20110713
DistributedDevelopment steering meeting held in #ubuntu-meeting on 2011-07-13 at 1200 UTC. <-- note time change
Chair: barry
Apologies
Agenda
- Attendance/apologies
- Action items
jelmer to study the feasibility of merge helper (bug 608675) as an intermediate step for quilt support
jelmer to look into bug 609187 (warn when package import is out of date)
Riddell to file bug saying that packaging branch pages on code.lp.net should be better labeled (done)
- Package importer progress
Get rid of bzr mark-uploaded ?
- Bugs of interest:
- Any other business?
Summary
- goal is to reduce import failures to under 200 by start of october
- jelmer made progress on support for multiple upstream tarballs
merge preview diffs are available at http://package-import.ubuntu.com/merges/ (but causes some import failures, maybe move it out-of-line?)
vila working on NoFinalPath bug fix
- post-mortem thread on lp/bzr update and its effect on bzr-svn
- poolie plans to stop the package importer pretending to be james_w
New Actions
- jelmer to post summary of user visible changes with lp rollout
- jelmer to post feature summary about inline diffs
riddell to update udd docs to use bzr tag instead of bzr mark-uploaded
Log
<barry> #startmeeting
<MootBot> Meeting started at 07:01. The chair is barry.
<MootBot> Commands Available: [TOPIC], [IDEA], [ACTION], [AGREED], [LINK],
[VOTE]
<jelmer> hi barry!
<barry> jelmer: hi!
<poolie> hi barry, jelmer
<barry> hi poolie
<barry> do you know if anyone else is joining us today?
<barry> well, anyway, let's get started
<barry> [TOPIC] agenda
<MootBot> New Topic: agenda
<barry> [LINK] https://wiki.ubuntu.com/DistributedDevelopment/20110713
<MootBot> LINK received:
https://wiki.ubuntu.com/DistributedDevelopment/20110713
<barry> [TOPIC] action items [08:03]
<MootBot> New Topic: action items
<barry> * jelmer to study the feasibility of merge helper
([[https://bugs.launchpad.net/bzr-builddeb/+bug/608675|bug 608675]])
as an intermediate step for quilt support
<barry>
<ubottu> Ubuntu bug 608675 in bzr-builddeb "merge-package should have support
for manipulating quilt v3 patch stacks" [High,In progress]
<jelmer> I've been away for the last week so haven't spent any time on this
since [08:04]
<jelmer> at the last meeting (at the rally) we discussed it, and my
conclusions there were that it seems feasible and I've started
working on implementing it.
<barry> jelmer: awesome. we'll just keep it on for next time [08:05]
<poolie> that's great
<barry> * jelmer to look into
[[https://bugs.launchpad.net/udd/+bug/609187|bug 609187]] (warn when
package import is out of date)
<ubottu> Ubuntu bug 609187 in Ubuntu Distributed Development "users are not
warned when branching ubuntu:foo (or lp:ubuntu/foo) and the package
import of foo is out of date" [High,Triaged]
<poolie> there was some discussion recently of this being useful even for
non-quilt branches
<poolie> hm
<poolie> or, more generally, of being able to merge both the patched and
unpatched forms
<jelmer> jam mentioned on IRC earlier that he's working on bug 609187
<poolie> barry, i think jam and others worked on that [08:06]
<poolie> oh, and maxb in london
<barry> very nice. so, in progress?
<jelmer> yep
<jelmer> should perhaps be owned by jam though?
<james_w> hi [08:07]
<maxb> jam has taken my proof of concept "How to talk to LP without
launchpadlib" and is pursuing it
<barry> sure, i can update that.
<poolie> hi maxb
<barry> hi maxb
<barry> maxb: do you just make straight up http calls?
<poolie> the point of avoiding lplib is that it's much faster to just make the
single call we care about [08:08]
<poolie> yes
<barry> makes sense
<maxb> Minimizing roundtrips and costly WADL downloads really matters if we're
going to do this interactively
<barry> +1
<Pendulum>
<barry> * Riddell to file bug saying that packaging branch pages on
code.lp.net should be better labeled [08:09]
<barry> anybody know if the bug has been filed?
<poolie> istr that he did
<jelmer> ... and fixed the bug :)
<poolie> and, perhaps he worked on it at the rally
<poolie> hooray
<barry> yay!
<barry> do you happen to have the bug #?
