Reland "[test] Print hanging tests on linux on test-runner termination"

This is a reland of 3fc96631

The private method on_event in timeout proc is now renamed to be truly
private.

Original change's description:
> [test] Print hanging tests on linux on test-runner termination
>
> This will print the list of processes still running before and after
> joining workers during termination. This will help debugging hanging
> tests during flake-bisect or with num-fuzzer, which both terminate
> on total timeout and currently still sometimes hang without printing
> processes.
>
> Bug: v8:8292
> Change-Id: I124b65fa35b8d7a6aa198fcf50f2c20df94dc51a
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1735312
> Reviewed-by: Tamer Tas <tmrts@chromium.org>
> Commit-Queue: Michael Achenbach <machenbach@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#63065}

Bug: v8:8292
Change-Id: Ibad1172666d6f4d2c07884a54edfe9d6499b57fe
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1735318Reviewed-by: Tamer Tas <tmrts@chromium.org>
Commit-Queue: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#63072}

tools/testrunner/local/pool.py

@@ -115,7 +115,15 @@ class Pool():
   # Necessary to not overflow the queue's pipe if a keyboard interrupt happens.
   BUFFER_FACTOR = 4
 
-  def __init__(self, num_workers, heartbeat_timeout=1):
+  def __init__(self, num_workers, heartbeat_timeout=1, notify_fun=None):
+    """
+    Args:
+      num_workers: Number of worker processes to run in parallel.
+      heartbeat_timeout: Timeout in seconds for waiting for results. Each time
+          the timeout is reached, a heartbeat is signalled and timeout is reset.
+      notify_fun: Callable called to signale some events like termination. The
+          event name is passed as string.
+    """
     self.num_workers = num_workers
     self.processes = []
     self.terminated = False
@@ -130,6 +138,7 @@ class Pool():
     # work_queue.
     self.processing_count = 0
     self.heartbeat_timeout = heartbeat_timeout
+    self.notify = notify_fun or (lambda x: x)
 
     # Disable sigint and sigterm to prevent subprocesses from capturing the
     # signals.
@@ -261,11 +270,13 @@ class Pool():
       for p in self.processes:
         os.kill(p.pid, signal.SIGTERM)
 
+    self.notify("Joining workers")
     for p in self.processes:
       p.join()
 
     # Drain the queues to prevent stderr chatter when queues are garbage
     # collected.
+    self.notify("Draining queues")
     try:
       while True: self.work_queue.get(False)
     except:

tools/testrunner/testproc/base.py

@@ -109,6 +109,19 @@ class TestProc(object):
 
   ### Communication
 
+  def notify_previous(self, event):
+    self._on_event(event)
+    if self._prev_proc:
+      self._prev_proc.notify_previous(event)
+
+  def _on_event(self, event):
+    """Called when processors to the right signal events, e.g. termination.
+
+    Args:
+      event: A text describing the signalled event.
+    """
+    pass
+
   def _send_test(self, test):
     """Helper method for sending test to the next processor."""
     return self._next_proc.next_test(test)
@@ -120,7 +133,6 @@ class TestProc(object):
     self._prev_proc.result_for(test, result)
 
 
-
 class TestProcObserver(TestProc):
   """Processor used for observing the data."""
   def __init__(self):

tools/testrunner/testproc/execution.py

@@ -45,7 +45,7 @@ class ExecutionProc(base.TestProc):
 
   def __init__(self, jobs, outproc_factory=None):
     super(ExecutionProc, self).__init__()
-    self._pool = pool.Pool(jobs)
+    self._pool = pool.Pool(jobs, notify_fun=self.notify_previous)
     self._outproc_factory = outproc_factory or (lambda t: t.output_proc)
     self._tests = {}
 

tools/testrunner/testproc/progress.py

@@ -149,6 +149,10 @@ class VerboseProgressIndicator(SimpleProgressIndicator):
       self._print('Still working...')
       self._print_processes_linux()
 
+  def _on_event(self, event):
+    self._print(event)
+    self._print_processes_linux()
+
 
 class DotsProgressIndicator(SimpleProgressIndicator):
   def __init__(self):

tools/testrunner/testproc/timeout.py

@@ -14,15 +14,15 @@ class TimeoutProc(base.TestProcObserver):
     self._start = time.time()
 
   def _on_next_test(self, test):
-    self._on_event()
+    self.__on_event()
 
   def _on_result_for(self, test, result):
-    self._on_event()
+    self.__on_event()
 
   def _on_heartbeat(self):
-    self._on_event()
+    self.__on_event()
 
-  def _on_event(self):
+  def __on_event(self):
     if not self.is_stopped:
       if time.time() - self._start > self._duration_sec:
         print('>>> Total timeout reached.')