Here is an issue that caused Kid templates (used by
Turbogears) to malfunction in Python 2.4.3c1.

The problem shows up with the following code:

class s1:
    def __repr__(self):
        return '\\n'

class s2:
    def __repr__(self):
        return u'\\n'

print repr(s1()), repr(s2())

I get the following results:

Python 2.3.5: \n \n
Python 2.4.2: \n \n
Python 2.4.3c1: \n \\n 

In the output for Python 2.4.3c1, the backslash in the
representation of class2 appears doubled. This did not
happen in earlier Python versions and seems to be a bug.

My vague guess is that the issue may have crept in with
an attempted fix of Bug #1379994.

-- Christoph

Logged In: YES 
user_id=29957

Confirmed - it's also broken in the trunk, and backing out
the patch for http://www.python.org/sf/1379994 (r41728)
fixes the problem. Perky, you checked this in - can you look
at this soon, please? I don't want to release 2.4.3 until
it's fixed, but I also want to get 2.4.3 out this week.

Thanks for the bug report!

Logged In: YES 
user_id=33168

Attached a patch for the test case to be added with fix.

Logged In: YES 
user_id=55188

Looking the C code, unicode_repr is doing correct.
But the inconsistency came from PyObject_Repr.
This change made it which is intended:

------------------------------------------------------------------------
r16198 | effbot | 2000-07-09 02:43:32 +0900 (일, 09  7 2000)
| 6 lines


- changed __repr__ to use "unicode escape" encoding for unicode
  strings, instead of the default encoding.
  (see "minidom" thread for discussion, and also patch #100706)

Logged In: YES 
user_id=55188

Found it!:
http://mail.python.org/pipermail/python-dev/2000-July/005353.html
But their intention had never applied before 2.4.3.
What problem would be if we change PyObject_Repr to use the
default encoding not unicode-escape? (revert r16198)

Logged In: YES 
user_id=29957

I'm confused how a checkin from 5+ years ago broke a change
from 3 months ago?

Or am I misunderstanding you?

Logged In: YES 
user_id=55188

Because unicode-escape codec didn't escape \,
PyObject_Repr(u'\\') bypassed backslashes.  But Martin and
Fredrik made PyObject_Repr to use unicode-escape codec for
unicode repr-returns 5 years ago.  So by fixing
unicode-escape codec, their intention could be applied for
the first time, 3 months ago.

Logged In: YES 
user_id=33168

We need to retain the old behaviour, but also fix the bug. 
How can we do that?  

Logged In: YES 
user_id=29957

Ok. After talking to perky, I reverted the fix for 1379994
on the release24-maint branch, and reverted /F's ancient
change on the trunk. This seemed the best combination of
practicality and purity. Fix will be in 2.4.3 final. 

Thanks for the bug report. Man, unicode and repr is a twisty
ball of horrors.