run-task: wait on child processes when we're pid 1.

src/taskgraph/run-task/run-task

@@ -28,6 +28,7 @@ import socket
 import stat
 import subprocess
 import sys
+import threading
 import time
 import urllib.error
 import urllib.request
@@ -121,6 +122,17 @@ def print_line(prefix, m):
     sys.stdout.buffer.flush()
 
 
+def reap_zombies(main_subprocess):
+    """Wait for main_subprocess to exit, while awaiting any other child processes"""
+    while main_subprocess.poll() is None:
+        with main_subprocess._waitpid_lock:
+            if main_subprocess.returncode is not None:
+                return
+            pid, status = os.wait()
+            if pid == main_subprocess.pid:
+                main_subprocess._handle_exitstatus(status)
+
+
 def _call_windows_retry(func, args=(), retry_max=5, retry_delay=0.5):
     """
     It's possible to see spurious errors on Windows due to various things
@@ -278,6 +290,13 @@ def run_command(prefix, args, *, extra_env=None, cwd=None):
 
     stdout = io.TextIOWrapper(p.stdout, encoding="latin1")
 
+    if os.getpid() == 1:
+        # in docker we're init, so we get to adopt unawaited zombies
+        reaper_thread = threading.Thread(target=reap_zombies, args=(p,))
+        reaper_thread.start()
+    else:
+        reaper_thread = None
+
     while True:
         data = stdout.readline().encode("latin1")
 
@@ -286,6 +305,8 @@ def run_command(prefix, args, *, extra_env=None, cwd=None):
 
         print_line(prefix, data)
 
+    if reaper_thread:
+        reaper_thread.join()
     return p.wait()