[blinkpy] Block and wait for killed processes

In Executive.kill_process, drop WNOHANG from os.waitpid and use the
default blocking wait instead.

This makes sure we can reliably collect the defunct process, which fixes
a peculiar issue on macOS: when calling os.killpg on a defunct process
group, we get EACCES instead of failing silently, which makes a unit
test flaky on macOS.

Drive-by:
* Update the docstring to reflect the reality (we no longer ignore
  permission errors).

Bug: 1119341
Change-Id: I1efa79d8e610e67cede7d9492a82ba67abc8e0a7
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2459326Reviewed-by: Dirk Pranke <dpranke@google.com>
Commit-Queue: Robert Ma <robertma@chromium.org>
Cr-Commit-Position: refs/heads/master@{#815265}

third_party/blink/tools/blinkpy/common/system/executive.py

@@ -110,7 +110,7 @@ class Executive(object):
 
         if kill_tree is True, the whole process group will be killed.
 
-        Will fail silently if pid does not exist or insufficient permissions.
+        Will fail silently if pid does not exist.
         """
         if sys.platform == 'win32':
             # Workaround for race condition that occurs when the browser is
@@ -136,7 +136,9 @@ class Executive(object):
                 os.killpg(os.getpgid(pid), signal.SIGKILL)
             else:
                 os.kill(pid, signal.SIGKILL)
-            os.waitpid(pid, os.WNOHANG)
+            # At this point if no exception has been raised, the kill has
+            # succeeded, so we can safely use a blocking wait.
+            os.waitpid(pid, 0)
         except OSError as error:
             if error.errno == errno.ESRCH:
                 _log.debug("PID %s does not exist.", pid)