Hmm, it actually passed on my Linux machine. Perhaps related to system version of Python?

RE in question:

_XML_BAD_CHAR_REGEX = lazy_re.compile(u'[^\u0009\u000A\u000D'
                                      u'\u0020-\uD7FF\uE000-\uFFFD'
                                      u'\U00010000-\U0010FFFF]')

Also fails for me on Windows.

Windows has Python 2.7.3, Mac has Python 2.7.10 and Linux has Python 2.7.6.

Very strange it would related to Python version since it would mean it was something that errors in 2.7.6 but not in 2.7.3 or 2.7.10...

So it's this part that's causing the issue:

re.compile(u'[\U00010000-\U0010FFFF]')

Looks like it's related to this:

http://stackoverflow.com/questions/31603075/how-can-i-represent-this-regex-to-not-get-a-bad-character-range-error

So it depends if its a "wide" or "narrow" Python build.

Looks it was added by newt@ here:

https://codereview.chromium.org/1440313003

... who no longer appears to work on Chrome.

However, this error isn't hit presumably during Chrome build process.

Is a different version of Python used there than system Python. Nico, do you know? (Or is this code not executed during Chrome build?)

I don't think we run grit tests as part of the build, only on presubmit.

I think we don' tuse system python but depot_tools python on windows (?)

This repros for me with using depot_tools\python.bat

.. on Windows which is Python version 2.7.6.

Yeah, I'd guess newt ran the presubmit on linux :-/

I guess we just don't run the XMB codepath during the build process then? I didn't know what it was, but looking a bit - I guess it's the format used to deliver to translators? Maybe that workflow is only done on Linux?

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/56dfab6710187ab2f41a917c1ce99529836bd6e7

commit 56dfab6710187ab2f41a917c1ce99529836bd6e7
Author: asvitkine <asvitkine@chromium.org>
Date: Fri Feb 17 20:23:30 2017

Fix xmb_unittest.py in GRIT.

It was broken on systems with a "narrow" build of Python, see:
http://stackoverflow.com/questions/31603075/how-can-i-represent-this-regex-to-not-get-a-bad-character-range-error

It turns out the regex was not needed at all - as invalid characters do not even
reach that point. This removes the check and adds a unit test to verify that
the tool produces an error when characters outside the valid range are specified.

BUG= 693086 

Review-Url: https://codereview.chromium.org/2697383002
Cr-Commit-Position: refs/heads/master@{#451370}

[modify] https://crrev.com/56dfab6710187ab2f41a917c1ce99529836bd6e7/tools/grit/grit/tool/xmb.py
[modify] https://crrev.com/56dfab6710187ab2f41a917c1ce99529836bd6e7/tools/grit/grit/tool/xmb_unittest.py