Use trace_child_process.py on Windows and save information about the trace in json.

tools/isolate/trace_child_process.py

Index: tools/isolate/trace_child_process.py
diff --git a/tools/isolate/trace_child_process.py b/tools/isolate/trace_child_process.py
index 3a8367bbbaa2462fcf8cebb9b400b68babb8b236..07d9915c50c813656f812b42054d2fd1d60dead4 100755
--- a/tools/isolate/trace_child_process.py
+++ b/tools/isolate/trace_child_process.py
@@ -3,30 +3,32 @@
 # Use of this source code is governed by a BSD-style license that can be
 # found in the LICENSE file.
 
-"""Waits for the go signal and replaces itself with the command to be run.
+"""Executes with the command to be run and optionally waits for the go signal.
 
 Not meant to be used directly, only meant to be used by trace_inputs.py.
+
+This script is used by the tracer as a signal for the log parser that the child
+process is the process we care about. It is especially important on kernel based
+tracer because we want it to trace the relevant process tree.
 """
 
-import os
 import subprocess
 import sys
 
 
 def main():
-  signal = 'Go!'
-  value = sys.stdin.read(len(signal))
-  assert value == signal
-  sys.stdin.close()
-  # Replace the executable with an absolute path to make it easier to grok.
   cmd = sys.argv[1:]
-  cmd[0] = os.path.abspath(cmd[0])
-  if cmd[0].endswith('.py'):
-    cmd.insert(0, sys.executable)
-  p = subprocess.Popen(cmd)
-  #print 'Child pid: %d' % p.pid
-  p.wait()
-  return p.returncode
+  if sys.argv[1] == '--wait':
+    cmd = cmd[1:]
+    signal = 'Go!'
+    value = sys.stdin.read(len(signal))
+    assert value == signal
+    sys.stdin.close()
+
+  # The reason os.execve() is not used is that we don't want the modules
+  # imported here to influence the executable being traced, so we need a fresh
+  # pid and need to fork.
+  return subprocess.call(cmd)
 
 
 if __name__ == '__main__':