Add a comment explaining why readline is the only way we can avoid buffering

tools/valgrind/common.py

 # Copyright (c) 2012 The Chromium Authors. All rights reserved.
 # Use of this source code is governed by a BSD-style license that can be
 # found in the LICENSE file.
 
 import logging
 import platform
 import os
 import signal
 import subprocess
 import sys
@@ skipping 37 matching lines @@
   p = subprocess.Popen(proc, universal_newlines=True,
                        bufsize=0,  # unbuffered
                        stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
 
   logging.info("started subprocess")
 
   did_timeout = False
   if timeout > 0:
     wait_until = time.time() + timeout
   while p.poll() is None and not did_timeout:
+    # Have to use readline rather than readlines() or "for line in p.stdout:",
+    # otherwise we get buffered even with bufsize=0.
     line = p.stdout.readline()
     while line and not did_timeout:
       sys.stdout.write(line)
       sys.stdout.flush()
       line = p.stdout.readline()
       if timeout > 0:
         did_timeout = time.time() > wait_until
 
   if did_timeout:
     logging.info("process timed out")
@@ skipping 173 matching lines @@
     return False
 
   print "-----------------------------------------------------"
   print "Suppressions used:"
   print "  count name"
   for (name, count) in sorted(suppcounts.items(), key=lambda (k,v): (v,k)):
     print "%7d %s" % (count, name)
   print "-----------------------------------------------------"
   sys.stdout.flush()
   return True