From 0d966c7a7d62f41b5d200fd7fd6e74cd2557dfa2 Mon Sep 17 00:00:00 2001
From: Tele Oyekunle <toyekunle@quora.com>
Date: Mon, 30 Mar 2026 21:20:33 +0000
Subject: [PATCH] Fix _sock_recv infinite loop when StatusDB TCP connection
 drops

When using --reruns with pytest-xdist, every test makes two blocking TCP
calls to the StatusDB server (get_test_failures and set_test_reruns in
pytest_runtest_protocol). If the server-side connection drops, _sock_recv
enters an infinite loop because recv(1) returns b'' (empty bytes) on a
closed socket, but the code only checks for the newline delimiter:

    while True:
        b = conn.recv(1)
        if b == self.delim:  # b'' != b'\n' -> never breaks
            break
        buf += b

This causes xdist workers to hang indefinitely at ~90% CPU, appearing
stuck on a test that never completes. The hang persists until the process
is manually killed.

The fix adds a check for empty bytes from recv(1) and raises
ConnectionError, which surfaces as an INTERNALERROR that xdist handles
by replacing the worker.
---
 CHANGES.rst                        |  4 ++++
 src/pytest_rerunfailures.py        |  2 ++
 tests/test_pytest_rerunfailures.py | 21 ++++++++++++++++++++-
 3 files changed, 26 insertions(+), 1 deletion(-)

diff --git a/CHANGES.rst b/CHANGES.rst
index 8638e1c..43dd973 100644
--- a/CHANGES.rst
+++ b/CHANGES.rst
@@ -7,6 +7,10 @@ Changelog
 - Fix missing teardown for session and module scoped fixtures when fixture teardown fails.
   Fixes `#314 <https://github.com/pytest-dev/pytest-rerunfailures/issues/314>`_.
 
+- Fix ``_sock_recv`` infinite loop when the StatusDB TCP connection drops.
+  ``recv(1)`` returning empty bytes (closed connection) was not handled,
+  causing workers to spin at 100% CPU indefinitely during xdist runs.
+
 
 16.1 (2025-10-10)
 -----------------
diff --git a/src/pytest_rerunfailures.py b/src/pytest_rerunfailures.py
index 63223ae..9e0ffb0 100644
--- a/src/pytest_rerunfailures.py
+++ b/src/pytest_rerunfailures.py
@@ -436,6 +436,8 @@ def _sock_recv(self, conn) -> str:
         buf = b""
         while True:
             b = conn.recv(1)
+            if not b:
+                raise ConnectionError("StatusDB connection closed unexpectedly")
             if b == self.delim:
                 break
             buf += b
diff --git a/tests/test_pytest_rerunfailures.py b/tests/test_pytest_rerunfailures.py
index 7156bc9..088aea5 100644
--- a/tests/test_pytest_rerunfailures.py
+++ b/tests/test_pytest_rerunfailures.py
@@ -1,10 +1,11 @@
 import random
+import socket
 import time
 from unittest import mock
 
 import pytest
 
-from pytest_rerunfailures import HAS_PYTEST_HANDLECRASHITEM
+from pytest_rerunfailures import HAS_PYTEST_HANDLECRASHITEM, SocketDB
 
 pytest_plugins = "pytester"
 
@@ -1407,3 +1408,21 @@ def test_fail():
 
     result = testdir.runpytest("--force-reruns", "3")
     assert_outcomes(result, passed=0, failed=1, rerun=3)
+
+
+def test_sock_recv_raises_on_closed_connection():
+    """_sock_recv should raise ConnectionError when recv returns empty bytes
+    (connection closed), not loop forever.
+
+    Previously, _sock_recv had no check for empty bytes from recv(1), causing
+    an infinite CPU-spinning loop when the server-side connection dropped.
+    This manifested as indefinite hangs during xdist test runs.
+    """
+    s1, s2 = socket.socketpair()
+    s2.close()  # Close one end — recv on s1 will return b""
+
+    db = SocketDB()
+    with pytest.raises(ConnectionError, match="closed unexpectedly"):
+        db._sock_recv(s1)
+
+    s1.close()