[TransWarp] problem with commit and activated query-link

[Phillip J. Eby]

At 01:06 PM 2/18/03 +0100, Ulrich Eck wrote:
>Traceback (most recent call last):
>  File "<stdin>", line 1, in ?
>  File "/usr/lib/python2.2/site-packages/peak/storage/transactions.py", 
> line 20, in commitTransaction
>    getTransaction(subject).commit()
>  File "/usr/lib/python2.2/site-packages/peak/storage/transactions.py", 
> line 223, in commit
>    self._prepare()
>  File "/usr/lib/python2.2/site-packages/peak/storage/transactions.py", 
> line 183, in _prepare
>    unready = [p for p in state.participants if not p.readyToVote(self)]
>  File "/usr/lib/python2.2/site-packages/peak/storage/data_managers.py", 
> line 498, in readyToVote
>    self.flush()
>  File "/usr/lib/python2.2/site-packages/peak/storage/data_managers.py", 
> line 444, in flush
>    del dirty[key]
>KeyError: 137744044
>
>this probably happens because flush is called when oidFor and readyToVote 
>is called.
>the flush in oidFor already clears the dirty-flag, and the flush in 
>readyToVote seems
>not to get this right (dirty.values() seem to do something wrong in this 
>case).
>
>would a simple condition in flush():
>
>if key in dirty:
>  del dirty[key]
>
>solve the problem (objects might have been saved twice in this case)
>or is it a problem with my implementation?

Nope, it's a problem with *my* implementation.  :)  You're right, this part 
of flush() is not re-entrance safe, and your proposed change should fix it.