== Meeting information == * #ubuntu-meeting Meeting, 16 Oct at 15:06 — 16:03 UTC * Full logs at [[http://ubottu.com/meetingology/logs/ubuntu-meeting/2014/ubuntu-meeting.2014-10-16-15.06.log.html]] == Meeting summary == === Lightning round === The discussion about "Lightning round" started at 15:07. === AOB === The discussion about "AOB" started at 15:27. === git-dpm === The discussion about "git-dpm" started at 15:30. * ''LINK:'' http://anonscm.debian.org/cgit/python-modules/packages/pycurl.git/ * ''LINK:'' http://anonscm.debian.org/cgit/users/barry/import-dscs.git/ == Vote results == == Done items == * (none) == People present (lines said) == * barry (115) * slangasek (56) * cjwatson (43) * mvo (30) * doko (20) * bdmurray (19) * sil2100 (18) * Caribou (13) * ubottu (10) * jodh (6) * bhuey (6) * meetingology (3) == Full Log == 15:06 #startmeeting 15:06 Meeting started Thu Oct 16 15:06:47 2014 UTC. The chair is slangasek. Information about MeetBot at http://wiki.ubuntu.com/meetingology. 15:06 15:06 Available commands: action commands idea info link nick 15:07 [TOPIC] Lightning round 15:07 $ echo $(shuf -e barry doko stgraber jodh bdmurray slangasek cjwatson caribou infinity mvo bhuey sil2100 robru) 15:07 jodh infinity mvo stgraber caribou bdmurray doko cjwatson slangasek sil2100 bhuey barry robru 15:07 jodh: you're up! :) 15:07 * system-image: 15:07 - Created an ubuntu-core-upgrader package. 15:07 - Writing documentation. 15:07 - Debugging mount issues. 15:07 - Currently improving upgrader. 15:08 ᧠ 15:09 PAHTAMASAT - ephemeris data for amateur satellites 15:09 no infinity today (Plumbing) 15:09 mvo: 15:09 apt: 15:09 - work on privsep code and backwards compatbility 15:09 - work on experimental branch 15:09 click: 15:09 - Address review comments for lp:~mvo/click/repository 15:09 - Debug vala dbus releated crash 15:09 - Improve table output for updates, add click update --machine-readable 15:09 - Look at bulk query interface to find updates 15:09 - Test sso+acquire branch with latest public.apps.ubuntu.com - success 15:10 - Work on lp:~mvo/click/sso+acquire - click update, install works now (with sso 15:10 ) 15:10 - fix click user-hook systemd job 15:10 misc: 15:10 - travel preparing 15:10 system-image: 15:10 - lots of work on this (gtimelog has 25 items for it this week :) 15:10 - image very usable at this point 15:10 utopic: 15:10 - ddtp-update 15:10 - command-not-found update 15:10 - app-install-data update 15:10 - upgrade test 15:10 - Debug/fix spectacular postinst upgrade failure during trusty->utopic (#1381570) 15:10 (done) 15:11 stgraber: around ? 15:12 Caribou: I don't think so 15:12 ok, I'll go ahead, he can always catch up 15:12 * Completed Python training in Paris 15:12 * sosreport 3.2 released on Debian 15:13 * First set of tests on networked kdump tools at run-level S 15:13 * Working on makedumpfile/kdump-tools test environment 15:13 * Need to ramp up on systemd 15:13 (done) 15:13 tested updated of daisy to revision 542 15:13 r545 daisy/submit.py: don't try to insert into systemoopshashes if system_token is missing, keep a metric of missing_system_tokens and whoopsie versions 15:13 submitted RT 75776 regarding update of daisy to r545 updated daisy code to resolve a KeyError trying to find HTTP_X_WHOOPSIE_VERION r546 15:13 Caribou: python 3 right? :) 15:13 r547 - daisy/submit.py: increment a counter for duplicate reports and a counter for duplicate reports including the whoopsie version 15:13 submitted RT to have daisy updated to revision 547 15:13 reviewed daisy log files for duplicate core submissions (numbers are much much lower) 15:13 why train Python in Paris? ;) 15:13 irc discussion regarding retracer backlog and declaring bankruptcy 15:14 submitted RT 75838 regarding declaring retracer queue bankruptcy 15:14 research into packages using /etc/os-release to resolve LP: #1362496 15:14 merged xnox's whoopsie changes fixing LP: #1339916 regarding system identifier 15:14 Launchpad bug 1362496 in base-files (Ubuntu-rtm 14.09) "LSB release and /etc/os-release still say "Utopic", needs to be RTM" [High,Fix released] https://launchpad.net/bugs/1362496 15:14 upload whoopsie to utopic 15:14 Launchpad bug 1339916 in whoopsie (Ubuntu RTM) "SystemIdentifier can change between reboots" [Undecided,New] https://launchpad.net/bugs/1339916 15:14 barry: unfortunately not but I stayed attentive to the differences 15:14 updated / uploaded lxc-android-config to make /var/lib/whoopsie writable 15:14 submitted apport merge proposal fixing LP: #1345569 15:14 uploaded apport to utopic fixing LP: #1345569 15:14 investigation into whoopsie test failure (callback-triggered-once) 15:14 reported glib2.0 bug LP: #1381804 regarding whoopsie test failure 15:14 Launchpad bug 1345569 in apport (Ubuntu) "recoverable_problem crashed with ValueError