Traceback when sending # or ; character from remote search or alert field

[mahfiaz]

This is a regression compared to 1.9.7
Maybe there are more characters which need special treatment besides # and ;

[mahfiaz (mahfiaz)]

It turned out that urlparse uses & and ; as equivalent separators.

urlparse.parse_qsl('data={"request":"da;ta=some"}')
returns
[('data', '{"request":"da'), ('ta', 'some"}')]

urllib.urlencode([('data', '{"request":"da;ta=some"}')])
returns
'data=%7B%22request%22%3A%22da%3Bta%3Dsome%22%7D'

Although we could replace these strings in httpserver.py in ready_read() right before passing the string to urlparse.parse_qs(), it seems like a wrong place to fix it.

[mahfiaz (mahfiaz)]

The problem seems to be, that we use QtCore.QByteArray.fromPercentEncoding(data) and unicode(data, 'utf8') too early (introduced by me several months ago), before JSON decoding, if it was done later, none of that would be a problem and fronends could have gone untouched.

Tim, what do you think about it?
http://paste.ubuntu.com/882195/