store: set aside a torn last journal line on open; server waits for lock

A crash or power loss mid-append can leave a partial last line. The
writer acknowledges an append only after its whole line is synced, so
that line was never acknowledged: opening now saves those bytes to
journal.jsonl.torn.<time>, truncates to the last newline, and journals
journal.tail_repaired (replay and projection ignore it; the GUI shows
it; books.digest reports it in journal_repairs). Damage before the last
newline is still corruption and still refuses to open. Tested at every
byte of the last append.

The MCP server now waits up to ten seconds for the writer lock: a
restarted server (a client reconnecting) raced the exiting one and died
with Locked, which also made mcp_schedules_and_recurring_over_stdio
flaky. The test harness now waits for each server to exit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GWVo7x3UHw5kwCUaWxiejX
This commit is contained in:
swaitsandClaude Opus 5.5 committed 2026-10-13 17:58:29 -06:00
1 parent 55f9e8f6e0
commit 54f632dfea
9 files changed
+293 -22

No files matched your search

+38 -1
View File
@@ -237,6 +237,8 @@ impl Store {
}
// The writer lock first: nothing is read until no one else writes.
let lock = lock_books(data_dir)?;
// An append cut off mid-line was never acknowledged: set it aside.
let torn = crate::journal::repair_torn_tail(&journal_path)?;
// Single verified scan; the writer reuses it.
let lines = read_and_verify_journal(&journal_path)?;
let journal = JournalWriter::from_lines(&journal_path, &lines)?;
@@ -305,6 +307,23 @@ impl Store {
base_currency,
};
store.catch_up(&lines)?;
if let Some(t) = torn {
// On the record, so the repair is part of the audit trail.
let line = store.append_journal(
"journal.tail_repaired",
json!({
"bytes": t.bytes,
"sha256": t.sha256,
"saved_as": t.saved_as.file_name().map(|n| n.to_string_lossy().into_owned()),
}),
)?;
apply_line_to_sqlite(&store.conn, &line)?;
eprintln!(
"wanderbooks: the journal's last line was cut off (an append that never finished); set aside {} bytes in {}",
t.bytes,
t.saved_as.display()
);
}
Ok(store)
}
@@ -414,6 +433,23 @@ impl Store {
self.base_currency
}
/// Every torn journal tail set aside on open, oldest first, as
/// (when, bytes, file it was saved to).
pub fn tail_repairs(&self) -> Result<Vec<(String, u64, String)>, StoreError> {
let lines = crate::journal::read_and_verify_journal(&self.data_dir.join(JOURNAL_FILE))?;
Ok(lines
.iter()
.filter(|l| l.line_type == "journal.tail_repaired")
.map(|l| {
(
l.ts.clone(),
l.payload["bytes"].as_u64().unwrap_or(0),
l.payload["saved_as"].as_str().unwrap_or("").to_string(),
)
})
.collect())
}
/// Data directory.
pub fn data_dir(&self) -> &Path {
&self.data_dir
@@ -3340,6 +3376,7 @@ pub(crate) fn replay_line_into_ledger(
p.get("class_id").and_then(|v| v.as_u64()).map(|v| v as u32),
)?;
}
"journal.tail_repaired" => {}
other => {
return Err(StoreError::JournalCorrupt(format!(
"unknown line type: {other}"
@@ -3545,7 +3582,7 @@ pub(crate) fn apply_line_to_sqlite(
fn apply_line_rows(conn: &Connection, line: &JournalLine) -> Result<(), StoreError> {
let p = &line.payload;
match line.line_type.as_str() {
"store.created" => {}
"store.created" | "journal.tail_repaired" => {}
"account.added" => {
conn.execute(
"INSERT INTO accounts(id, code, name, type, parent, placeholder, currency) VALUES (?,?,?,?,?,?,?)",