in add_proc_info(): invalid process" [High,Fix released] https://launchpad.net/bugs/1345569 15:14 Launchpad bug 1381804 in whoopsie (Ubuntu) "whoopsie test failure since glib2.0 2.41.2-1 uploaded" [High,Fix released] https://launchpad.net/bugs/1381804 15:14 doko: because I live near Paris & needed python training :) 15:14 ✔ done 15:14 heh 15:14 - Python 3.4.2 prepared for the trusty SRU 15:14 - a lot of binutils fixes 15:14 - slof ftbfs fix 15:14 - keyutils ftbfs fix 15:14 - crash ftbfs fix 15:14 - flask ftbfs fix 15:15 - refit ftbfs fix 15:15 - wagon2 ftbfs fix 15:15 - libaio ftbfs fix 15:15 - libunwind ftbfs fix 15:15 - openjdk-6 security update (utopic, trusty, precise, lucid) 15:15 - openjdk-8 update (still needs the aarch64 hotspot security updates) 15:15 - gcc-4.9.2 (something before the release candidate) 15:15 - binutils 2.25 branch 15:15 - cross toolchain updates 15:15 - the usual pestering about ftbfs, MIR, ... 15:15 - some syncs, removals, component overrides 15:15 (done) 15:16 Upgraded utopic to Perl 5.20.1. 15:16 Dragged in on Saturday for a weekend broken-images panic following Friday night's Mir landing. 15:16 Did a good part of the work to move a number of click packages from the ubuntu-touch rootfs into the custom tarball, which was an RTM blocker. 15:16 Spent a considerable amount of time analysing and thinking about the terrible bug 1265192. This resulted in four changes which I think cover all the bases: 15:16 bug 1265192 in ubiquity (Ubuntu Trusty) "Install/reinstall wipes out all/other partitions" [Critical,Triaged] https://launchpad.net/bugs/1265192 15:16 - Exclude free space from count of deleted partitions. 15:16 - Describe use_device consistently, avoiding language that is ambiguous in the event that OS detection gets it wrong. 15:16 - Offer separate replace option even if we believe that the disk only contains Ubuntu. 15:16 - Always show a confirmation dialog before committing partitioning changes. 15:16 Fixed (again) GRUB installation on PowerKVM, which lacks nvram. 15:16 Failed to finish native-dbus branch, but belatedly pushed a first draft so that Michael could have a look. 15:16 Various +1 maintenance work (mostly removals) in an attempt to prepare for release. 15:16 Off tomorrow to try to get my head together a bit before the sprint. 15:16 .. 15:17 native-dbus branch \o/ 15:18 slangasek: your turn o/ 15:18 bdmurray: regarding /var/lib/whoopsie, it was noticed yesterday that this fix isn't on ubuntu-rtm/14.09 yet, and it rather ought to be - we can't sync lxc-android-config because there's an earlier upload on utopic from stgraber that bumps an lxc dependency. We should talk about getting this landed 15:19 * slangasek nods 15:19 * short week, due to eating Canadian turkey and Canadian stuffing on Monday 15:19 slangasek: okay and also getting the new whoopsie there 15:19 * worked through a system-image server bug where it could not import deltas for images that dropped files with non-ascii filenames 15:19 * still working on embargoed security update which is awaiting vendor signatures 15:19 * with Colin and cwayne, finished splitting click core apps out of the rootfs... landed at the very last minute for our phone image release 15:19 * helped shepherding of landings and image prep yesterday to get our RTM milestone out (still being validated) 15:19 * internal list discussions around daisy data and how to use it to drive RTm 15:19 (done) 15:19 bdmurray: yep 15:19 sil2100: your turn :) 15:19 slangasek, cjwatson: really good work with the click-app split o/ 15:20 the good work was all cjwatson's 15:20 Works like a charm so far 15:20 - Landing team work, preparing landing e-mails 15:20 - CI Train maintenance and features: 15:20 * Fix problems with jobs succeeding on failure 15:20 * Fix problem with source package extraction, often seen in sync silos 15:20 * Handle errors during package builds more gracefully 15:20 * Preparing additions to the dual-landing functionality (not enabled yet) 15:20 - Coordination of many key landings for image promotion for ubuntu-rtm 15:20 - Preparing silo for the qtmir cherry-pick of unity8 lifecycle fixes 15:20 - Preparing and testing silo with the indicator-sound fix 15:20 - Writing a lot of announcements 15:20 - Pushing on fixes, keeping management up-to-date 15:20 - Preparations for travel 15:20 - Attending many meetings regarding our promotion plans for ubuntu-rtm 15:20 (done) 15:21 This week 15:21 -built icedtea prelease packages for the next security release. Work with Matthias to work around tarball problems 15:21 -wrote scripts to support easier analysis of jtreg test log. This is to replace the diff -u method I've been using and reporting 15:21 -commit the jtreg work directory to my ppa 15:21 -post jtregs result for utopic/trusty/precise 15:21 (done) 15:22 system-image: LP: #1373467 (on hold for...) LP: #1374459 (in progress) 15:22 Launchpad bug 1373467 in Ubuntu system image "Support config.d directory" [High,In progress] https://launchpad.net/bugs/1373467 15:22 Launchpad bug 1374459 in Ubuntu system image "Support alternative downloaders" [Low,Triaged] https://launchpad.net/bugs/1374459 15:22 debuntu: LP: #1295833; LP: #1380814; upstream pycurl issue #210 (debug callback UnicodeDecodeError on Python 3) - uploaded fix to Ubuntu. nose2_0.4.7-2ubuntu1 (remove unused tox B-D) and nose2_0.5.0-1 to Debian. LP: #1381564 (pyparsing 2.0.3+dfsg1-1 to Debian, awaiting landing for syncpackage to Utopic). 