<poolie> bug 797688 [08:11]
<ubottu> Launchpad bug 797688 in Launchpad itself "Ubuntu Packaging Branches
Should be Labeled Clearer" [Low,Fix released]
https://launchpad.net/bugs/797688
<barry> thanks (you're quicker at searching than me at this hour :)
<poolie> i have kanban fu
<poolie> http://people.canonical.com/~mbp/kanban/jr-kanban.html
<MootBot> LINK received:
http://people.canonical.com/~mbp/kanban/jr-kanban.html
<poolie> scan down the right hand column until you find it
<barry> ah, nice!
<poolie> i like it [08:12]
<barry> i didn't realize each person had their own kanban
<barry> [TOPIC] package importer progress [08:13]
<MootBot> New Topic: package importer progress
<poolie> yes, there's one aggregated per team and one per person
<poolie> let's see
<poolie> i think max said some new things started failing
<maxb> There does seem to have been a small uptick in the count
<poolie> hm, so it stepped down a bit, but now has drifted up
<maxb> Besides the append-revisions-only stuff, which I'm taking account of
<poolie> back from 400 to 425?
<poolie> some of these are AppendRevisionsOnlyViolation which is a recent
regression [08:14]
<poolie> so i'm proposing to make this rather more of a priority for the
canonical bzr team
<maxb> The append-revisions-only stuff accounts for ~25, but I've fixed about
~20 imports in the last 48 hours, which is why were are back at around
400
<poolie> putting it at the top of our quarterly planning commitments, and
trying to get it below 200 by the start of october
<poolie> which would be much faster than our current trend [08:15]
<barry> poolie: that would be amazing
<poolie> thanks max
<jelmer> I've made some progress on support for multiple upstream tarballs,
which isn't finished yet but should help a bit too.
<poolie> oh, that's great
<maxb> One thing is that there seems to be a gradual uptick in the number of
packages using multiple upstream tarballs. That's also the top single
failure cause at the moment
<maxb> Another thing to consider is that quite a number of the failures are
failures during the generation of merge preview diffs - *not* the
import itself [08:16]
<maxb> A subcategory of those are, I believe, being addressed by a bugfix of
vila's
<poolie> why does it generate merge preview diffs? [08:17]
<maxb> I'm guessing because james_w thought it would be useful?
<maxb> I doubt many people know they exist.
<maxb> http://package-import.ubuntu.com/merges/
<MootBot> LINK received: http://package-import.ubuntu.com/merges/
<james_w> I started generating them as a test [08:18]
<poolie> hi james
<james_w> but never followed it through to be the replacement for
merges.ubuntu.com
<james_w> hi
<poolie> huh
<barry> james_w: hi. i didn't know that existed either
<poolie> so, what shall we do
<poolie> perhaps we should move it out of line so that it doesn't cause
failure of the actual package import? [08:19]
<maxb> I estimate ~44 failures are in the merge generation. They are probably
legitimate bzrlib bugs
<maxb> I think we should be mindful of that possibility, but see how many of
the failures are cleared up by vila's NoFinalPath bugfix before
investing time [08:20]
<james_w> that seems to be a fair number of them
<jelmer> Merges are one of the goals of the UDD branches, so it seems useful
to keep it in. [08:21]
<james_w> given no-one is using the output currently it wouldn't really cost
anything to delete the code
<barry> or maybe comment it out for now?
<poolie> ok, well, now i know it's there :) [08:22]
<barry> :)
<poolie> i agree with max, let's see how it goes with current in progress work
landed, and then we can think about pruning it
<poolie> or, alternatively, at least linking to ti
<barry> +1
<barry> so, not really on the agenda, but i see lp and bzr got updated last
night? how'd that go? [08:23]
<poolie> it was a bit bumpy
<barry> dang
<barry> but everything worked out?
<poolie> since we went several revisions forward, and there were some bad
interactions between changes in bzr and in bzr-svn when importing
large branches
<poolie> we are having a postmortem thread at the moment [08:24]
<poolie> to try to understand what happened
<poolie> but i think it will be ok
<barry> cool. i'm on the mlist so i'll follow along
<poolie> we will make it safer to deploy and roll back, and then deploy more
often
<poolie> similar to the approach being taken with other bits of lp
<poolie> and speaking of which, can i just say how much i'm looking forward to
their new shorter upgrade windows [08:25]
<poolie> (https://lists.launchpad.net/launchpad-dev/msg07623.html but it's a
bit offtopic)
<poolie> jelmer, can you tell us about what this is likely to change with
imports?
<jelmer> poolie: Can you rephrase that, do you mean the vcs imports? [08:26]
<poolie> yes [08:27]
<barry> also, what are the top-level user visible changes we'll see in
bzr/udd/codehosting now?
<poolie> we can hope to see more of them succeed after the lp upgrade?
