grit: c_format: rework how we build the escaped string

Python 3 doesn't support string_escape on bytes.  It doesn't have an
official supported form for this currently either.  It could be done
ad-hoc (iterate over every byte and build up the string ourselves),
but it turns out the codecs module has an escape_encode helper that
does what we need with a bit of effort.

Also cleanup the associated unittest while we're here.  It was mixing
bytes & string data inline and expecting it to be interpreted as UTF8
encoding.  Python 3 didn't like this at all.  Make it a proper unicode
string and use \u sequences to get the codepoints we want.

Bug: 983071
Test: `./grit/test_suite_all.py` passes
Change-Id: I1c4de0e3ab5c129d4999d6685887a2c54bd7ab9c
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/1755388
Commit-Queue: Mike Frysinger <vapier@chromium.org>
Reviewed-by: Robert Flack <flackr@chromium.org>
Cr-Commit-Position: refs/heads/master@{#687323}

tools/grit/grit/format/c_format.py

@@ -7,6 +7,7 @@
 
 from __future__ import print_function
 
+import codecs
 import os
 import re
 
@@ -65,10 +66,16 @@ def _FormatMessage(item, lang):
   """Format a single <message> element."""
 
   message = item.ws_at_start + item.Translate(lang) + item.ws_at_end
-  # output message with non-ascii chars escaped as octal numbers
-  # C's grammar allows escaped hexadecimal numbers to be infinite,
-  # but octal is always of the form \OOO
-  message = message.encode('utf-8').encode('string_escape')
+  # Output message with non-ascii chars escaped as octal numbers C's grammar
+  # allows escaped hexadecimal numbers to be infinite, but octal is always of
+  # the form \OOO.  Python 3 doesn't support string-escape, so we have to jump
+  # through some hoops here via codecs.escape_encode.
+  # This basically does:
+  #   - message - the starting string
+  #   - message.encode(...) - convert to bytes
+  #   - codecs.escape_encode(...) - convert non-ASCII bytes to \x## escapes
+  #   - (...).decode() - convert bytes back to a string
+  message = codecs.escape_encode(message.encode('utf-8'))[0].decode('utf-8')
   # an escaped char is (\xHH)+ but only if the initial
   # backslash is not escaped.
   not_a_backslash = r"(^|[^\\])"  # beginning of line or a non-backslash char

tools/grit/grit/format/c_format_unittest.py

@@ -24,7 +24,7 @@ from grit.tool import build
 class CFormatUnittest(unittest.TestCase):
 
   def testMessages(self):
-    root = util.ParseGrdForUnittest("""
+    root = util.ParseGrdForUnittest(u"""
     <messages>
       <message name="IDS_QUESTIONS">Do you want to play questions?</message>
       <message name="IDS_QUOTES">
@@ -36,7 +36,7 @@ No.
 Statement.  Two all.  Game point.
 </message>
       <message name="IDS_NON_ASCII">
-         \xc3\xb5\\xc2\\xa4\\\xc2\xa4\\\\xc3\\xb5\xe4\xa4\xa4
+         \u00f5\\xc2\\xa4\\\u00a4\\\\xc3\\xb5\u4924
       </message>
     </messages>
       """)