Error with non-ASCII project description

[eblot]

If the project description in the configuration file (conf\trac.ini) contains non-ASCII characters, such as:

[project]
; ...
descr = Tést

Trac [3094] fails with the following error:

Python traceback

Traceback (most recent call last):
  File "trac\web\main.py", line 299, in dispatch_request
    dispatcher.dispatch(req)
  File "trac\web\main.py", line 153, in dispatch
    populate_hdf(req.hdf, self.env, req)
  File "trac\web\main.py", line 92, in populate_hdf
    hdf['project'] = {
  File "trac\web\clearsilver.py", line 194, in __setitem__
    self.set_value(name, value, True)
  File "trac\web\clearsilver.py", line 236, in set_value
    add_value(name, value)
  File "trac\web\clearsilver.py", line 228, in add_value
    add_value('%s.%s' % (prefix, k), value[k])
  File "trac\web\clearsilver.py", line 218, in add_value
    self.hdf.setValue(prefix, markup.escape(value))
UnicodeEncodeError: 'ascii' codec can't encode character u'\xe9' in position 1: ordinal not in range(128)

It seems the trac.ini file encoding is irrelevant. At least, the same trouble occurs with iso-8859-1 and utf-8 encodings.

The trouble occurs with the project description, but it probably occurs with other fields as well.

[cboos]

Thanks for having tested that. I suspected problems like this
from looking at some of the (various) backtraces in #2905.

I think it should be OK to require that the trac.ini file
should have the same encoding as the one specified in

[trac]
default_charset = ...

[eblot]

Is this option still necessary ? (I mean, can't UTF-8 always be used)

[cboos]

default_charset is used when there's no other clue to
determine what charset is used by some external content
(like an attachment file, a source file in the repository,
etc.) so I think using it for the TracIni file itself is
a good fit.

Note: using iso-8859-15 or similar here rather than utf-8
has the nice property of making a file always decipherable,
as this is a fixed byte encoding.
If you'd use utf-8 for that config option, you'll likely go into
trouble whenever you'll manipulate anything but ascii or utf-8.

[cboos]

Possible fix for the issue; consider trac.ini is an UTF-8 encoded file, or encoded using the system encoding

[cboos]

Well, finally in the attachment:text_to_unicode_r3109.diff​
I took the approach of defaulting to UTF-8, and if this doesn't
work, use the system encoding, after a suggestion from cmlenz.

[cboos]

This should now be fixed in r3118.