Strip invalid utf-8 characters from mc.exe header output

build/win/message_compiler.py

 # Copyright 2015 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.
 
 # Runs the Microsoft Message Compiler (mc.exe). This Python adapter is for the
 # GN build, which can only run Python and not native binaries.
 #
 # Usage: message_compiler.py <environment_file> [<args to mc.exe>*]
 
+import os
 import subprocess
 import sys
 
-# Read the environment block from the file. This is stored in the format used
-# by CreateProcess. Drop last 2 NULs, one for list terminator, one for trailing
-# vs. separator.
-env_pairs = open(sys.argv[1]).read()[:-2].split('\0')
-env_dict = dict([item.split('=', 1) for item in env_pairs])
+def main():
+  # Read the environment block from the file. This is stored in the format used
+  # by CreateProcess. Drop last 2 NULs, one for list terminator, one for
+  # trailing vs. separator.
+  env_pairs = open(sys.argv[1]).read()[:-2].split('\0')
+  env_dict = dict([item.split('=', 1) for item in env_pairs])
 
-# mc writes to stderr, so this explicitly redirects to stdout and eats it.
-try:
-  # This needs shell=True to search the path in env_dict for the mc executable.
-  subprocess.check_output(["mc.exe"] + sys.argv[2:],
-                          env=env_dict,
-                          stderr=subprocess.STDOUT,
-                          shell=True)
-except subprocess.CalledProcessError as e:
-  print e.output
-  sys.exit(e.returncode)
+  # mc writes to stderr, so this explicitly redirects to stdout and eats it.
+  try:
+    # This needs shell=True to search the path in env_dict for the mc
+    # executable.
+    rest = sys.argv[2:]
+    subprocess.check_output(['mc.exe'] + rest,
+                            env=env_dict,
+                            stderr=subprocess.STDOUT,
+                            shell=True)
+    # We require all source code (in particular, the header generated here) to
+    # be UTF-8. jinja can output the intermediate .mc file in UTF-8 or UTF-16LE.
+    # However, mc.exe only supports Unicode via the -u flag, and it assumes when
+    # that is specified that the input is UTF-16LE (and errors out on UTF-8
+    # files, assuming they're ANSI). Even with -u specified and UTF16-LE input,
+    # it generates an ANSI header, and includes broken versions of the message
+    # text in the comment before the value. To work around this, for any invalid
+    # // comment lines, we simply drop the line in the header after building it.

[Nico, 2016/11/22 17:45:21]

(did you file an msconnect for getting support for this in mc.exe in the
future?)

+    header_dir = None
+    input_file = None
+    for i, arg in enumerate(rest):
+      if arg == '-h' and len(rest) > i + 1:
+        assert header_dir == None
+        header_dir = rest[i + 1]
+      elif arg.endswith('.mc') or arg.endswith('.man'):
+        assert input_file == None
+        input_file = arg
+    if header_dir:
+      header_file = os.path.join(
+          header_dir, os.path.splitext(os.path.basename(input_file))[0] + '.h')
+      header_contents = []
+      with open(header_file, 'rb') as f:
+        for line in f.readlines():
+          if line.startswith('//') and '?' in line:
+            continue
+          header_contents.append(line)
+      with open(header_file, 'wb') as f:
+        f.write(''.join(header_contents))
+  except subprocess.CalledProcessError as e:
+    print e.output
+    sys.exit(e.returncode)
+
+if __name__ == '__main__':
+  main()