Merging code to PQM - update

[Robert Collins]

So we did an experiment recently using Launchpad to more closely
manage and monitor the landing of branches in PQM via the Launchpad
'merge queues' feature [which isn't finished - this was a dogfooding
experiment]. I should have sent more updates about this than I did -
sorry!

Firstly, I'd like to document what we've achieved:
 - we found some limits/issues in the launchpadlib API which have been
filed - they affect efficiency and robustness of using it, not whether
we can or not. Bugs have been filed for all of these things - and they
are launchpad API bugs, not launchpad-code (the code.launchpad.net
website) bugs.
 - we found out more about what we will need from a polished merge
daemon API in Launchpad; The Launchpad code team likely knew much of
this, but I think its useful to have a list of bugs we can review.
Most of them (pre-existing ones may not be tagged) are listed here:
https://bugs.edge.launchpad.net/launchpad-code/+bugs?field.tag=merge-queues
 - PQM is broadly ready for use with Launchpad queues when the are
available - and we know what we need to change to work with that.
 - This also helps us prepare for a switch to Tarmac: we've got a list
of the things needed to manage our process in Launchpad, and can thus
estimate what it will take for us to switch. Until queues are ready,
we're not really in a position to switch though.
 - hydrazine is ready to work with queues when we want it to.

Now, I think that this gave us a better workflow, once the teething
problems were addressed. We can identify a few things that we should
do better next time:
 - we should have a staging environment for PQM; there was much more
disruption there than desirable.
 - we should figure out some way to test code using launchpad via
launchpadlib - its really not acceptable as it stands.
 - I should have communicated significantly more verbosely about this
whole process

However, Tim, the launchpad code team manager has some concerns about
security for merge daemons like PQM and Tarmac in this area, and has
requested that until a couple of key issues are addressed, that we
stop dogfooding. This means that I'm going to file an RT to disable
the Launchpad OAuth token in our PQM configuration, and it will stop
reading from Launchpad as a result. We can resume dogfooding when the
bugs in this list:
https://bugs.edge.launchpad.net/launchpad-code/+bugs?field.tag=dogfood-requirement++merge-queues&field.tags_combinator=ALL
have been closed.

For the moment, I'm not going to be putting work time into moving this
forward, but I am likely to put personal time in: I'd love for us to
be able to get rid of the email glue to getting things landed, and I
really loved the ease that we achieved. If anyone else wants to help
out on this, grab Tim (thumper on IRC) and chat with him :)

-Rob