From a950f978d3c3aea3eb3649333039b2976bfe2b13 Mon Sep 17 00:00:00 2001
From: biondizzle <biondizzle@gmail.com>
Date: Fri, 22 May 2026 17:08:12 +0000
Subject: [PATCH] run_test.sh: SIGKILL all children of screen session on
 cleanup

Deadlocked GPU processes ignore SIGHUP from screen -X quit.
Now kills the entire process group with SIGKILL, plus a catch-all
pkill for any python test_ processes.
---
 tests/run_test.sh | 16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)

diff --git a/tests/run_test.sh b/tests/run_test.sh
index 55d694f7..964f3cee 100755
--- a/tests/run_test.sh
+++ b/tests/run_test.sh
@@ -12,13 +12,23 @@ if [ -z "$TEST_FILE" ]; then
 fi
 
 # --- CLEANUP ---
+# Kill the screen AND every process inside it (handles deadlocked GPU procs)
 if screen -list | grep -q kernel-test; then
-    echo "Killing existing kernel-test screen..."
+    echo "Killing existing kernel-test screen and children..."
+    # Find PIDs belonging to the screen session and SIGKILL them
+    session_pid=$(screen -ls | grep kernel-test | grep -o '[0-9]*' | head -1)
+    if [ -n "$session_pid" ]; then
+        # Kill the entire process group (screen's children)
+        pkill -9 -P "$session_pid" 2>/dev/null || true
+        # Also nuke any python test process just in case
+        pkill -9 -f "python.*test_" 2>/dev/null || true
+    fi
     screen -S kernel-test -X quit 2>/dev/null || true
     sleep 2
-    pkill -f "python.*test_fmha" 2>/dev/null || true
-    sleep 1
 fi
+# Belt and suspenders: kill any leftover python test processes
+pkill -9 -f "python.*test_" 2>/dev/null || true
+sleep 1
 
 rm -f /tmp/kernel-test.log