<jelmer> poolie: What was the recent lp upgrade from last night barry was
referring to exactly? [08:28]
<poolie> there should be a big new feature in lp which is showing diffs inline
in the branch pages
<poolie> lp had an upgrade outage today (australian time)
<poolie> hm
<barry> that will be *awesome*
<poolie> perhaps we should get our story straight offline :)
<poolie> and then post something on the lp blog about user visible changes, if
any
<barry> poolie: +1, can i give you that action? [08:29]
<poolie> actually could i put that on jelmer, since he landed the change iirc
<poolie> the inline content feature is
http://people.canonical.com/~andrew/Inline-diff-screenshot.png
[08:30]
<poolie> not landed yet
<barry> [ACTION] jelmer to post summary of user visible changes with lp
rollout
<MootBot> ACTION received: jelmer to post summary of user visible changes
with lp rollout
<jelmer> I haven't caught up on all that email yet, but I'd be happy to take
an action item about the post-morten analysis
<jam> poolie: that specific screenshot is for merge proposal pages
<poolie> true
<jam> danilo's change was for Branch pages, which I certainly don't see, but I
don't know whether or not his expander code has landed.
<poolie> i think the code changes proposed now may only touch mp pages but we
should be able to follow through and show stuff on the branch too
<poolie> jelmer so i guess there's two things to follow on, one is the
postmortem (what went wrong) and the other is the feature
announcement, if any (what will go right) [08:31]
<poolie> i had the idea that the updated bzr-git etc are going to help a lot
more branch imports complete
<poolie> am i confused?
<jelmer> I haven't landed anything related to the expander code [08:32]
<jelmer> I do have a fix for having bzr-svn deal with tags better (not use so
much memory), but IIRC I haven't proposed that for landing on lp yet
<poolie> didn't you land several updates into lp when we were at the rally?
[08:33]
<poolie> perhaps we should move on; this isn't really udd specific [08:35]
<maxb> I have something I'd like people to have a think about: What are we
going to do about developer-maintained package branches, which don't
use upstream-* tags because they merge directly from upstream's bzr,
and thus the UDD importer always fails on currently? I'm thinking of,
e.g. apport
<jelmer> (sorry, looking through the revision log atm..)
<barry> poolie: okay
<barry> [TOPIC] * Get rid of `bzr mark-uploaded` ?
<MootBot> New Topic: * Get rid of `bzr mark-uploaded` ?
<jelmer> taking that offline (in the post-mortem thread perhaps?) might indeed
be better
<Riddell> hi, sorry I'm late [08:36]
<barry> so, if i'm not mistaken, mark-uploaded is more or less unnecessary
now, right?
<jelmer> somewhat; debcommit -r will tag too and "bzr tag" without argument
can also behave the same way as mark-uploaded [08:37]
<Riddell> hmm, I was hoping to get away from using debcommit
<barry> jelmer: what does mark-uploaded do that bzr tag or debcommit doesn't?
<barry> i guess there are two things leading these questions: whether to
change the udd documentation to just use bzr tag, and whether to
actually deprecate or get rid of the command?
<jelmer> barry: I don't think there's anything it does that bzr tag doesn't,
other than perhaps being a bit more clearly named
<barry> jelmer: cool. any objections to me changing the udd docs to use tag
instead?
<poolie> james_w ^?
<barry> Riddell: i'd like to get away from recommending debcommit too
<james_w> not really
<poolie> i wonder if bzr builddeb could instead hook into the tag command
(through adding a hook, not in some gross way) to give people policy
checks? [08:40]
<james_w> probably worth a note that if bzr says that they have to give a tag
name they need a newer bzr/bzr-builddeb, or their bzr-builddeb is
incorrectly installed
<poolie> if those checks are seen as useful
<barry> james_w: good idea [08:41]
<barry> [ACTION] barry to update udd docs to use `bzr tag` instead of `bzr
mark-uploaded`
<MootBot> ACTION received: barry to update udd docs to use `bzr tag` instead
of `bzr mark-uploaded`
<poolie> barry i suggested Riddell might put his packaging background to use
in working on the packaging guide with yourself and daniel
<Riddell> yes, I'm working my way through it fixing things and working out
what's missing or not clear [08:42]
<poolie> great
<barry> Riddell: fantastic. you can add me as a reviewer for any mp
<poolie> shall we talk about max's question?
<barry> Riddell: do you want to take the above action instead of me?
<ScottK> Riddell: Does that include non-UDD packaging?
