Avoid replicator wait during export close

LEAPP export does not use Replicator workflows, so avoid waiting for Replicator during SimulationApp shutdown. This keeps teardown from blocking on unrelated queued Replicator work and keeps traceback diagnostics around close for future CI timeouts.

Author: AntoineRichard

scripts/reinforcement_learning/leapp/rsl_rl/export.py

@@ -11,6 +11,7 @@
 
 import argparse
 import contextlib
+import faulthandler
 import importlib.metadata as metadata
 import os
 import sys
@@ -40,17 +41,22 @@
 
 _TIMING_PREFIX = "[LEAPP_EXPORT_TIMING]"
 _PROCESS_START = time.perf_counter()
+_CLOSE_HANG_TRACEBACK_TIMEOUT = 120
+
+
+def _emit_timing(message: str) -> None:
+    """Print timing diagnostics to the original stderr stream."""
+    print(message, file=sys.__stderr__ or sys.stderr, flush=True)
 
 
 @contextlib.contextmanager
 def _timed_phase(task_name: str | None, phase: str):
     """Print timing information for a single export phase."""
     start = time.perf_counter()
     task_label = task_name or "<unknown>"
-    print(
+    _emit_timing(
         f"{_TIMING_PREFIX} task={task_label} phase={phase} status=start "
-        f"pid={os.getpid()} total_elapsed={start - _PROCESS_START:.2f}s",
-        flush=True,
+        f"pid={os.getpid()} total_elapsed={start - _PROCESS_START:.2f}s"
     )
     status = "done"
     try:
@@ -60,13 +66,26 @@ def _timed_phase(task_name: str | None, phase: str):
         raise
     finally:
         end = time.perf_counter()
-        print(
+        _emit_timing(
             f"{_TIMING_PREFIX} task={task_label} phase={phase} status={status} "
-            f"phase_elapsed={end - start:.2f}s total_elapsed={end - _PROCESS_START:.2f}s",
-            flush=True,
+            f"phase_elapsed={end - start:.2f}s total_elapsed={end - _PROCESS_START:.2f}s"
         )
 
 
+def _close_simulation_app() -> None:
+    """Close Isaac Sim without waiting for Replicator workflows."""
+    traceback_stream = sys.__stderr__ or sys.stderr
+    faulthandler.dump_traceback_later(
+        _CLOSE_HANG_TRACEBACK_TIMEOUT,
+        repeat=True,
+        file=traceback_stream,
+    )
+    try:
+        simulation_app.close(wait_for_replicator=False)
+    finally:
+        faulthandler.cancel_dump_traceback_later()
+
+
 parser = argparse.ArgumentParser(description="Train an RL agent with RSL-RL.")
 parser.add_argument(
     "--disable_fabric", action="store_true", default=False, help="Disable fabric and use USD I/O operations."
@@ -125,7 +144,7 @@ def _timed_phase(task_name: str | None, phase: str):
 
 installed_version = metadata.version("rsl-rl-lib")
 
-print(f"[INFO] LEAPP version: {metadata.version('leapp')}", flush=True)
+_emit_timing(f"[INFO] LEAPP version: {metadata.version('leapp')}")
 with _timed_phase(args_cli.task, "app_launcher"):
     app_launcher = AppLauncher(args_cli)
     simulation_app = app_launcher.app
@@ -202,7 +221,7 @@ def main(env_cfg: ManagerBasedRLEnvCfg, agent_cfg: RslRlBaseRunnerCfg):
     task_name = args_cli.task.split(":")[-1]
     train_task_name = task_name.replace("-Play", "")
     task_start = time.perf_counter()
-    print(f"{_TIMING_PREFIX} task={task_name} phase=main status=start", flush=True)
+    _emit_timing(f"{_TIMING_PREFIX} task={task_name} phase=main status=start")
 
     with _timed_phase(task_name, "agent_config"):
         agent_cfg: RslRlBaseRunnerCfg = cli_args.update_rsl_rl_cfg(agent_cfg, args_cli)
@@ -317,10 +336,9 @@ def main(env_cfg: ManagerBasedRLEnvCfg, agent_cfg: RslRlBaseRunnerCfg):
     with _timed_phase(task_name, "env_close"):
         env.close()
 
-    print(
+    _emit_timing(
         f"{_TIMING_PREFIX} task={task_name} phase=main status=done "
-        f"phase_elapsed={time.perf_counter() - task_start:.2f}s",
-        flush=True,
+        f"phase_elapsed={time.perf_counter() - task_start:.2f}s"
     )
 
 
@@ -329,4 +347,4 @@ def main(env_cfg: ManagerBasedRLEnvCfg, agent_cfg: RslRlBaseRunnerCfg):
         main()
     finally:
         with _timed_phase(args_cli.task, "simulation_app_close"):
-            simulation_app.close()
+            _close_simulation_app()