Do not redirect stderr of Apache process

Tools/Scripts/webkitpy/layout_tests/servers/apache_http.py

 # Copyright (C) 2011 Google Inc. All rights reserved.
 #
 # Redistribution and use in source and binary forms, with or without
 # modification, are permitted provided that the following conditions are
 # met:
 #
 #     * Redistributions of source code must retain the above copyright
 # notice, this list of conditions and the following disclaimer.
 #     * Redistributions in binary form must reproduce the above
 # copyright notice, this list of conditions and the following disclaimer
@@ skipping 108 matching lines @@
                 start_cmd += ['-c', 'Alias %s "%s"' % (alias, path),
                         # Disable CGI handler for additional dirs.
                         '-c', '<Location %s>' % alias,
                         '-c', 'RemoveHandler .cgi .pl',
                         '-c', '</Location>']
 
         self._start_cmd = start_cmd
 
     def _spawn_process(self):
         _log.debug('Starting %s server, cmd="%s"' % (self._name, str(self._start_cmd)))
-        self._process = self._executive.popen(self._start_cmd, stderr=self._executive.PIPE)
-        if self._process.returncode is not None:
-            retval = self._process.returncode
-            err = self._process.stderr.read()
-            if retval or len(err):
-                raise server_base.ServerError('Failed to start %s: %s' % (self._name, err))
+        self._process = self._executive.popen(self._start_cmd)
+        retval = self._process.returncode
+        if retval:
+            raise server_base.ServerError('Failed to start %s: %s' % (self._name, retval))

[Dirk Pranke, 2014/11/11 16:31:53]

This change worries me. I fairly strongly recall that there were cases where
Apache did actually fail to start up properly but didn't return a nonzero error
code.

However, I can see how opening a pipe and never reading from it might cause
problems also.

I guess we can try this for now. If we see issues, we'll probably need to change
the code to launch a thread to read from the pipe to ensure that we don't get
the blocking you've seen.

[Kunihiko Sakamoto, 2014/11/12 01:02:39]

Got it. Let's try this for now.

 
         # For some reason apache isn't guaranteed to have created the pid file before
         # the process exits, so we wait a little while longer.
         if not self._wait_for_action(lambda: self._filesystem.exists(self._pid_file)):
             self._log_errors_from_subprocess()
             raise server_base.ServerError('Failed to start %s: no pid file found' % self._name)
 
         return int(self._filesystem.read_text_file(self._pid_file))
 
     def stop(self):
@@ skipping 18 matching lines @@
         retval = proc.returncode
         err = proc.stderr.read()
         if retval or len(err):
             raise server_base.ServerError('Failed to stop %s: %s' % (self._name, err))
 
         # For some reason apache isn't guaranteed to have actually stopped after
         # the stop command returns, so we wait a little while longer for the
         # pid file to be removed.
         if not self._wait_for_action(lambda: not self._filesystem.exists(self._pid_file)):
             raise server_base.ServerError('Failed to stop %s: pid file still exists' % self._name)