Panel redesign (accepted mockups, design contract) | shipped `2.0.0` | 27/27 | [plans/panel-redesign.md](plans/panel-redesign.md) |
 | schema-squash | Squash SQLite migrations into a 2.x baseline | built in panel-redesign (stage 1) | — | [plans/panel-redesign.md](plans/panel-redesign.md) |
-| audit-2026-10-12 | Follow-ups of the 2026-10-12 full-tree code audit, seven stages | in progress (stage 0 shipped `2.4.2`, stages 1+2 `2.5.0`; next stage 2b → `2.5.1`) | 22/62 | [audit-2026-10-12.md](audit-2026-10-12.md) |
+| audit-2026-10-12 | Follow-ups of the 2026-10-12 full-tree code audit, seven stages | in progress (stage 0 shipped `2.4.2`, stages 1+2 `2.5.0`; next stage 2b → `2.5.1`) | 23/62 | [audit-2026-10-12.md](audit-2026-10-12.md) |
 
 **Recommended order** (not binding; owner, 2026-10-09: feature order follows
 development convenience): **panel-redesign** shipped as `2.0.0` (2026-10-11) and

internal/logtail/fileid_other.go

@@ -0,0 +1,9 @@
+//go:build !unix
+
+package logtail
+
+import "os"
+
+// fileID is empty where the system has no inode to give (the panel runs on
+// Linux; this keeps the package building elsewhere for its tests).
+func fileID(*os.File) string { return "" }

internal/logtail/fileid_unix.go

@@ -0,0 +1,24 @@
+//go:build unix
+
+package logtail
+
+import (
+	"fmt"
+	"os"
+	"syscall"
+)
+
+// fileID names the file itself - device and inode - which a rename keeps and
+// a new file does not share. It is part of the log's fingerprint, so two log
+// files whose first bytes happen to match are still told apart.
+func fileID(f *os.File) string {
+	info, err := f.Stat()
+	if err != nil {
+		return ""
+	}
+	st, ok := info.Sys().(*syscall.Stat_t)
+	if !ok {
+		return ""
+	}
+	return fmt.Sprintf("%d.%d", st.Dev, st.Ino)
+}

internal/logtail/logtail.go

@@ -606,6 +606,10 @@ func follow(ctx context.Context, path string, tr *tracker, caughtUp chan struct{
 		log.Printf("log-tailer: seek %s: %v", path, err)
 	}
 	r.Reset(f) // the reader buffered from the pre-seek position
+	// Rotated while the panel was down: the end of the old file first.
+	if tr.readRotated(handle) {
+		flush()
+	}
 	defer func() {
 		if f != nil {
 			f.Close()

