mirror of
https://github.com/openai/codex.git
synced 2026-10-21 16:38:24 +00:00
Enforce writer ownership for thread archive and deletion (#35031)
## Why Paginated threads allow only one app-server process to write at a time. Archive and delete operations must not mutate a thread or its spawned descendants while another process owns any of them. ## What changed - Acquire paginated writer locks for the full archive or deletion set before changing rollout files, including descendants whose rollout has not materialized yet. - Add a batch archive store operation so ownership conflicts are detected before any thread in the subtree is archived. - Return ownership conflicts as JSON-RPC invalid-request errors and document the behavior. ## Testing - Cover archive and deletion conflicts for owned threads and descendants, including unmaterialized rollouts. - Verify deletion still removes rollouts with unreadable metadata. GitOrigin-RevId: f62a3e0c1ee2631cea48dd9145b588a1486e2425
This commit is contained in:
@@ -324,7 +324,7 @@ By default, `thread/resume` includes the reconstructed turn history in `thread.t
|
||||
|
||||
Paginated threads keep the same resume contract as legacy threads. A default resume materializes the full projected history into `thread.turns`; `excludeTurns: true` keeps that array empty and includes `turnsBackwardsCursor` and `itemsBackwardsCursor` for the durable history visible at the resume boundary. Pass each cursor directly to its matching list API with `sortDirection: "desc"`; the first page includes the cursor's head row, while newer records arrive through live notifications. Either cursor is `null` when there is no durable row yet.
|
||||
|
||||
Only one app-server process can hold a paginated thread open for writing at a time. If another process already owns the thread, `thread/resume` fails with JSON-RPC error `-32600`. Read-only requests remain available without resuming the thread.
|
||||
Only one app-server process can hold a paginated thread open for writing at a time. If another process already owns the thread, `thread/resume`, `thread/archive`, and `thread/delete` fail with JSON-RPC error `-32600`. Archive and deletion also fail if another process owns any spawned descendant. Read-only requests remain available without resuming the thread.
|
||||
|
||||
Experimental clients that want the live resume subscription plus a turns page in one round trip can pass `initialTurnsPage`. It accepts the same `limit`, `sortDirection`, and `itemsView` controls as `thread/turns/list`; omitted controls use its defaults. The response includes `initialTurnsPage` with `nextCursor` and `backwardsCursor` for follow-up pagination.
|
||||
|
||||
|
||||
@@ -465,6 +465,7 @@ use codex_rollout::state_db::reconcile_rollout;
|
||||
use codex_state::ThreadMetadata;
|
||||
use codex_state::log_db::LogDbLayer;
|
||||
use codex_thread_store::ArchiveThreadParams as StoreArchiveThreadParams;
|
||||
use codex_thread_store::ArchiveThreadsParams as StoreArchiveThreadsParams;
|
||||
use codex_thread_store::DeleteThreadsParams as StoreDeleteThreadsParams;
|
||||
use codex_thread_store::GitInfoPatch as StoreGitInfoPatch;
|
||||
use codex_thread_store::ItemSortKey as StoreItemSortKey;
|
||||
|
||||
@@ -149,7 +149,9 @@ fn thread_store_delete_error(err: ThreadStoreError) -> JSONRPCErrorError {
|
||||
ThreadStoreError::ThreadNotFound { thread_id } => {
|
||||
invalid_request(format!("thread not found: {thread_id}"))
|
||||
}
|
||||
ThreadStoreError::InvalidRequest { message } => invalid_request(message),
|
||||
ThreadStoreError::InvalidRequest { message } | ThreadStoreError::Conflict { message } => {
|
||||
invalid_request(message)
|
||||
}
|
||||
ThreadStoreError::Unsupported { operation } => {
|
||||
