broken national characters while viewing the changeset

[tiger@…]

We're running subversion repository server on linux (system's LANG=ru_RU.KOI8-R), while clients are on windows (codepage windows-1251). Almost all sources in the repository contains characters in windows-1251 codepage. Trac's repository browser shows sources in UTF-8 where national chars are broken (e.g. - unreadable)… Seems it assumes that if the server's LANG is KOI8-R then sources are also KOI8-R and converts sources to UTF-8 from KOI8-R. This is wrong. Is there a hack exists to tell Trac the encoding of the sources to be converted to UTF-8 from?

[daniel+trac@…]

My guess is that Trac should honour the svn:mime-type property and its charset parameter, which is how Subversion keeps track of charsets. Eg.:

svn propset svn:mime-type 'text/plain; charset=iso-8859-1' README

or in the above case,

svn propset svn:mime-type 'text/plain; charset=windows-1251' whatever

Trac should probably convert from that given charset to UTF-8 before displaying it to the browser. Currently, Trac seems to completely ignore the charset specified there, and always assume UTF-8.

[Jonas Borgström]

Trac currently assumes the charset is utf-8 if it isn't iso-8859-1. Honoring charsets specified in the mime-type property is probably the way to go.

[Jonas Borgström]

Fixed in [801]. The new "default_charset" trac.ini option can be used to specify default charset, see TracIni.

[tiger@…]

Changeset [801] closes the problem in the browser, but the same problem happily exists in the changeset viewer.

[Jonas Borgström]

Changeset.py now uses the same charset guessing behavoir as File.py.