internal/logtail/logtail_test.go

@@ -815,3 +815,79 @@ func TestSplitTimestamp(t *testing.T) {
 		})
 	}
 }
+
+// A log rotated while the panel was down is read to its end before the new
+// one: what Postfix wrote between the stop and the rotation is in mail.log.1
+// (audit 2026-10-12, stage 2b). Both halves are read, once each.
+func TestALogRotatedWhileDownIsReadToItsEnd(t *testing.T) {
+	old := pollInterval
+	pollInterval = 10 * time.Millisecond
+	t.Cleanup(func() { pollInterval = old })
+
+	dir := t.TempDir()
+	path := filepath.Join(dir, "mail.log")
+	seed := strings.Repeat("host postfix/qmgr[1]: seed line, not a delivery\n", 20)
+	if err := os.WriteFile(path, []byte(seed), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	cs := &captureStore{}
+	stop := startRun(t, path, cs)
+	appendLine(t, path, "host postfix/smtp[1]: Q1: to=<a@example.net>, relay=x, dsn=2.0.0, status=sent (ok)")
+	waitFor(t, func() bool { return contains(cs.snapshot(), "Q1|a@example.net|sent") })
+	stop()
+
+	// Down: Postfix writes, then logrotate renames and creates, and Postfix
+	// goes on in the new file.
+	appendLine(t, path, "host postfix/smtp[1]: Q2: to=<b@example.net>, relay=x, dsn=2.0.0, status=sent (ok)")
+	if err := os.Rename(path, path+".1"); err != nil {
+		t.Fatal(err)
+	}
+	if err := os.WriteFile(path, []byte(strings.Repeat("host postfix/qmgr[1]: the new log\n", 20)+
+		"host postfix/smtp[1]: Q3: to=<c@example.net>, relay=x, dsn=5.1.1, status=bounced (nope)\n"), 0o644); err != nil {
+		t.Fatal(err)
+	}
+
+	cs.reset()
+	stop = startRun(t, path, cs)
+	waitFor(t, func() bool {
+		got := cs.snapshot()
+		return contains(got, "Q2|b@example.net|sent") && contains(got, "Q3|c@example.net|bounced")
+	})
+	stop()
+	n := 0
+	for _, c := range cs.snapshot() {
+		if strings.HasPrefix(c, "Q1|") {
+			t.Errorf("Q1 was read again: %s", c)
+		}
+		if strings.HasPrefix(c, "Q2|") {
+			n++
+		}
+	}
+	if n != 1 {
+		t.Errorf("Q2 read %d times, want once", n)
+	}
+}
+
+// A mail.log.1 that is not the file the saved place belongs to is not read
+// from that place: the new file is read from its start and the gap is said.
+func TestARotatedLogThatIsNotTheSavedOneIsNotRead(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "mail.log")
+	if err := os.WriteFile(path+".1", []byte(strings.Repeat("host postfix/qmgr[1]: another file altogether\n", 20)+
+		"host postfix/smtp[1]: QX: to=<x@example.net>, relay=x, dsn=2.0.0, status=sent (ok)\n"), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	tr := &tracker{path: path, behind: true, prev: store.LogtailState{Fingerprint: "not-this-one", Offset: 10}}
+	var read []string
+	if tr.readRotated(func(l string) { read = append(read, l) }) || len(read) != 0 {
+		t.Fatalf("read %d lines of a rotated file that is not the saved one", len(read))
+	}
+}
+
+// A fingerprint saved before it carried the file's identity still matches the
+// file by its hash.
+func TestSameLogAcceptsAFingerprintOfBefore(t *testing.T) {
+	if !sameLog("abc", "abc:1.2") || !sameLog("abc:1.2", "abc:1.2") || sameLog("abc:1.2", "abc:1.3") || sameLog("", "abc") {
+		t.Fatal("sameLog")
+	}
+}

internal/logtail/offset.go

@@ -1,12 +1,15 @@
 package logtail
 
 import (
+	"bufio"
 	"crypto/sha256"
 	"encoding/hex"
 	"errors"
+	"fmt"
 	"io"
 	"log"
 	"os"
+	"strings"
 	"time"
 
 	"github.com/mixeme/selfpost/internal/store"
@@ -38,6 +41,11 @@ type tracker struct {
 	fp       string // fingerprint of the file currently open ("" if too short)
 	saved    int64  // last offset written to the store
 	lastSave time.Time
+
+	// behind is set by resume when the saved place belongs to a file that is
+	// no longer the log - rotated while the panel was down; prev is that place.
+	behind bool
+	prev   store.LogtailState
 }
 
 // resume returns the byte offset the tailer should start reading f at, having
@@ -68,7 +76,7 @@ func (t *tracker) resume(f *os.File) int64 {
 	case !ok:
 		t.saved = size
 		return size
-	case prev.Fingerprint != "" && prev.Fingerprint == t.fp && prev.Offset <= size:
+	case prev.Fingerprint != "" && sameLog(prev.Fingerprint, t.fp) && prev.Offset <= size:
 		t.saved = prev.Offset
 		if prev.Offset < size {
 			log.Printf("log-tailer: resuming %s at offset %d (%d bytes to catch up)",
@@ -78,10 +86,74 @@ func (t *tracker) resume(f *os.File) int64 {
 	default:
 		log.Printf("log-tailer: %s changed while the panel was down; reading from the start", t.path)
 		t.saved = 0
+		t.behind, t.prev = prev.Fingerprint != "", prev
 		return 0
 	}
 }
 
