Issue1830

I am using Jython 2.5.1 and find a strange behavior for assert and print. Pls see the following lines.

1.  assert False == True, u"\u4eee\u60f3\u30a4\u30e1\u30fc\u30b8\u300c"

   Traceback (most recent call last):
    File "<stdin>", line 1, in <module>
   Error in sys.excepthook:
   UnicodeEncodeError: 'ascii' codec can't encode characters in position 0-6: ordinal not in range(128)

2. print u"\u4eee\u60f3\u30a4\u30e1\u30fc\u30b8\u300c"

   output:  仮想イメージ「

3. assert False == True, u"\u4eee\u60f3\u30a4\u30e1\u30fc\u30b8\u300c".encode('utf8')
   Traceback (most recent call last):
   File "<stdin>", line 1, in <module>
  AssertionError: 仮想イメージ「

2 and 3 can show the message correctly, but 1 cannot. Do assert and print have different mechanism to display the message?
BTW, my sys.stdout.encoding is UTF-8.

Confirmed on trunk. In fact the trouble is that exceptions can't handle unicode in the message.

Interestingly, this does not work on CPython 2.7, but does on CPython 3.4.

Should be an easy fix.

Target beta 4

For the record: in C-Python 2.7.6 under Ubuntu 14.04 with OpenJDK 1.7.0_55 you get an AssertionError without a message:

  >>> assert False, u"\u4eee\u60f3\u30a4\u30e1\u30fc\u30b8\u300c"
  Traceback (most recent call last):
    File "<stdin>", line 1, in <module>
  AssertionError

In Python 3.4.0 you get an AssertionError with unicode characters.

With Jython 2.7.2a1 I get something like the opposite behaviour:

>>> sys.stderr.encoding, sys.stderr.errors
('ms936', 'backslashreplace')
>>> assert False, u"\u4eee\u60f3\u30a4\u30e1\u30fc\u30b8\u300c"
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AssertionError: 仮想イメージ「
>>> print u"\u4eee\u60f3\u30a4\u30e1\u30fc\u30b8\u300c"
仮想イメージ「
>>> assert False, u"\u4eee\u60f3\u30a4\u30e1\u30fc\u30b8\u300c".encode('utf8')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AssertionError: \xe4\xbb\xae\xe6\x83\xb3\xe3\x82¤\xe3\x83\xa1\xe3\x83\xbc\xe3\x82\xb8\xe3\x80\x8c

I wouldn't expect a good result from emitting utf-8 bytes onto an ms936 console, but this is not much better:
>>> assert False, u"\u4eee\u60f3\u30a4\u30e1\u30fc\u30b8\u300c".encode('ms936')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AssertionError: \x81\xa2\xcf\xeb\xa5¤\xa5á\xa9`\xa5\xb8\xa1\xb8

Since we can cope with:
>>> print u"\u4eee\u60f3\u30a4\u30e1\u30fc\u30b8\u300c".encode('ms936')
仮想イメージ「
I think error message generation is defending itself too well against the decoding problems that replace the message you wanted with one about codecs.