running 1 test
PUMP_IPC_READER: spawned
=== P1b ACK-DEADLOCK GATE: flood_n=2000 flood_sent=false broker_alive(net_status)=true sessions_answered=true concurrent_attach_subscribed=false attach_received_output=false (fixed = flood_sent + broker_alive + sessions_answered true; pre-fix: the N>pipe-buffer applied-ack stream fills the return direction → the per-conn handler deadlocks → sessions_answered=false, and even flood_sent stalls as the operator's own writes back up) ===

thread 'input_flood_through_serve_attach_does_not_deadlock_broker' (213480) panicked at crates\spt-daemon\tests\input_ack_deadlock.rs:532:5:
all 2000 flooded input records must SEND through serve_attach without stalling ΓÇö pre-fix the applied-ack stream fills the target brainΓåöbroker return direction and the resulting full-duplex deadlock backs up the operator's own writes (REQ-HAZARD-INPUT-ACK-BACKPRESSURE). Reverting the fix (serve_attach's controller arm to send_effect with ack=true, or dropping the dispatch_input `if req.ack` gate) re-introduces the deadlock and flips this RED.
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
test input_flood_through_serve_attach_does_not_deadlock_broker ... FAILED

failures:

failures:
    input_flood_through_serve_attach_does_not_deadlock_broker

test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 51.98s