unsupported_thread_store_operation(operation)
|
||||
}
|
||||
|
||||
@@ -1430,7 +1430,7 @@ impl ThreadRequestProcessor {
|
||||
let thread_id = ThreadId::from_string(¶ms.thread_id)
|
||||
.map_err(|err| invalid_request(format!("invalid session id: {err}")))?;
|
||||
|
||||
let thread_ids = self.state_db_spawn_subtree_thread_ids(thread_id).await?;
|
||||
let subtree_thread_ids = self.state_db_spawn_subtree_thread_ids(thread_id).await?;
|
||||
|
||||
let mut archive_thread_ids = Vec::new();
|
||||
match self
|
||||
@@ -1449,7 +1449,7 @@ impl ThreadRequestProcessor {
|
||||
}
|
||||
Err(err) => return Err(thread_store_archive_error("archive", err)),
|
||||
}
|
||||
for descendant_thread_id in thread_ids.into_iter().skip(1) {
|
||||
for descendant_thread_id in subtree_thread_ids.iter().copied().skip(1) {
|
||||
match self
|
||||
.thread_store
|
||||
.read_thread(StoreReadThreadParams {
|
||||
@@ -1472,46 +1472,26 @@ impl ThreadRequestProcessor {
|
||||
}
|
||||
}
|
||||
|
||||
let mut archived_thread_ids = Vec::new();
|
||||
let Some((parent_thread_id, descendant_thread_ids)) = archive_thread_ids.split_first()
|
||||
else {
|
||||
return Ok((ThreadArchiveResponse {}, archived_thread_ids));
|
||||
};
|
||||
if archive_thread_ids.is_empty() {
|
||||
return Ok((ThreadArchiveResponse {}, Vec::new()));
|
||||
}
|
||||
|
||||
self.prepare_thread_for_archive(*parent_thread_id).await;
|
||||
match self
|
||||
archive_thread_ids[1..].reverse();
|
||||
for &thread_id_to_archive in &archive_thread_ids {
|
||||
self.prepare_thread_for_archive(thread_id_to_archive).await;
|
||||
}
|
||||
|
||||
let archived_thread_ids = self
|
||||
.thread_store
|
||||
.archive_thread(StoreArchiveThreadParams {
|
||||
thread_id: *parent_thread_id,
|
||||
.archive_threads(StoreArchiveThreadsParams {
|
||||
thread_ids: archive_thread_ids,
|
||||
writer_lock_thread_ids: subtree_thread_ids,
|
||||
})
|
||||
.await
|
||||
{
|
||||
Ok(()) => {
|
||||
archived_thread_ids.push(parent_thread_id.to_string());
|
||||
}
|
||||
Err(err) => return Err(thread_store_archive_error("archive", err)),
|
||||
}
|
||||
|
||||
for descendant_thread_id in descendant_thread_ids.iter().rev().copied() {
|
||||
self.prepare_thread_for_archive(descendant_thread_id).await;
|
||||
match self
|
||||
.thread_store
|
||||
.archive_thread(StoreArchiveThreadParams {
|
||||
thread_id: descendant_thread_id,
|
||||
})
|
||||
.await
|
||||
{
|
||||
Ok(()) => {
|
||||
archived_thread_ids.push(descendant_thread_id.to_string());
|
||||
}
|
||||
Err(err) => {
|
||||
warn!(
|
||||
"failed to archive spawned descendant thread {descendant_thread_id} while archiving {thread_id}: {err}"
|
||||
);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
.map_err(|err| thread_store_archive_error("archive", err))?
|
||||
.into_iter()
|
||||
.map(|thread_id| thread_id.to_string())
|
||||
.collect();
|
||||
Ok((ThreadArchiveResponse {}, archived_thread_ids))
|
||||
}
|
||||
|
||||
@@ -4953,7 +4933,9 @@ pub(super) fn core_thread_write_error(operation: &str, err: CodexErr) -> JSONRPC
|
||||
|
||||
fn thread_store_archive_error(operation: &str, err: ThreadStoreError) -> JSONRPCErrorError {
|
||||
match err {
|
||||
ThreadStoreError::InvalidRequest { message } => invalid_request(message),
|
||||
ThreadStoreError::InvalidRequest { message } | ThreadStoreError::Conflict { message } => {
|
||||
invalid_request(message)
|
||||
}
|
||||
ThreadStoreError::Unsupported {
|
||||
operation: unsupported_operation,
|
||||
} => unsupported_thread_store_operation(unsupported_operation),
|
||||
|
||||
@@ -9,6 +9,7 @@ use codex_app_server_protocol::RequestId;
|
||||
use codex_app_server_protocol::ThreadArchiveParams;
|
||||
use codex_app_server_protocol::ThreadArchiveResponse;
|
||||
