wboayue / rust-ibapi / 25999805501

92%

push

github

fix(transport): clear stale metadata on reconnect; stop clobbering sync is_connected (#580)

Port of v2-stable PRs #576 + #579 to main.

Async: connection_metadata() was observably stale during a reconnect
window — callers saw the prior session's server_version, next_order_id,
managed_accounts, connection_time until the new handshake completed. Add
reset_connection_metadata() and call from reconnect() after socket
replacement, before establish_connection(). server_version_cache is
zeroed at the same time.

Sync: same metadata bug as async — same fix.

Sync transport: reset() stored connected=false at its tail. The dispatch
loop calls reset() *immediately after* restoring connected=true on
successful reconnect, so the flag was clobbered back to false within
microseconds. is_connected() was therefore permanently false after the
first network blip; examples/sync/connection_monitoring.rs would never
print "Connection restored". reset()'s job is channels, not lifecycle
state — drop the store.

Three regression tests, all using the existing MemoryStream / MockSocket
test fixtures (no real-TCP harness needed):

- connection::async::reconnect_clears_metadata_while_waiting_for_handshake
- connection::sync::reconnect_clears_metadata_while_waiting_for_handshake
- transport::sync::test_is_connected_stays_true_after_reconnect

The first two spawn reconnect in a task/thread with no inbound handshake
bytes queued, wait for the new handshake's wire bytes to appear (signals
reset_connection_metadata has run), and assert cleared state — then
release the handshake and assert restored state.

17 of 17 new or added lines in 2 files covered. (100.0%)

18956 of 20602 relevant lines covered (92.01%)

30.11 hits per line