launchpad-dev team mailing list archive

[Gary Poster <gary.poster@xxxxxxxxxxxxx>]

= Attendance =

Brad, Benji, Gary

(Francesco swap day, Graham Product Summit/UDS)

= Project status =

Francis approved: when we get a 90% rate of green test runs, we will
start work on stretch goals.  The stretch goal will get a WIP limit of
two cards, and we will continue to work on the test run errors with a
WIP limit of one card.  We must have at least 95% rate of green test
runs to deploy.

RT 50242 for data center temporary machine is ready for review.

= Tricks =

* Brad sent out email to launchpad-dev list (subject: "bin/test
--load-list") of how he is identifying test isolation errors.

* benji: Once you identify a pair of tests that cause test isolation
errors, he does a binary search of code by commenting out parts of the
initial test to determine which part is causing the isolation problem.

= Successes =

* Benji and Brad did a great job collapsing five bugs into one (992692).
 Anything we can learn from that?

   * benji: Already known: Global state is awful, especially for tests
   * bac: More already known: see
https://dev.launchpad.net/ParallelTests/TestIsolation

= Pain =

* benji: make doesn’t work for him on lxc any more: he has to move some
files and directories around.  He needs to find out why. (No suggestions)

* bac: locations.conf for using branch name in bzr push is not set up by
setuplxc.  gary: doesn't work if you are using lightweight checkouts in
a shared sandbox anyway.

* duplicating work: Gary flubbed by not identifying at least two of the
identical test isolation bugs as dupes initially.  We didn't identify
anything I could have done better other than "be more careful," which is
not a process change.  We also did not talk about a success with this:
because Brad talked about the bug on daily standup, he and Benji
identified the similarities and were able to pair.