use codex_app_server_protocol::ThreadArchivedNotification;
|
||||
use codex_app_server_protocol::ThreadHistoryMode;
|
||||
use codex_app_server_protocol::ThreadResumeParams;
|
||||
use codex_app_server_protocol::ThreadResumeResponse;
|
||||
use codex_app_server_protocol::ThreadStartParams;
|
||||
@@ -33,6 +34,75 @@ use tokio::time::timeout;
|
||||
|
||||
const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);
|
||||
|
||||
#[tokio::test]
|
||||
async fn thread_archive_rejects_owned_unmaterialized_paginated_descendant() -> Result<()> {
|
||||
let server = create_mock_responses_server_repeating_assistant("Done").await;
|
||||
let codex_home = TempDir::new()?;
|
||||
MockResponsesConfig::new(&server.uri()).write(codex_home.path())?;
|
||||
let parent_id = create_fake_rollout(
|
||||
codex_home.path(),
|
||||
"2025-01-01T00-00-00",
|
||||
"2025-01-01T00:00:00Z",
|
||||
"parent",
|
||||
Some("mock_provider"),
|
||||
/*git_info*/ None,
|
||||
)?;
|
||||
let parent_thread_id = ThreadId::from_string(&parent_id)?;
|
||||
let mut owner = TestAppServer::builder()
|
||||
.with_codex_home(codex_home.path())
|
||||
.build_initialized()
|
||||
.await?;
|
||||
let ThreadStartResponse { thread: child, .. } = owner
|
||||
.start_thread(ThreadStartParams {
|
||||
history_mode: Some(ThreadHistoryMode::Paginated),
|
||||
..Default::default()
|
||||
})
|
||||
.await?;
|
||||
let child_thread_id = ThreadId::from_string(&child.id)?;
|
||||
let state_db = StateRuntime::init(
|
||||
codex_state::SqliteConfig::new_for_testing(codex_home.path().abs()),
|
||||
"mock_provider".into(),
|
||||
)
|
||||
.await?;
|
||||
state_db
|
||||
.upsert_thread_spawn_edge(
|
||||
parent_thread_id,
|
||||
child_thread_id,
|
||||
DirectionalThreadSpawnEdgeStatus::Open,
|
||||
)
|
||||
.await?;
|
||||
|
||||
let mut other = TestAppServer::builder()
|
||||
.with_codex_home(codex_home.path())
|
||||
.build_initialized()
|
||||
.await?;
|
||||
let request_id = other
|
||||
.send_thread_archive_request(ThreadArchiveParams {
|
||||
thread_id: parent_id.clone(),
|
||||
})
|
||||
.await?;
|
||||
let error: JSONRPCError = timeout(
|
||||
DEFAULT_READ_TIMEOUT,
|
||||
other.read_stream_until_error_message(RequestId::Integer(request_id)),
|
||||
)
|
||||
.await??;
|
||||
assert_eq!(error.error.code, -32600);
|
||||
assert_eq!(
|
||||
error.error.message,
|
||||
format!("thread {} already has an active writer", child.id)
|
||||
);
|
||||
timeout(DEFAULT_READ_TIMEOUT, owner.shutdown_gracefully()).await??;
|
||||
let _: ThreadArchiveResponse = other
|
||||
.request(|request_id| ClientRequest::ThreadArchive {
|
||||
request_id,
|
||||
params: ThreadArchiveParams {
|
||||
thread_id: parent_id,
|
||||
},
|
||||
})
|
||||
.await?;
|
||||
Ok(())
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn thread_archive_requires_materialized_rollout() -> Result<()> {
|
||||
let server = create_mock_responses_server_repeating_assistant("Done").await;
|
||||
|
||||
@@ -1,6 +1,9 @@
|
||||
use anyhow::Result;
|
||||
use app_test_support::MockResponsesConfig;
|
||||
use app_test_support::TestAppServer;
|
||||
use app_test_support::create_fake_paginated_rollout;
|
||||
use app_test_support::create_fake_rollout;
|
||||
use app_test_support::create_mock_responses_server_repeating_assistant;
|
||||
use codex_app_server_protocol::ClientRequest;
|
||||
use codex_app_server_protocol::JSONRPCError;
|
||||
use codex_app_server_protocol::RequestId;
|
||||
@@ -9,6 +12,8 @@ use codex_app_server_protocol::ThreadDeleteResponse;
|
||||
use codex_app_server_protocol::ThreadDeletedNotification;
|
||||
use codex_app_server_protocol::ThreadLoadedListParams;
|
||||
use codex_app_server_protocol::ThreadLoadedListResponse;
|
||||
use codex_app_server_protocol::ThreadResumeParams;
