http://trac.osgeo.org/proj/timeline <- Internal Error with UnicodeDecodeError

[neteler]

​http://trac.osgeo.org/proj/timeline reports only this Internal Error:

Repository checkins event provider (ChangesetModule) failed:
UnicodeDecodeError: 'utf8' codec can't decode bytes in position 41-43: invalid data

likely due to CVS -> SVN migration of PROJ4.

Here is some material: ​http://trac.edgewall.org/wiki/UnicodeDecodeError which may be helpful.

Markus

[martin]

I'm yet not up to having a clue here ....

According to the mentioned page at "edgewall" I understand that either a different character conversion should have been applied while converting from CVS to SVN or a more recent release of Trac might solve the problem.

Anyone's experienced about configuring Trac !?

Cheers, Martin.

[warmerdam]

Circa line 69 of /usr/lib/python2.3/site-packages/trac/db/sqlite_backend.py I have changed this:

        def _convert_row(self, row):
            return tuple([(isinstance(v, str) and [v.decode('utf-8')] or [v])[0]
                          for v in row])

to:

        def _convert_row(self, row):
            result = []
            for v in row:
                if isinstance(v,str):
                    try:
                        item = v.decode('utf-8')
                    except:
                        item = '(undecodable utf-8 string)'
                else:
                    item = v

                result.append(item)
            return tuple(result)
        
            #return tuple([(isinstance(v, str) and [v.decode('utf-8')] or [v])[0]
            #              for v in row])

This seems to successfully replace untranslatable unicode messages with "(undecodable utf-8 string)". It turns out one of the commit messages include special characters.

If we upgrade Trac we may need to re-apply a variation of this patch.

I did review cvs2svn and it was not immediately obvious how to rerun the translation in such a way that the corrupt string would not be in the log messages.

[neteler]

Replying to warmerdam: ...

This seems to successfully replace untranslatable unicode messages with "(undecodable utf-8 string)". It turns out one of the commit messages include special characters.

If we upgrade Trac we may need to re-apply a variation of this patch.

I did review cvs2svn and it was not immediately obvious how to rerun the translation in such a way that the corrupt string would not be in the log messages.

Here is the GRASS conversion script which does some encoding magic: ​http://josef.fsv.cvut.cz/~landa/grass-cvs2svn/grass-cvs2svn-base.sh

Maybe helpful...