UnexpectedFormData: Invalid moderation action for held message

[Curtis Hovey]

OOPS-1720EB821 was cause while moderating the last batch of a mailing list moderation queue
     https://edge.launchpad.net/%7Eopenjdk/+mailinglist-moderate?start=2950&batch=50

The tail of the oops is:
  Module lp.registry.browser.team, line 859, in moderate_action
    (message.message_id, action_name))
UnexpectedFormData: Invalid moderation action for held message <email address hidden>: None

^ This message was already moderated. This error might relate to the fact that we are process messages in the action by batch or because the message's status has yet to transition from DISCARD_PENDING to DISCARDED by the XMLRPC polling processes

[Curtis Hovey (sinzui)]

This bug seems to mutate, which leads me to think the issue relates to skipping messages in a batch. On set of oopses were address by going back to the first batch and moderating the first message. The notification reported that I moderated the problem message in the oops

[Robert Collins (lifeless)]

I think this could also happen with the single page form if folk used it rapidly twice in a row, but the batching may well have exacerbated it?

It seems to me we need to have the message id in the form data to fix this properly; do you agree?

[Curtis Hovey (sinzui)]

I do agree. I think we want to extend MessageApprovalSet.getMessageByMessageID(), or add getMessageByMessageIDs(id_list)

[Launchpad QA Bot (lpqabot)]

Fixed in stable r11573 <http://bazaar.launchpad.net/~launchpad-pqm/launchpad/stable/revision/11573>.

[Robert Collins (lifeless)]

I'll need help QAing this, AFAIK I'm not in any teams with nontrivial traffic flow (such that looking at the last batch can cause the issue).

Any suggestions of someone in such a team that can help us?

[Curtis Hovey (sinzui)]

I still have access to openjdk and I did not get an error moderating a couple hundred messages from the head and tail of the batch. I am marking this as qa-ok since I was seeing the error frequently when I discovered this bug,