|
||||
use codex_app_server_protocol::ThreadResumeResponse;
|
||||
use codex_app_server_protocol::ThreadStartParams;
|
||||
use codex_app_server_protocol::ThreadStartResponse;
|
||||
use codex_core::find_thread_path_by_id_str;
|
||||
@@ -25,6 +30,63 @@ use tokio::time::timeout;
|
||||
|
||||
const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);
|
||||
|
||||
#[tokio::test]
|
||||
async fn thread_delete_rejects_paginated_writer_owned_by_another_process() -> Result<()> {
|
||||
let server = create_mock_responses_server_repeating_assistant("Done").await;
|
||||
let codex_home = TempDir::new()?;
|
||||
MockResponsesConfig::new(&server.uri()).write(codex_home.path())?;
|
||||
let thread_id = create_fake_paginated_rollout(
|
||||
codex_home.path(),
|
||||
"2025-01-01T00-00-00",
|
||||
"2025-01-01T00:00:00Z",
|
||||
"owned",
|
||||
Some("mock_provider"),
|
||||
/*git_info*/ None,
|
||||
)?;
|
||||
let mut owner = TestAppServer::builder()
|
||||
.with_codex_home(codex_home.path())
|
||||
.build_initialized()
|
||||
.await?;
|
||||
let _: ThreadResumeResponse = owner
|
||||
.request(|request_id| ClientRequest::ThreadResume {
|
||||
request_id,
|
||||
params: ThreadResumeParams {
|
||||
thread_id: thread_id.clone(),
|
||||
exclude_turns: true,
|
||||
..Default::default()
|
||||
},
|
||||
})
|
||||
.await?;
|
||||
|
||||
let mut other = TestAppServer::builder()
|
||||
.with_codex_home(codex_home.path())
|
||||
.build_initialized()
|
||||
.await?;
|
||||
let request_id = other
|
||||
.send_thread_delete_request(ThreadDeleteParams {
|
||||
thread_id: thread_id.clone(),
|
||||
})
|
||||
.await?;
|
||||
let error: JSONRPCError = timeout(
|
||||
DEFAULT_READ_TIMEOUT,
|
||||
other.read_stream_until_error_message(RequestId::Integer(request_id)),
|
||||
)
|
||||
.await??;
|
||||
assert_eq!(error.error.code, -32600);
|
||||
assert_eq!(
|
||||
error.error.message,
|
||||
format!("thread {thread_id} already has an active writer")
|
||||
);
|
||||
timeout(DEFAULT_READ_TIMEOUT, owner.shutdown_gracefully()).await??;
|
||||
let _: ThreadDeleteResponse = other
|
||||
.request(|request_id| ClientRequest::ThreadDelete {
|
||||
request_id,
|
||||
params: ThreadDeleteParams { thread_id },
|
||||
})
|
||||
.await?;
|
||||
Ok(())
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn thread_delete_deletes_spawned_descendants() -> Result<()> {
|
||||
let codex_home = TempDir::new()?;
|
||||
|
||||
@@ -24,6 +24,7 @@ pub use store::ThreadStore;
|
||||
pub use store::ThreadStoreFuture;
|
||||
pub use types::AppendThreadItemsParams;
|
||||
pub use types::ArchiveThreadParams;
|
||||
pub use types::ArchiveThreadsParams;
|
||||
pub use types::ClearableField;
|
||||
pub use types::CreateThreadParams;
|
||||
pub use types::DeleteThreadParams;
|
||||
|
||||
@@ -2,10 +2,58 @@ use super::LocalThreadStore;
|
||||
use super::helpers::matching_rollout_file_name;
|
||||
use super::helpers::scoped_rollout_path;
|
||||
use crate::ArchiveThreadParams;
|
||||
use crate::ArchiveThreadsParams;
|
||||
use crate::ThreadStoreError;
|
||||
use crate::ThreadStoreResult;
|
||||
use chrono::Utc;
|
||||
use codex_rollout::find_thread_path_by_id_str;
|
||||
use tracing::warn;
|
||||
|
||||
pub(super) async fn archive_threads(
|
||||
store: &LocalThreadStore,
|
||||
params: ArchiveThreadsParams,
|
||||
) -> ThreadStoreResult<Vec<codex_protocol::ThreadId>> {
|
||||
let thread_ids = params.thread_ids;
|
||||
if thread_ids.is_empty() {
|
||||
return Ok(Vec::new());
|
||||
}
|
||||
|
||||
let mut lock_thread_ids = params.writer_lock_thread_ids;
|
||||
lock_thread_ids.extend(thread_ids.iter().copied());
|
||||
lock_thread_ids.sort_unstable_by_key(ToString::to_string);
|
||||
lock_thread_ids.dedup();
|
||||
