From 717f37a6383c536d9414d902e3a3597364b8a930 Mon Sep 17 00:00:00 2001
From: Sprite <sprite@sprites.dev>
Date: Tue, 10 Feb 2026 04:58:38 +0000
Subject: [PATCH] Fix process hangs: add timeouts and fix atexit handler
 conflict

- Add timeout=30 to quickstart example command to prevent indefinite
  blocking on WebSocket hangs
- Change checkpoint HTTP client timeout from None to 300s to prevent
  indefinite blocking on slow API responses
- Fix atexit handler to skip cleanup if the event loop was already
  stopped by loop.py's handler, preventing creation of a new loop
  during shutdown
---
 examples/quickstart.py    | 2 +-
 src/sprites/checkpoint.py | 8 ++++----
 src/sprites/control.py    | 7 ++++++-
 3 files changed, 11 insertions(+), 6 deletions(-)

diff --git a/examples/quickstart.py b/examples/quickstart.py
index 4a8f745..33d33eb 100644
--- a/examples/quickstart.py
+++ b/examples/quickstart.py
@@ -13,7 +13,7 @@
 client.create_sprite(os.environ["SPRITE_NAME"])
 
 # step: Run Python
-output = client.sprite(os.environ["SPRITE_NAME"]).command("python", "-c", "print(2+2)").output()
+output = client.sprite(os.environ["SPRITE_NAME"]).command("python", "-c", "print(2+2)", timeout=30).output()
 print(output.decode(), end="")
 
 # step: Clean up
diff --git a/src/sprites/checkpoint.py b/src/sprites/checkpoint.py
index a6aaf71..fc2561c 100644
--- a/src/sprites/checkpoint.py
+++ b/src/sprites/checkpoint.py
@@ -217,9 +217,9 @@ def create_checkpoint(sprite: Sprite, comment: str = "") -> CheckpointStream:
     if comment:
         payload["comment"] = comment
 
-    # Use a separate client for streaming with no timeout
+    # Use a separate client for streaming with a generous timeout
     with httpx.Client(
-        timeout=None,
+        timeout=300.0,
         headers={"Authorization": f"Bearer {sprite.client.token}"},
     ) as client:
         try:
@@ -269,9 +269,9 @@ def restore_checkpoint(sprite: Sprite, checkpoint_id: str) -> RestoreStream:
     """
     url = f"{sprite.client.base_url}/v1/sprites/{sprite.name}/checkpoints/{checkpoint_id}/restore"
 
-    # Use a separate client for streaming with no timeout
+    # Use a separate client for streaming with a generous timeout
     with httpx.Client(
-        timeout=None,
+        timeout=300.0,
         headers={"Authorization": f"Bearer {sprite.client.token}"},
     ) as client:
         try:
diff --git a/src/sprites/control.py b/src/sprites/control.py
index a93646c..1e7418a 100644
--- a/src/sprites/control.py
+++ b/src/sprites/control.py
@@ -645,7 +645,12 @@ def _cleanup_on_exit() -> None:
         return
 
     try:
-        from sprites.loop import get_loop, stop_loop
+        from sprites.loop import get_loop, stop_loop, _loop
+
+        # Don't create a new loop just for cleanup — if the loop was already
+        # stopped by loop.py's atexit handler, just skip cleanup
+        if _loop is None:
+            return
 
         loop = get_loop()
         future = asyncio.run_coroutine_threadsafe(_close_all_pools(), loop)