+// readRotated reads, when the log was rotated while the panel was down, the
+// rest of the file the saved place belongs to: logrotate renamed it to
+// mail.log.1 (delaycompress leaves it plain), and what Postfix wrote there
+// between the panel's stop and the rotation - deliveries, refusals, deferrals -
+// was otherwise never read (audit 2026-10-12, stage 2b). It is read from the
+// saved offset to its end, line by line through handle, before the new file
+// is read from its start. A mail.log.1 that is not that file - the log
+// rotated more than once, or the file is gone - is said in the log, with the
+// time from which nothing was read. It reports whether it read anything.
+func (t *tracker) readRotated(handle func(string)) bool {
+	if !t.behind {
+		return false
+	}
+	t.behind = false
+	rotated := t.path + ".1"
+	lost := func(why string) {
+		log.Printf("log-tailer: %s; what was written to %s from %s until it was rotated was not read",
+			why, t.path, t.prev.UpdatedAt.UTC().Format(time.RFC3339))
+	}
+	f, err := os.Open(rotated)
+	if err != nil {
+		lost(fmt.Sprintf("%s cannot be opened (%v)", rotated, err))
+		return false
+	}
+	defer f.Close()
+	size, err := fileSize(f)
+	if err != nil || !sameLog(t.prev.Fingerprint, fingerprintOf(f)) || t.prev.Offset > size {
+		lost(fmt.Sprintf("%s is not the file the saved place belongs to (rotated more than once while the panel was down?)", rotated))
+		return false
+	}
+	if _, err := f.Seek(t.prev.Offset, io.SeekStart); err != nil {
+		lost(fmt.Sprintf("seek %s: %v", rotated, err))
+		return false
+	}
+	r := bufio.NewReader(f)
+	for {
+		line, err := r.ReadString('\n')
+		if line != "" {
+			handle(strings.TrimRight(line, "\r\n"))
+		}
+		if err != nil {
+			break
+		}
+	}
+	log.Printf("log-tailer: read %s from offset %d to its end (%d bytes) before the new %s",
+		rotated, t.prev.Offset, size-t.prev.Offset, t.path)
+	return true
+}
+
+// sameLog says whether a saved fingerprint is the current file's. One saved
+// before the fingerprint carried the file's identity is its hash alone, and is
+// compared with the hash.
+func sameLog(saved, current string) bool {
+	if saved == current {
+		return true
+	}
+	if !strings.Contains(saved, ":") {
+		hash, _, _ := strings.Cut(current, ":")
+		return saved != "" && saved == hash
+	}
+	return false
+}
+
 // adopt re-fingerprints after the follow loop switched to a rotated-in file and
 // persists the fresh start immediately, so a restart right after a rotation
 // does not resume at the old file's offset.
@@ -113,9 +185,11 @@ func (t *tracker) record(f *os.File, offset int64, force bool) {
 	t.lastSave = time.Now()
 }
 
-// fingerprintOf hashes the head of the file. It returns "" for a file too short
+// fingerprintOf hashes the head of the file and adds the file's identity
-// to identify — the head would still change as Postfix appends, so such a
+// (device and inode, where there is one): two log files whose first bytes
-// fingerprint could not be compared meaningfully on the next start.
+// match are still two files. It returns "" for a file too short to identify —
+// the head would still change as Postfix appends, so such a fingerprint could
+// not be compared meaningfully on the next start.
 func fingerprintOf(f *os.File) string {
 	buf := make([]byte, fingerprintSize)
 	n, err := f.ReadAt(buf, 0)
@@ -127,7 +201,11 @@ func fingerprintOf(f *os.File) string {
 		return ""
 	}
 	sum := sha256.Sum256(buf)
-	return hex.EncodeToString(sum[:])
+	fp := hex.EncodeToString(sum[:])
+	if id := fileID(f); id != "" {
+		fp += ":" + id
+	}
+	return fp
 }
 
 func fileSize(f *os.File) (int64, error) {

internal/store/logtail.go

@@ -14,17 +14,23 @@ import (
 type LogtailState struct {
 	Fingerprint string
 	Offset      int64
+	// UpdatedAt is when the position was saved; read back only (zero on save).
+	UpdatedAt time.Time
 }
 
 // LogtailState returns the stored read position for path. ok is false the first
 // time a path is followed (nothing persisted yet), which the tailer treats as
 // "start at the end".
 func (s *Store) LogtailState(path string) (LogtailState, bool, error) {
-	var st LogtailState
+	var (
+		st      LogtailState
+		updated string
+	)
 	err := s.db.QueryRow(
-		`SELECT fingerprint, read_offset FROM logtail_state WHERE path = ?`,
+		`SELECT fingerprint, read_offset, updated_at FROM logtail_state WHERE path = ?`,
 		path,
-	).Scan(&st.Fingerprint, &st.Offset)
+	).Scan(&st.Fingerprint, &st.Offset, &updated)
+	st.UpdatedAt, _ = time.Parse(time.RFC3339, updated)
 	if errors.Is(err, sql.ErrNoRows) {
 		return LogtailState{}, false, nil
 	}