let mut _live_writer_guards = Vec::with_capacity(lock_thread_ids.len());
|
||||
for thread_id in &lock_thread_ids {
|
||||
_live_writer_guards.push(store.live_writer_locks.lock(*thread_id).await);
|
||||
if store
|
||||
.live_recorders
|
||||
.lock()
|
||||
.await
|
||||
.get(thread_id)
|
||||
.is_some_and(|entry| entry.writer_lock.is_some())
|
||||
{
|
||||
return Err(ThreadStoreError::Conflict {
|
||||
message: format!("thread {thread_id} already has an active writer"),
|
||||
});
|
||||
}
|
||||
}
|
||||
let _writer_guards = store
|
||||
.acquire_paginated_writer_locks(&lock_thread_ids)
|
||||
.await?;
|
||||
|
||||
let parent_thread_id = thread_ids[0];
|
||||
let mut archived_thread_ids = Vec::new();
|
||||
for thread_id in thread_ids {
|
||||
match archive_thread(store, ArchiveThreadParams { thread_id }).await {
|
||||
Ok(()) => archived_thread_ids.push(thread_id),
|
||||
Err(err) if archived_thread_ids.is_empty() => return Err(err),
|
||||
Err(err) => warn!(
|
||||
"failed to archive spawned descendant thread {thread_id} while archiving {parent_thread_id}: {err}"
|
||||
),
|
||||
}
|
||||
}
|
||||
Ok(archived_thread_ids)
|
||||
}
|
||||
|
||||
pub(super) async fn archive_thread(
|
||||
store: &LocalThreadStore,
|
||||
@@ -64,6 +112,7 @@ mod tests {
|
||||
use chrono::Utc;
|
||||
use codex_protocol::ThreadId;
|
||||
use codex_protocol::protocol::SessionSource;
|
||||
use codex_protocol::protocol::ThreadHistoryMode;
|
||||
use codex_rollout::ARCHIVED_SESSIONS_SUBDIR;
|
||||
use codex_utils_absolute_path::test_support::PathExt;
|
||||
use pretty_assertions::assert_eq;
|
||||
@@ -77,6 +126,50 @@ mod tests {
|
||||
use crate::local::LocalThreadStore;
|
||||
use crate::local::test_support::test_config;
|
||||
use crate::local::test_support::write_session_file;
|
||||
use crate::local::test_support::write_session_file_with_history_mode;
|
||||
|
||||
#[tokio::test]
|
||||
async fn archive_threads_rejects_owned_descendants_before_archiving_anything() {
|
||||
let home = TempDir::new().expect("temp dir");
|
||||
let store = LocalThreadStore::new(test_config(home.path()), /*state_db*/ None);
|
||||
let owner = LocalThreadStore::new(test_config(home.path()), /*state_db*/ None);
|
||||
let parent_uuid = Uuid::from_u128(203);
|
||||
let parent_thread_id =
|
||||
ThreadId::from_string(&parent_uuid.to_string()).expect("valid parent thread id");
|
||||
let parent_path = write_session_file_with_history_mode(
|
||||
home.path(),
|
||||
"2025-01-03T12-00-00",
|
||||
parent_uuid,
|
||||
ThreadHistoryMode::Paginated,
|
||||
)
|
||||
.expect("parent session file");
|
||||
let child_uuid = Uuid::from_u128(204);
|
||||
let child_thread_id =
|
||||
ThreadId::from_string(&child_uuid.to_string()).expect("valid child thread id");
|
||||
let child_path = write_session_file_with_history_mode(
|
||||
home.path(),
|
||||
"2025-01-03T12-00-01",
|
||||
child_uuid,
|
||||
ThreadHistoryMode::Paginated,
|
||||
)
|
||||
.expect("child session file");
|
||||
let _owner_guard = owner
|
||||
.writer_lock_coordinator
|
||||
.acquire(child_thread_id)
|
||||
.expect("acquire child writer lock");
|
||||
|
||||
let error = store
|
||||
.archive_threads(ArchiveThreadsParams {
|
||||
thread_ids: vec![parent_thread_id, child_thread_id],
|
||||
writer_lock_thread_ids: Vec::new(),
|
||||
})
|
||||
.await
|
||||
.expect_err("owned descendant should block archive");
|
||||
|
||||
assert!(matches!(error, ThreadStoreError::Conflict { .. }));
|
||||
assert!(parent_path.exists());
|
||||
assert!(child_path.exists());
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn archive_thread_moves_rollout_to_archived_collection() {
|
||||
|
||||
@@ -34,7 +34,8 @@ pub(super) async fn delete_thread(
|
||||
if reference_index.reference_count(thread_id) > 0 {
|
||||
