feat: Disable wal_autocheckpoint

From https://www.sqlite.org/wal.html:
> The default strategy is to allow successive write transactions to grow the WAL until the WAL
  becomes about 1000 pages in size, then to run a checkpoint operation for each subsequent COMMIT
  until the WAL is reset to be smaller than 1000 pages. By default, the checkpoint will be run
  automatically by the same thread that does the COMMIT that pushes the WAL over its size
  limit. This has the effect of causing most COMMIT operations to be very fast but an occasional
  COMMIT (those that trigger a checkpoint) to be much slower.

And while autocheckpoint runs in the `PASSIVE` mode and thus doesn't block concurrent readers and
writers, in our design it blocks writers because it's done under `write_mutex` locked and thus may
cause the app to stuck for noticeable time. Let's disable autocheckpointing then, we can't rely on
it anyway. Instead, run a `TRUNCATE` checkpoint from `inbox_loop()` if the WAL is >= 4K pages and a
`PASSIVE` checkpoint otherwise.

Author: iequidoo

src/context.rs

@@ -413,7 +413,7 @@ impl Context {
 
     /// Changes encrypted database passphrase.
     pub async fn change_passphrase(&self, passphrase: String) -> Result<()> {
-        self.sql.change_passphrase(passphrase).await?;
+        self.sql.change_passphrase(self, passphrase).await?;
         Ok(())
     }
 

src/scheduler.rs

@@ -26,7 +26,7 @@ use crate::location;
 use crate::log::{LogExt, error, info, warn};
 use crate::message::MsgId;
 use crate::smtp::{Smtp, send_smtp_messages};
-use crate::sql;
+use crate::sql::{self, Sql};
 use crate::tools::{self, duration_to_str, maybe_add_time_based_warnings, time, time_elapsed};
 
 pub(crate) mod connectivity;
@@ -506,6 +506,11 @@ async fn inbox_fetch_idle(ctx: &Context, imap: &mut Imap, mut session: Session)
                 last_housekeeping_time.saturating_add(constants::HOUSEKEEPING_PERIOD);
             if next_housekeeping_time <= time() {
                 sql::housekeeping(ctx).await.log_err(ctx).ok();
+            } else {
+                let force_truncate = false;
+                if let Err(err) = Sql::wal_checkpoint(ctx, force_truncate).await {
+                    warn!(ctx, "wal_checkpoint() failed: {err:#}.");
+                }
             }
         }
         Err(err) => {

src/sql.rs

@@ -298,7 +298,11 @@ impl Sql {
     /// The database must already be encrypted and the passphrase cannot be empty.
     /// It is impossible to turn encrypted database into unencrypted
     /// and vice versa this way, use import/export for this.
-    pub async fn change_passphrase(&self, passphrase: String) -> Result<()> {
+    pub(crate) async fn change_passphrase(
+        &self,
+        _context: &Context,
+        passphrase: String,
+    ) -> Result<()> {
         let mut lock = self.pool.write().await;
 
         let pool = lock.take().context("SQL connection pool is not open")?;
@@ -646,8 +650,12 @@ impl Sql {
         &self.config_cache
     }
 
-    /// Runs a checkpoint operation in TRUNCATE mode, so the WAL file is truncated to 0 bytes.
-    pub(crate) async fn wal_checkpoint(context: &Context) -> Result<()> {
+    /// Runs a WAL checkpoint operation.
+    ///
+    /// * `force_truncate` - Force TRUNCATE mode to truncate the WAL file to 0 bytes, otherwise only
+    ///   run PASSIVE mode if the WAL isn't too large. NB: Truncating blocks all db connections for
+    ///   some time.
+    pub(crate) async fn wal_checkpoint(context: &Context, force_truncate: bool) -> Result<()> {
         let t_start = Time::now();
         let lock = context.sql.pool.read().await;
         let Some(pool) = lock.as_ref() else {
@@ -658,13 +666,19 @@ impl Sql {
         // Do as much work as possible without blocking anybody.
         let query_only = true;
         let conn = pool.get(query_only).await?;
-        tokio::task::block_in_place(|| {
+        let pages_total = tokio::task::block_in_place(|| {
             // Execute some transaction causing the WAL file to be opened so that the
             // `wal_checkpoint()` can proceed, otherwise it fails when called the first time,
             // see https://sqlite.org/forum/forumpost/7512d76a05268fc8.
             conn.query_row("PRAGMA table_list", [], |_| Ok(()))?;
-            conn.query_row("PRAGMA wal_checkpoint(PASSIVE)", [], |_| Ok(()))
+            conn.query_row("PRAGMA wal_checkpoint(PASSIVE)", [], |row| {
+                let pages_total: i64 = row.get(1)?;
+                Ok(pages_total)
+            })
         })?;
+        if !force_truncate && pages_total < 4096 {
+            return Ok(());
+        }
 
         // Kick out writers.
         const _: () = assert!(Sql::N_DB_CONNECTIONS > 1, "Deadlock possible");
@@ -735,6 +749,7 @@ fn new_connection(path: &Path, passphrase: &str) -> Result<Connection> {
          PRAGMA busy_timeout = 0; -- fail immediately
          PRAGMA soft_heap_limit = 8388608; -- 8 MiB limit, same as set in Android SQLiteDatabase.
          PRAGMA foreign_keys=on;
+         PRAGMA wal_autocheckpoint=N;
          ",
     )?;
 
@@ -843,7 +858,8 @@ pub async fn housekeeping(context: &Context) -> Result<()> {
     // bigger than 200M) and also make sure we truncate the WAL periodically. Auto-checkponting does
     // not normally truncate the WAL (unless the `journal_size_limit` pragma is set), see
     // https://www.sqlite.org/wal.html.
-    if let Err(err) = Sql::wal_checkpoint(context).await {
+    let force_truncate = true;
+    if let Err(err) = Sql::wal_checkpoint(context, force_truncate).await {
         warn!(context, "wal_checkpoint() failed: {err:#}.");
         debug_assert!(false);
     }

src/sql/sql_tests.rs

@@ -263,7 +263,7 @@ async fn test_sql_change_passphrase() -> Result<()> {
     sql.open(&t, "foo".to_string())
         .await
         .context("failed to open the database second time")?;
-    sql.change_passphrase("bar".to_string())
+    sql.change_passphrase(&t, "bar".to_string())
         .await
         .context("failed to change passphrase")?;