Traceback: "IntegrityError: columns sid, authenticated are not unique"

[peter.arrenbrecht@…]

I tried to log on while viewing a ticket I wanted to close. Then I got the following:

Traceback (most recent call last):
  File "/var/lib/python-support/python2.4/trac/web/main.py", line 387, in dispatch_request
    dispatcher.dispatch(req)
  File "/var/lib/python-support/python2.4/trac/web/main.py", line 184, in dispatch
    req.session = Session(self.env, req)
  File "/var/lib/python-support/python2.4/trac/web/session.py", line 52, in __init__
    self.promote_session(sid)
  File "/var/lib/python-support/python2.4/trac/web/session.py", line 152, in promote_session
    (self.req.authname, int(time.time())))
  File "/var/lib/python-support/python2.4/trac/db/util.py", line 50, in execute
    return self.cursor.execute(sql_escape_percent(sql), args)
  File "/var/lib/python-support/python2.4/trac/db/sqlite_backend.py", line 56, in execute
    args or [])
  File "/var/lib/python-support/python2.4/trac/db/sqlite_backend.py", line 48, in _rollback_on_error
    return function(self, *args, **kwargs)
IntegrityError: columns sid, authenticated are not unique

[Emmanuel Blot]

Duplicate, search the Trac DB before submitting a new ticket.

[Christian Boos]

This backtrace looks similar to the ones previously reported in #6569 and #6175. However those tickets were closed as an "installation issue", which is not really a satisfying resolution for an internal error without a clearly understood cause.

A similar issue was reported today for a Trac 0.12.1 instance, in ticket:3563#comment:13.

I remember having written something about a race condition in promote_session, but I can't find that information right now.

Also, as this is an "old" problem and not a regression, let's not delay 0.12.2 because of that.

[Carl Fürstenberg <carl@…>]

I assume as 0.12.2 is released, someone could look into this problem now? If possible, we can pay for you to look into this issue.

[Remy Blank]

The only race condition I can see in promote_session() is if two concurrent processes interfere between the initial SELECT and the INSERT INTO of the else: clause. I have added the same workaround as for #3563 in [10825].