20101117
DistributedDevelopment steering meeting to be held in #ubuntu-meeting on 2010-11-17.
Agenda
- Action items
- barry to start some sphinx docs to be well-integrated w/ wiki.u.c (ongoing)
- poolie to start list thread to find problems which can be carried on at UDS (done)
- barry to contact poolie about getting his UDS session notes into the blueprint (done)
UDD survey results (poolie) (see also Bazaar/Plate)
- Bugs of interest:
http://pad.lv/674353 - graph packages built through UDD
http://pad.lv/556132 - don't drop SSH connection after sending 1GB; requested by kiko
http://pad.lv/375013 - support committing direct to stacked branches
- bzr branches are too expensive to use for casual sponsoring, compared with downloading packages from my local mirror (slangasek)
(watch file support) - james_w and barry to sprint on that at uds-n
bzr commit in a heavyweight checkout does not propagate new tags
Interface to dpkg-buildpackage inconsistent and not well documented
- changelog merge problem (probably caused by dpkg-mergechangelogs). need more investigation by barry
- Sponsoring/Patch Pilot
- AOB
Summary
- Top concerns:
- package import failures
- gather data for usage graphs (header to mark built-from-branch)
- network performance holistically
- poolie is looking for someone (preferably from platform) to do a bzr team rotation
New Actions
- poolie to capture uds session notes in a wiki page, with more detail (turn it into a blueprint?)
- poolie to send dump of udd survey results to mailing list
- ajmitch to come up with questions/topics for next meeting (re: REVU)
- poolie to send bzr rotation pitch to platform mailing list
Log
<barry_> #startmeeting
<MootBot> Meeting started at 15:00. The chair is barry_.
<MootBot> Commands Available: [TOPIC], [IDEA], [ACTION], [AGREED], [LINK],
[VOTE]
<barry_> thumper, poolie, james_w ping
<flacoste> me
<james_w> hi
<james_w> is it that time already?
<barry_> ajmitch: ping
<barry_> hi, yeah. i know the usa (canada too?) went off dst so it comes even
earlier for us now [16:03]
* slangasek waves
<poolie> hi
<barry_> slangasek: hi
* ajmitch is here
<barry_> [TOPIC] agenda [16:04]
<MootBot> New Topic: agenda
<thumper> barry_: hi [16:05]
<barry_> thumper: hi
<barry_> [TOPIC] action items
<MootBot> New Topic: action items
<poolie> ?
<poolie> no agenda? i added something...
<barry_> * barry to start some sphinx docs to be well-integrated w/
wiki.u.c (ongoing)
<barry_>
<barry_> er, yes, sorry [16:06]
<barry_> i've been a little ill so kind of unprepared ;)
<ajmitch> I think the wiki formatting is a bit off
<poolie> sorry to hear that
<barry_> fixing... thanks...
<barry_> [LINK] https://wiki.ubuntu.com/DistributedDevelopment/20101117
<MootBot> LINK received:
https://wiki.ubuntu.com/DistributedDevelopment/20101117
<barry_> agenda: action items, udd survey results, bugs [16:07]
<barry_> * barry to start some sphinx docs to be well-integrated w/
wiki.u.c (ongoing)
<barry_>
<barry_> not done, but i'd like to keep it on the list
<poolie> agree
<barry_> * poolie to start list thread to find problems which can be
carried on at UDS (ongoing)
<barry_>
<barry_> i didn't want to take this off without poolie's input
<barry_> and... [16:08]
<barry_> * barry to contact poolie about getting his UDS session notes into
the blueprint
<barry_>
<poolie> ah, i don't think we had a new thread, but i did gather stuff from
old ones
<poolie> i sent a brief mail
<barry_> poolie: i know you posted to ml, but i haven't read it yet
<poolie> with a link
<poolie> should we turn them into a blueprint? it's more a bunch of ideas and
issues than an actual blueprint
<barry_> poolie: thanks. i think that covers both items
<poolie> perhaps we should capture them on a wiki page in more detail than my
mail
<poolie> i'll take that on
<barry_> poolie: thanks. i do think we should have a blueprint that we can
use to track work items
<barry_> [ACTION] poolie to capture notes from email into wiki page with more
detail
<MootBot> ACTION received: poolie to capture notes from email into wiki page
with more detail
<barry_> poolie: which reminds me, will jelmer be joining us here after the
1st of the year? [16:10]
<poolie> from the 10th
<poolie> oh, in this meeting in particular - i hope so
<barry_> cool
<poolie> mm, but what time will that be? about 10pm?