return Err(referenced_thread_error(thread_id));
|
||||
}
|
||||
delete_thread_after_reference_check(store, thread_id).await
|
||||
let mut writer_guards = store.acquire_paginated_writer_locks(&[thread_id]).await?;
|
||||
delete_thread_after_reference_check(store, thread_id, &mut writer_guards).await
|
||||
}
|
||||
|
||||
pub(super) async fn delete_threads(
|
||||
@@ -50,7 +51,7 @@ pub(super) async fn delete_threads(
|
||||
let mut lock_thread_ids: Vec<_> = deletion_set.iter().copied().collect();
|
||||
lock_thread_ids.sort_unstable_by_key(ToString::to_string);
|
||||
let mut _live_writer_guards = Vec::with_capacity(lock_thread_ids.len());
|
||||
for thread_id in lock_thread_ids {
|
||||
for &thread_id in &lock_thread_ids {
|
||||
_live_writer_guards.push(store.live_writer_locks.lock(thread_id).await);
|
||||
}
|
||||
|
||||
@@ -78,8 +79,11 @@ pub(super) async fn delete_threads(
|
||||
}
|
||||
}
|
||||
|
||||
let mut writer_guards = store
|
||||
.acquire_paginated_writer_locks(&lock_thread_ids)
|
||||
.await?;
|
||||
for thread_id in thread_ids {
|
||||
match delete_thread_after_reference_check(store, thread_id).await {
|
||||
match delete_thread_after_reference_check(store, thread_id, &mut writer_guards).await {
|
||||
Ok(()) | Err(ThreadStoreError::ThreadNotFound { .. }) => {}
|
||||
Err(err) => return Err(err),
|
||||
}
|
||||
@@ -106,6 +110,7 @@ fn referenced_thread_error(thread_id: codex_protocol::ThreadId) -> ThreadStoreEr
|
||||
async fn delete_thread_after_reference_check(
|
||||
store: &LocalThreadStore,
|
||||
thread_id: codex_protocol::ThreadId,
|
||||
writer_guards: &mut Vec<super::writer_lock::WriterLockGuard>,
|
||||
) -> ThreadStoreResult<()> {
|
||||
let thread_id_str = thread_id.to_string();
|
||||
let state_db_ctx = store.state_db().await;
|
||||
@@ -145,12 +150,15 @@ async fn delete_thread_after_reference_check(
|
||||
}
|
||||
}
|
||||
// Drop the recorder before removing files, but retain its writer lock until cleanup finishes.
|
||||
let _writer_lock = store
|
||||
if let Some(writer_lock) = store
|
||||
.live_recorders
|
||||
.lock()
|
||||
.await
|
||||
.remove(&thread_id)
|
||||
.and_then(|entry| entry._writer_lock);
|
||||
.and_then(|entry| entry.writer_lock)
|
||||
{
|
||||
writer_guards.push(writer_lock);
|
||||
}
|
||||
let found_rollout_path = !rollout_paths.is_empty();
|
||||
for rollout_path in rollout_paths {
|
||||
delete_rollout_file(store, rollout_path.as_path(), thread_id)?;
|
||||
@@ -395,6 +403,89 @@ mod tests {
|
||||
assert!(!child_path.exists());
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn delete_threads_rejects_owned_descendants_before_deleting_anything() {
|
||||
let home = TempDir::new().expect("temp dir");
|
||||
let store = LocalThreadStore::new(test_config(home.path()), /*state_db*/ None);
|
||||
let owner = LocalThreadStore::new(test_config(home.path()), /*state_db*/ None);
|
||||
let parent_uuid = Uuid::from_u128(309);
|
||||
let parent_thread_id =
|
||||
ThreadId::from_string(&parent_uuid.to_string()).expect("valid parent thread id");
|
||||
let parent_path = write_session_file_with_history_mode(
|
||||
home.path(),
|
||||
"2025-01-03T12-00-00",
|
||||
parent_uuid,
|
||||
ThreadHistoryMode::Paginated,
|
||||
)
|
||||
.expect("parent session file");
|
||||
let child_uuid = Uuid::from_u128(310);
|
||||
let child_thread_id =
|
||||
ThreadId::from_string(&child_uuid.to_string()).expect("valid child thread id");
|
||||
let child_path = write_session_file_with_history_mode(
|
||||
home.path(),
|
||||
"2025-01-03T12-00-01",
|
||||
child_uuid,
|
||||
ThreadHistoryMode::Paginated,
|
||||
)
|
||||
.expect("child session file");
|
||||
let _owner_guard = owner
|
||||
.writer_lock_coordinator
|
||||
.acquire(child_thread_id)
|
||||