<Riddell> barry: yes can do
<poolie> hi scott [08:43]
<barry> [ACTION] Riddell to update udd docs to use `bzr tag` instead of `bzr
mark-uploaded`
<MootBot> ACTION received: Riddell to update udd docs to use `bzr tag`
instead of `bzr mark-uploaded`
<barry> poolie: yes, let's take up maxb's question
<maxb> <maxb> I have something I'd like people to have a think about: What are
we going to do about developer-maintained package branches, which don't
use upstream-* tags because they merge directly from upstream's bzr,
and thus the UDD importer always fails on currently? I'm thinking of,
e.g. apport
<Riddell> ScottK: I'd like to discuss that with dholbach, it needs quite a lot
of changes to include non-UDD bits and I don't know if he has
thought about how to do that (but essentially i'm working on it from
a bzr developer perspective so no)
<ScottK> That's one of the guide's biggest problems at the moment. [08:44]
<Riddell> ScottK: but the guide should either be renamed or fixed for that
<ScottK> Agreed.
<Riddell> yep
<barry> ScottK: yep
<jelmer> maxb: It would seem reasonable to just directly tag the upstream
versions with upstream-X if they have the same contents
<poolie> by detecting there's a revision identical to the orig.tgz? [08:46]
<jelmer> either by detecting or having some place where the upstream tag
algorithm is specified [08:47]
<maxb> So perhaps the way forward is to discuss with the packagers of those
packages how much they are willing to tweak their workflow for UDD
compatibility
<poolie> can we do something towards recognizing the actual tag in the
upstream branch?
<poolie> there very likely is one, istm
<poolie> right
<jelmer> bzr-builddeb already has a config variable export-upstream-revision
which can be set to the tag pattern for finding upstream versions
<poolie> maxb, what kind of tweaks? [08:48]
<jelmer> e.g. "export-upstream-revision = tag:bzr-svn-$UPSTREAM_VERSION" as
bzr-svn's tagging algorithm uses the tag name "bzr-svn-1.0a" for
version "1.0a" [08:49]
<jelmer> The proper place to store that sort of information would be in
Launchpad's release information I think [08:50]
<poolie> that would be useful in other places
<poolie> for instance automatically presenting the releases in the ui and ws
<maxb> As in, potentially making use of "bzr merge-upstream", or setting tags
manually
<barry> it's getting late, so let's move on [08:52]
<barry> [TOPIC] bugs of interest
<MootBot> New Topic: bugs of interest
<barry> [LINK]
http://people.canonical.com/~mbp/kanban/canonical-bazaar-kanban.html
<MootBot> LINK received:
http://people.canonical.com/~mbp/kanban/canonical-bazaar-kanban.html
<poolie> good idea
<barry> anything in particular to point out?
<poolie> i plan to stop the package importer pretending to be james tomorrow
[08:53]
<barry> nice :)
<poolie> lots of bugs there
<poolie> moving across
<poolie> jelmer it looks like you should do a bzr-svn release [08:54]
<poolie> barry, james, maxb, is there anything especially biting you, or that
you think especially needs attention?
<jam> barry: [LINK] lp:///~jameinel/bzr/2.2-is-up-to-date
<jelmer> poolie: I agree; the tags memory usage issue is the main thing that's
blocking that at the moment.
<jam> [LINK] lp:///~jameinel/bzr/2.2-is-up-to-date
<MootBot> LINK received: lp:///~jameinel/bzr/2.2-is-up-to-date
<poolie> jam, what is that? [08:55]
<jam> I just added a "Branch.open" hook
<jam> so that doing "bzr info lp:ubuntu/bzr" tells you if the branch is
up-to-date
<poolie> oh i see
<poolie> great
<barry> poolie: atm, nope. things are going well
<poolie> we need to try again with a 2.3.4 sru into natty
<poolie> ok
<poolie> it's getting late here for sure
<poolie> next?
<jam> http://paste.ubuntu.com/643225/ [08:56]
<MootBot> LINK received: http://paste.ubuntu.com/643225/
<barry> [TOPIC] aob
<MootBot> New Topic: aob
<barry> just one small note: poolie the meeting 2 weeks from now is at the
wrong time i think
<jam> barry: I think that goes with the earlier action item (that I missed),
for checking if import branches are up-to-date
<poolie> yes, it does, that's nice
<barry> jam: thanks [08:57]
<poolie> ok
<barry> jam: that looks great
<poolie> so we're going to keep it at the same time as today?
<barry> poolie: if that worked for you guys. it was much better for me ;)
<jam> wfm, poolie is the one most impacted by it [08:58]
<poolie> ok
<poolie> thanks everyone for coming, especially maxb
<barry> okay, with that...
<barry> #endmeetingDistributedDevelopment/20110713 (last edited 2011-07-27 11:46:10 by mail)