<poolie> if so, i hope he can [16:11]
<barry_> i have no problem moving the meeting earlier to accommodate, but i
think that will affect you and thumper, right?
<poolie> it's currently 8am for me; 7am wouldn't be a burden; 6 would be a
slight burden
<thumper> barry_: I'm ok with an hour earlier (normally)
<poolie> let's see how we go
<barry_> okay, we can take that up when jelmer's on officially [16:12]
<poolie> so he'll be joining us from the start of the bzr sprint at the
platform rally
<barry_> excellent. is that happening at the same time, or a week
before/after? or maybe that's the lp sprint i'm thinking about?
[16:13]
<barry_> (sorry, my brain is not with me today)
<poolie> overlapping in time and space with the (canonical internal) platform
rally,
<barry_> that'll be great
<poolie> and immediately prior to the lp epic, which we will also be attending
<poolie> the first week i plan for us to mostly work together as a small team,
of course also talking to ubuntu people
<poolie> and the second week i want us to do lots of work crossing the lp:bzr
boundary [16:14]
<poolie> eg bzr people fixing things in lp and vice versa
<barry_> nice. i won't be able to attend the lp epic [16:15]
<barry_> [TOPIC] udd survey results
<MootBot> New Topic: udd survey results
<barry_> -> poolie [16:16]
<poolie> ok, we got 82 great responses to our "how is udd for you"
<poolie> mostly pretty positive: net promoter score (of people who responded)
very widely distributed, but with ~60% of people voting 8, 9, or 10
[16:17]
<poolie> lots of textual answers
<poolie> perhaps i'll just send a dump of raw answers to the list?
<barry_> +1
<poolie> i don't think there's anything sensitive in there, but i'll check
first
<poolie> or edit that out if there is
<barry_> probably best to anonymize the responses [16:18]
<ajmitch> the answers which score low would be interesting to read
<barry_> ajmitch: yep. it would be nice to see which of those map to known
bugs
<poolie> so that's an action for me [16:19]
<barry_> [ACTION] poolie to send a dump of udd survey results to mlist
<MootBot> ACTION received: poolie to send a dump of udd survey results to
mlist
<barry_> poolie: any themes that stand out?
<poolie> good qn [16:20]
<poolie> people generally find the bzr interface nice
<barry_> (btw, that Bazaar/Plate page does not yet exist)
<poolie> reliability, including getting up to date imports
<poolie> error messages caan be confusing
<poolie> lp ui clunky
<poolie> it's another layer on top of a complicated stack
<poolie> when things work, it's great
<poolie> contradictory messages (i guess varying experiences) on whether git
imports are good or not
<poolie> also varying opinions on whether they're glad or frustrated we don't
use git [16:22]
<barry_> ;)
<ScottK> barry_: I'll show you my clamav mess at bacon-pig if you want more
use cases.
<poolie> perhaps not surprisingly the people with bad experiences seem to use
it the least
<poolie> so, we need to look for the place to insert a wedge, where making
things a bit better will make people use it more [16:23]
<barry_> ScottK: sounds great
<ScottK> poolie: I gave it a try, wrote a long mail to the UDD list about
issues I found. I don't have a sense that a lot has changed. Not
sure if that's accurate or not.
<poolie> ScottK: if it's the mail i'm thinking of, i think we got through
nearly half since it was sent earlier this year [16:24]
* ScottK would be glad to try it again if significant progress has been made.
<poolie> maybe half by count not size
<ScottK> poolie: OK. That's good to know.