.expect("acquire child writer lock");
|
||||
|
||||
let error = store
|
||||
.delete_threads(DeleteThreadsParams {
|
||||
thread_ids: vec![parent_thread_id, child_thread_id],
|
||||
})
|
||||
.await
|
||||
.expect_err("owned descendant should block deletion");
|
||||
|
||||
assert!(matches!(error, ThreadStoreError::Conflict { .. }));
|
||||
assert!(parent_path.exists());
|
||||
assert!(child_path.exists());
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn delete_threads_rejects_owned_thread_before_rollout_materializes() {
|
||||
let home = TempDir::new().expect("temp dir");
|
||||
let store = LocalThreadStore::new(test_config(home.path()), /*state_db*/ None);
|
||||
let owner = LocalThreadStore::new(test_config(home.path()), /*state_db*/ None);
|
||||
let thread_id = ThreadId::default();
|
||||
let _owner_guard = owner
|
||||
.writer_lock_coordinator
|
||||
.acquire(thread_id)
|
||||
.expect("acquire writer lock before rollout exists");
|
||||
|
||||
let error = store
|
||||
.delete_threads(DeleteThreadsParams {
|
||||
thread_ids: vec![thread_id],
|
||||
})
|
||||
.await
|
||||
.expect_err("owned thread should block deletion before rollout exists");
|
||||
|
||||
assert!(matches!(error, ThreadStoreError::Conflict { .. }));
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn delete_threads_removes_rollout_with_unreadable_metadata() {
|
||||
let home = TempDir::new().expect("temp dir");
|
||||
let store = LocalThreadStore::new(test_config(home.path()), /*state_db*/ None);
|
||||
let uuid = Uuid::from_u128(311);
|
||||
let thread_id = ThreadId::from_string(&uuid.to_string()).expect("valid thread id");
|
||||
let rollout_path =
|
||||
write_session_file(home.path(), "2025-01-03T12-00-00", uuid).expect("session file");
|
||||
std::fs::write(&rollout_path, "{not json}\n").expect("damage rollout metadata");
|
||||
|
||||
store
|
||||
.delete_threads(DeleteThreadsParams {
|
||||
thread_ids: vec![thread_id],
|
||||
})
|
||||
.await
|
||||
.expect("delete rollout with unreadable metadata");
|
||||
|
||||
assert!(!rollout_path.exists());
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn delete_rollout_file_treats_vanished_path_as_already_deleted() {
|
||||
let home = TempDir::new().expect("temp dir");
|
||||
|
||||
@@ -34,6 +34,7 @@ use tokio::sync::OwnedMutexGuard;
|
||||
|
||||
use crate::AppendThreadItemsParams;
|
||||
use crate::ArchiveThreadParams;
|
||||
use crate::ArchiveThreadsParams;
|
||||
use crate::CreateThreadParams;
|
||||
use crate::DeleteThreadParams;
|
||||
use crate::DeleteThreadsParams;
|
||||
@@ -91,7 +92,7 @@ struct LiveRecorderEntry {
|
||||
// canonical SessionMeta is durable. Retain the mode captured when live persistence was opened
|
||||
// so missing SQLite rows can still be seeded.
|
||||
history_mode: ThreadHistoryMode,
|
||||
_writer_lock: Option<WriterLockGuard>,
|
||||
writer_lock: Option<WriterLockGuard>,
|
||||
}
|
||||
|
||||
#[derive(Default)]
|
||||
@@ -207,6 +208,42 @@ impl LocalThreadStore {
|
||||
Ok(())
|
||||
}
|
||||
|
||||
async fn acquire_paginated_writer_locks(
|
||||
&self,
|
||||
thread_ids: &[ThreadId],
|
||||
) -> ThreadStoreResult<Vec<WriterLockGuard>> {
|
||||
let mut writer_locks = Vec::new();
|
||||
for &thread_id in thread_ids {
|
||||
if self
|
||||
.live_recorders
|
||||
.lock()
|
||||
.await
|
||||
.get(&thread_id)
|
||||
.is_some_and(|entry| entry.writer_lock.is_some())
|
||||
{
|
||||
continue;
|
||||
}
|
||||
|
||||
// Only a readable legacy header proves no paginated writer can own this id. Missing
|
||||
// lazy rollouts and damaged headers must conservatively try the lock.
|
||||
let history_mode = match read_thread::resolve_rollout_path(
|
||||
self, thread_id, /*include_archived*/ true,
|
||||
)
|
||||
.await?
