Browser requests w/ non-ASCII path trigger an error in the Python logger

[Emmanuel Blot]

When a request to browse a directory that contains non-ASCII characters, the Python logger triggers an un-noticed error:

192.168.25.1 - - [15/Apr/2007 01:37:06] "GET /project1/browser/branches/teste%CC%81 HTTP/1.1" 200 -
Traceback (most recent call last):
  File "/Library/Frameworks/Python.framework/Versions/2.4/lib/python2.4/logging/__init__.py", line 731, in emit
    msg = self.format(record)
  File "/Library/Frameworks/Python.framework/Versions/2.4/lib/python2.4/logging/__init__.py", line 617, in format
    return fmt.format(record)
  File "/Library/Frameworks/Python.framework/Versions/2.4/lib/python2.4/logging/__init__.py", line 408, in format
    s = self._fmt % record.__dict__
UnicodeEncodeError: 'ascii' codec can't encode characters in position 43-44: ordinal not in range(128)

The error can probably be triggered with any requests that contains non-ASCII characters in the path part of the URL.

• A possible workaround is to convert the URL path string to UTF-8:

  trunk/trac/web/api.py

  @@ -169 +169 @@
   
       def __repr__(self):
           return '<%s "%s %s">' % (self.__class__.__name__, self.method,
  -                                 self.path_info)
  +                                 self.path_info.encode('utf-8'))
   
       # Public API

This error occurs when logger output is stderr (the terminal being configured to handle UTF-8). I'm not sure about how it behaves with other output stream such as syslog or a file.

[Christian Boos]

Yeah, I'm still seeing this. Please commit your suggested fix.

[Emmanuel Blot]

I can't reproduce it anymore - weird.

I'm using the following setup:

Trac	0.11dev-r5785
Python	2.5 [GCC 4.0.1 (Apple Computer, Inc. build 5341)]
setuptools	0.6c5
SQLite	3.3.5
pysqlite	2.3.2
Genshi	0.5dev-r657
Pygments	0.7.1
Subversion	1.4.3 (r23084)
jQuery	1.1.2

Can you double check? I don't mind committing the fix, but it seems that the log output is ok w/o it, such as in:

01:29:06 Trac[main] DEBUG: Dispatching <Request "GET u'/browser/TV.com/test/Re\u0301pertoire a\u0300 proble\u0300mes'">

[Christian Boos]

Well, it seems r5646 did the trick.

I'm pretty sure I've seen the error in one circumstance in a later version, but I can't remember the config I used and steps to reproduce, so I'm closing it for now.