<poolie> the others are definitely still on the list
<poolie> one interesting thing to call out from udd, is that the general
things for the bzr team are
<poolie> - network speed, holistically (including mirroring, or shallow
checkouts, etc) [16:25]
<ScottK> It's difficult to know the state of the project as a casual observer
and has it progressed enough to be worth trying again.
<poolie> maybe we should send a better summary of what's changed since eg the
last ubuntu release?
* ajmitch will need to sort out something about REVU & branches if I'm to be
of use to this meeting [16:26]
<barry_> poolie: +1
<barry_> ajmitch: yes, would you like to prepare something for 2 weeks time?
[16:27]
<ajmitch> barry_: OK, I'll try & come up with some questions/suggestions at
least
<barry_> ajmitch: great! [16:28]
<barry_> who wants to talk about bugs? :)
<barry_> [TOPIC] bugs of interest / top bugs
<MootBot> New Topic: bugs of interest / top bugs
<slangasek> reliability is still a big problem right now; almost 2000 failures
according to http://package-import.ubuntu.com/status/
<barry_> i'm a little afraid this list is getting pretty long ;)
<barry_> slangasek: agreed. we really need to address the import failures.
can't use udd if the branches aren't there [16:29]
<james_w>
http://package-import.ubuntu.com/status/language-pack-gnome-lo-base.html
<MootBot> LINK received:
http://package-import.ubuntu.com/status/language-pack-gnome-lo-base.html
<james_w> that's a funny failure mode
<barry_> 696 packages failed with key
AssertionError:<module>:main:import_package:import_package [16:30]
<ajmitch> debian unstable is hardly getting any updates at the moment, due to
squeeze freezing
<ajmitch> oops, ECHAN
<poolie> so package imports in particular, as a special focus? ok
<slangasek> james_w: I see gourmet has landed back in that failure state again
in spite of your efforts :(
<ScottK> ajmitch: It actually fits.
<poolie> we're still having some trouble getting enough time to work on udd
things outside of bzr itself
<poolie> perhaps a big push on that specific thing will help us focus on it
[16:31]
<james_w> slangasek, yeah, I'm still working on it, but it's having to fit in
around other things unfortunately
* slangasek nods
<barry_> poolie: we have bug 653307 for a different failure case, i guess we
should have separate bugs for each major class of import failures?
<ubottu> Launchpad bug 653307 in Ubuntu Distributed Development "Import fails
with missing referenced chk root keys" [Critical,In progress]
https://launchpad.net/bugs/653307
<james_w> slangasek, I understand why we are getting the failure though now
<slangasek> oh, good :)
<barry_> we have 10 "bugs of interest". do we want to go through each one?
maybe take a look at the wiki page and see if there has been any
progress on any of them? [16:34]
<poolie> barry_: 10? i see 3 on the agenda..
<poolie> ... unless i reload [16:35]
<barry_> refresh :)
<poolie> let's go, quickly
<poolie> a graph of packages is flacoste's favorite nomination
<poolie> (and he sets my bonus :-)
<ajmitch> heh
<poolie> on the grounds that you should put metrics in place early in the use
of a feature
<poolie> this is mostly about inserting an X-SBC field to mark a package as
having been built from a branch [16:36]
<poolie> probably trivial to do; just needs figuring out what would be
tasteful to put in it
<ajmitch> how much of the BFB infrastructure is in place for that so far?
<flacoste> i'd also like a graph of the package-imports and/or hottest100
results
<poolie> according to jelmer there is infrastructure to scan it out there
<flacoste> so that we assess the quality of our service
<poolie> there are graphs of them [16:37]
<poolie> possibly there should be better graphs?
<poolie> if we're going to do 10 bugs, and stay roughly within our hour, i'm
going to go fairly fast
<barry_> poolie: please do [16:38]
<poolie> http://pad.lv/556132 is the top vote for Linaro - problems in conch -
probably needs to be fixed upstream in twisted; i will push it
<MootBot> LINK received: http://pad.lv/556132 is the top vote for Linaro -
problems in conch - probably needs to be fixed upstream in twisted;
i will push it
* barry_ remembers a uds conversation about this during a walk to/from dinner
<jam> poolie: I assume that is the point where ssh needs to renegotiate keys?