|
||||
{
|
||||
Some(rollout_path) => codex_rollout::read_session_meta_line(rollout_path.as_path())
|
||||
.await
|
||||
.ok()
|
||||
.map(|meta_line| meta_line.meta.history_mode),
|
||||
None => None,
|
||||
};
|
||||
if !matches!(history_mode, Some(ThreadHistoryMode::Legacy)) {
|
||||
writer_locks.push(self.writer_lock_coordinator.acquire(thread_id)?);
|
||||
}
|
||||
}
|
||||
Ok(writer_locks)
|
||||
}
|
||||
|
||||
async fn insert_live_recorder(
|
||||
&self,
|
||||
thread_id: ThreadId,
|
||||
@@ -222,7 +259,7 @@ impl LocalThreadStore {
|
||||
entry.insert(LiveRecorderEntry {
|
||||
recorder,
|
||||
history_mode,
|
||||
_writer_lock: writer_lock,
|
||||
writer_lock,
|
||||
});
|
||||
Ok(())
|
||||
}
|
||||
@@ -402,7 +439,24 @@ impl ThreadStore for LocalThreadStore {
|
||||
}
|
||||
|
||||
fn archive_thread(&self, params: ArchiveThreadParams) -> ThreadStoreFuture<'_, ()> {
|
||||
Box::pin(async move { archive_thread::archive_thread(self, params).await })
|
||||
Box::pin(async move {
|
||||
archive_thread::archive_threads(
|
||||
self,
|
||||
ArchiveThreadsParams {
|
||||
thread_ids: vec![params.thread_id],
|
||||
writer_lock_thread_ids: Vec::new(),
|
||||
},
|
||||
)
|
||||
.await
|
||||
.map(|_| ())
|
||||
})
|
||||
}
|
||||
|
||||
fn archive_threads(
|
||||
&self,
|
||||
params: ArchiveThreadsParams,
|
||||
) -> ThreadStoreFuture<'_, Vec<ThreadId>> {
|
||||
Box::pin(async move { archive_thread::archive_threads(self, params).await })
|
||||
}
|
||||
|
||||
fn unarchive_thread(&self, params: ArchiveThreadParams) -> ThreadStoreFuture<'_, StoredThread> {
|
||||
|
||||
@@ -6,6 +6,7 @@ use std::pin::Pin;
|
||||
|
||||
use crate::AppendThreadItemsParams;
|
||||
use crate::ArchiveThreadParams;
|
||||
use crate::ArchiveThreadsParams;
|
||||
use crate::CreateThreadParams;
|
||||
use crate::DeleteThreadParams;
|
||||
use crate::DeleteThreadsParams;
|
||||
@@ -167,6 +168,26 @@ pub trait ThreadStore: Any + Send + Sync {
|
||||
/// Archives a thread.
|
||||
fn archive_thread(&self, params: ArchiveThreadParams) -> ThreadStoreFuture<'_, ()>;
|
||||
|
||||
/// Archives threads in order, returning the successfully archived thread ids.
|
||||
///
|
||||
/// The first thread must archive successfully; later failures are best effort.
|
||||
fn archive_threads(
|
||||
&self,
|
||||
params: ArchiveThreadsParams,
|
||||
) -> ThreadStoreFuture<'_, Vec<ThreadId>> {
|
||||
Box::pin(async move {
|
||||
let mut archived_thread_ids = Vec::new();
|
||||
for thread_id in params.thread_ids {
|
||||
match self.archive_thread(ArchiveThreadParams { thread_id }).await {
|
||||
Ok(()) => archived_thread_ids.push(thread_id),
|
||||
Err(err) if archived_thread_ids.is_empty() => return Err(err),
|
||||
Err(err) => tracing::warn!("failed to archive thread {thread_id}: {err}"),
|
||||
}
|
||||
}
|
||||
Ok(archived_thread_ids)
|
||||
})
|
||||
}
|
||||
|
||||
/// Unarchives a thread and returns its updated metadata.
|
||||
fn unarchive_thread(&self, params: ArchiveThreadParams) -> ThreadStoreFuture<'_, StoredThread>;
|
||||
|
||||
|
||||
@@ -806,6 +806,17 @@ pub struct ArchiveThreadParams {
|
||||
pub thread_id: ThreadId,
|
||||
}
|
||||
|
||||
/// Parameters for archiving a set of threads as one store operation.
|
||||
#[derive(Clone, Debug, PartialEq, Eq, Serialize, Deserialize)]
|
||||
pub struct ArchiveThreadsParams {
|
||||
/// Thread ids to archive, in the order their persisted data should be moved.
|
||||
pub thread_ids: Vec<ThreadId>,
|
||||
/// Thread ids whose paginated writer ownership must be checked before archiving, including
|
||||
/// descendants whose rollout has not materialized yet.
|
||||
#[serde(default)]
|
||||
pub writer_lock_thread_ids: Vec<ThreadId>,
|
||||
}
|
||||
|
||||
/// Parameters for deleting a thread.
|
||||
#[derive(Clone, Debug, PartialEq, Eq, Serialize, Deserialize)]
|
||||
pub struct DeleteThreadParams {
|
||||
|
||||
Reference in New Issue
Block a user