launchpad-dev team mailing list archive

[Gary Poster <gary.poster@xxxxxxxxxxxxx>]

= Attendance =

full house (bac, benji, frankban, gmb, gary_poster)

= Project plan =

no change.

Discussed availability:
 * Benji is out Mon & Tues of next week, and Wed-Fri of following week
 * Francesco is out Tuesday and Wednesday of next week
 * Graham is unavailable for all practical purposes for the next two weeks.

= Tricks =

- gary proxying benji: To work around ./bin/test --load-list bug 987904, 
you can use --shuffle and start passing in values to --shuffle-seed 
until you get the test run order you want. If you get a failure, you can 
throw away the tests that came after it.

- gary: we had talked about comparing Python to Go and suspecting we 
would miss the REPL and pdb; in off-work reading found basic Go REPL 
(https://bitbucket.org/binet/igo) and gdb support 
(http://golang.org/doc/gdb).  Also mentioned shebang support with "go 
run" (apparently by niemeyer?)

= Successes =

- benji: Google hangouts have really changed our collaboration and work 
habits dramatically and for the better.  Terminal sharing really helps 
him a lot (mentioned previously @ 
https://dev.launchpad.net/yellow/RemoteTerminalBroadcasting), as opposed 
to screen sharing.  He is working on developing that further.

- gmb made hacking instructions much more concise; we attribute this to 
his innate brilliance + ability to prioritize what is initially important.

= Pain =

- gary: "buildbot results show subunit streams filtered by workers" card 
has been there for two weeks now. We talked about it last week and said 
it was OK.  Is it still OK?  It is OK that we are working on it--still 
bringing value--but we are not happy with how we are tracking it.  Benji 
has mentioned in the past that we should have a different way of 
tracking these kinds of tasks--tasks that we discover to require 
subdividing.  We will call these subdividable mini-feature cards "meta" 
cards.  Proposal: Create an active Meta lane with coding/review 
sub-lanes with limit of 1.  Meta upcoming sub-tasks tracked separately. 
 Experiment four weeks. Success criteria:  Meta tasks done faster, and 
true state easier to see.

- gary: we keep on finding new intermittent bugs.  They are real, and 
good to fix, but do we have to be fixing all of them for this project? 
Not all of them might actually bite us.  Reply: Ideally, yes.  We might 
consider turning off --shuffle within the worker process as pat of an 
exit strategy.  Revisit in two weeks.

- gary: In four weeks, let's have another "hey how is this meeting 
going" question.

- team: We discussed gmb's desire to get an AMI to speed up the juju LP 
deployment for his UDS session.