[16:39]
<poolie> http://pad.lv/375013, commit to stacked branches, is Tim's top vote
for the code team - it will ease some issues in building enormous
branches in moderate sized vms
<MootBot> LINK received: http://pad.lv/375013, commit to stacked branches, is
Tim's top vote for the code team - it will ease some issues in
building enormous branches in moderate sized vms
<poolie> jam: yes, apparently conch says "too hard, bye!"
<poolie> > bzr branches are too expensive to use for casual sponsoring,
compared with downloading packages from my local mirror (slangasek)
<poolie> needs to be turned into a proper bug, with a number - can you do that
steve?
<poolie> (watch file support) - james_w and barry to sprint on that at uds-n
[16:40]
<flacoste> poolie: couldn't we use stacked branch for that?
<poolie> rockin
<poolie> flacoste: for what?
<slangasek> poolie: ack
<flacoste> poolie: specifically, slangasek bug
<poolie> it can be part of the solution
* flacoste trying to push Tim's bug up the chain :-)
<poolie> this is the thing from uds i described as "network performance
holistically" [16:41]
<barry_> on bug 295274; i need to get back to that branch
<ubottu> Launchpad bug 295274 in bzr-builddeb "merge-upstream shouldn't
require --version when debian/watch is present" [High,Triaged]
https://launchpad.net/bugs/295274
<poolie> https://launchpad.net/bugs/653307 missing chk root keys
<jam> flacoste: the issue raised at UDS is that no matter how little we
download, a local mirror is still a lot faster
<jam> (even if we get the download to be ~ the same size as the tarball)
[16:42]
<flacoste> jam: that's compared to apt-get source right
<jam> right
<poolie> right
<slangasek> right - to be productive I need a good way to have the source
already be local
<jam> there was a discussion about being able to list the packages you are
interested, and having a cron-ish script keep those up to date in a
local mirror/shared repo
<poolie> there are a bunch of things we can usefully do - local mirrors, raw
speed, etc
<poolie> let's give it a number, then split it into bugs for each step
<poolie> re the missing chk root keys [16:43]
<poolie> i know spiv was working on this
<poolie> he didn't mention it recently; i'm not sure why he stopped [16:44]
<poolie> james_w: do you know?
<james_w> no, I don't
<poolie> i'll ping on the bug [16:45]
<james_w> I think the immediate cause was his bout of illness
<barry_> [ACTION] slangasek to file bug on "bzr branches are too expensive to
use for casual sponsoring, compared with downloading packages from my
local mirror"
<MootBot> ACTION received: slangasek to file bug on "bzr branches are too
expensive to use for casual sponsoring, compared with downloading
packages from my local mirror"
<poolie> probably
<ScottK> poolie: Why does the distro packages file need to know if something
was built from branch?
<poolie> https://bugs.launchpad.net/bzr/+bug/603395 commit in checkout doesn't
propagate tags [16:46]
<ScottK> (or flacoste)
<poolie> spiv is working on this and i think it should be done soon
<flacoste> ScottK: simply to allow us to track uptake of UDD across the distro
<poolie> ScottK: two things: firstly, to answer "how can i get the exact
source for this"; secondly and a bit pragmatically, so we can trace
how much of ubuntu is using this
<poolie> we should split it out from the 'want a graph' bug
<ScottK> Right, but making the packages file larger makes apt-get update
slower for every single user of Ubuntu.
<poolie> true
<ScottK> Surely there are less invasive ways to track this.
<poolie> perhaps it should not propagate to the binary? or perhaps there's a
better way?
<james_w> there probably is
<poolie> i'd love suggestions
<poolie> and that's what i meant by making sure it was tasteful
<james_w> X-C should be enough, but that might require some special-casing in
LP unfortunately [16:48]
<ScottK> I think that's better than adding more noise to the Packages files.
<james_w> we need to decide what data we want, and then we can come up with a
way to allow us to capture it
<poolie> ok
<james_w> if only there was a forum in which we could discuss what data to
capture to ensure we were tracking what we wanted [16:49]
<poolie> there's a bug, let's capture the constraints/ideas there
<poolie> https://launchpad.net/bugs/653832 "trying to import version
... again"
<poolie> new to me [16:50]
<james_w> that's one that I'm looking at currently
<james_w> I know the immediate cause, but I'm a bit worried about what will
happen when I unclog the pipe :-)
<poolie> cool, thanks
<poolie> there's a related thing that's on my plate, which is migrating the
package importer onto "proper" is-managed service [16:51]
<poolie> currently in a holding pattern waiting for losa timeslices
<poolie> https://launchpad.net/bugs/499684 "Interface to dpkg-buildpackage
inconsistent and not well documented" is also new to me
<barry_> i think that's ScottK's complaint about -- (one i share :) [16:52]
<poolie> it seems like a good bug report [16:53]
<ajmitch> it is a bit obscure when explaining how to use the tools to
newcomers
<barry_> on my changelog "bug" - i haven't hit it again yet, so i'm still
waiting for a specific example before i file a bug
<poolie> i don't think it should preempt others already in train?
<barry_> i should probably remove it from top bugs since it isn't one yet
<barry_> i think that's it for bugs, unless anybody has any other comments?
[16:54]
<barry_> [TOPIC] * Sponsoring/Patch Pilot [16:55]
<MootBot> New Topic: * Sponsoring/Patch Pilot
<poolie> please file a bug for the changelog/merge thing when you can describe
it
<poolie> nothing else from me
<barry_> poolie: yep, definitely
<barry_> so next week we start the new regime for sponsoring. i'm up on
monday and i plan to heavily encourage branches and merge proposals.
i'd like to give preferential treatment to reviews in mps. (note
though i can only review anyway) [16:56]
<ajmitch> this'll be a good way to get feedback about sponsoring workflow
[16:57]
<barry_> yep, i'll keep copious notes :)
<poolie> that's a great idea [16:58]
<barry_> i'll post my thoughts to the udd mlist afterward
<barry_> i just wanted to mention this; i don't have much else [16:59]
<barry_> unless anybody has suggestions, or things to emphasize, etc... ?
<barry_> [TOPIC] AOB [17:00]
<MootBot> New Topic: AOB
<barry_> anything else not on the agenda?
<poolie> sorry one more
<poolie> i'd like to look for someone to do a rotation onto the bzr team,
ideally from platform
<poolie> to free up some time for them to work on this
<poolie> if anyone here is interested in working on something different,
that's great [17:01]
<poolie> otherwise, think of someone else you might know who it would suit
<barry_> poolie: maybe a solicitation on the platform mlist? (i don't think i
can take it on until the py27 transition is done)
<poolie> maybe someone who's been working on the same group for a few years
and could do with a refresher
<poolie> sure, i just thought about it in the shower this morning [17:02]
<barry_> :)
<poolie> i will get the idea out there
<james_w> good idea
<barry_> awesome
<barry_> [ACTION] poolie to send bzr rotation pitch to platform list
<MootBot> ACTION received: poolie to send bzr rotation pitch to platform list
<poolie> i'd like to also approach likely candidates individually, if we can
get a shortlist
<barry_> cool, thanks. any other business? [17:04]
<poolie> i'll just mention https://wiki.ubuntu.com/Bazaar/Plate
<poolie> (private only, sorry)
<poolie> as a view of what we're up to
<poolie> nothing else from me [17:05]
<james_w> I have a merge proposal for barry_, but I think that's a bit to
"Other Business"
<james_w> poolie, ubuntu.com or canonical.com?
<poolie> sorry, canonical [17:06]
<poolie> https://wiki.canonical.com/Bazaar/Plate
<barry_> poolie: thanks [17:07]
<barry_> okay, folks. thanks again for another great meeting. see you in two
weeks
<barry_> #endmeetingDistributedDevelopment/20101117 (last edited 2010-11-29 21:07:11 by barry)