15:22 Launchpad bug 1295833 in Bazaar Fast Import "Import error in exporter.py - fastimport.helpers" [Critical,Fix committed] https://launchpad.net/bugs/1295833 15:22 Launchpad bug 1380814 in tox (Ubuntu) "[FFE] tox 1.8.0-1" [Wishlist,Fix released] https://launchpad.net/bugs/1380814 15:22 Launchpad bug 1381564 in pyparsing (Ubuntu) "pyparsing ParseResults.pop() fails with NameError: global name 'index' is not defined" [Medium,Fix released] https://launchpad.net/bugs/1381564 15:22 other: lots of trainguarding; citrain conference call 15:22 --done-- 15:23 robru: around? 15:24 any questions over people's status? 15:24 mvo: chsh> eew 15:25 slangasek: exactly, exploded really deep in the dependency chain, no fun 15:25 but fixed now, glad I found this before the release 15:25 * slangasek nods 15:26 how old is "older", btw? Should this have been caught by precise->trusty->utopic upgrades? 15:26 Fixed except that there's a systemd autopkgtest failure in the way, I think 15:27 Oh, possibly that's done. Not sure, the bug isn't closed at any rate 15:27 [TOPIC] AOB 15:28 anything else we should cover today? 15:28 (before barry gives us a brief presentation) 15:29 slangasek: I couldn't figure out when this changed, I don't know for sure, I can dig into it, I'm also puzzled that the precise->trusty->utopic has not found it 15:29 mvo: ok. fwiw I just checked my laptop, which has a uuidd user with /bin/false and a /var/log/installer that says it was installed with lucid 15:30 [TOPIC] git-dpm 15:30 * slangasek hands the mic to barry 15:30 thanks 15:30 \o/ 15:30 okay, so quick backstory: 15:30 slangasek: strange, just logged into a precise chroot and there my libuuid user has /bin/sh 15:30 slangasek: I will ask jibel about it 15:30 debian-python team uses svn to manage all team packages. the svn repos are debian/ only, i.e. not source-full 15:31 svn is pretty creaky these days so lots of folks want to use something more modern and distributed-y 15:31 git being the obvious choice 15:31 lots of open questions about a migration of team packages to git, and we did some experimentation with the options before debconf, and then had a meeting at debconf 15:32 when managing packages with git, first you have... git! 15:32 a lot of your package management can just be done with git commands 15:32 and git-buildpackage serves the same purpose as svn-buildpackage, which the debian-python team is quite familiar with 15:33 i.e. use that to build source packages and binary packages, etc. 15:33 tag for release 15:33 the tricky part comes in when you want to do patch management 15:33 you want a good interface with quilt since that's still the way you generally do patches against upstream in debuntu 15:34 there are two common choices here: 15:34 git pq 15:34 git-dpm 15:34 my first experience with git-dpm was with the six package, which cjwatson maintains in debian. it was quite nice 15:34 so i did some small conversions to both tools and found git-dpm to be so much simpler to use and teach 15:35 man git-dpm for lots of good details 15:35 (there's also dgit which is roughly equivalent to udd+bzr but we'll ignore that for now) 15:35 so, git-dpm 15:35 there's a third one that people keep going on about but isn't in Debian yet, right? git-cherrysoda or something? 15:35 you can use it to manage your branches, and we are recommending *source full* repos 15:36 git-cherrypick but i don't even think it's in the archive yet 15:36 or wasn't last time i looked 15:36 * slangasek nods 15:36 git-dpm has some very simple and well documented workflows for importing new upstream releases 15:36 (although i have some suggestions for improvements) 15:36 and let's say you need to add a quilt patch 15:37 git-dpm checkout-patches 15:37 git-debcherry I think it is 15:37 I still have the PDF queued up to look at ... 15:37 that puts you in a 'patched' branch, with only upstream source in your working tree (no debian/) 15:37 (git-stonefruit) 15:37 cjwatson: yeah 15:38 in the patched branch, you just edit the files as needed to fix whatever bug you need, then git commit as usual 15:38 when you're happy with your changes: 15:38 git-dpm update-patches 15:38 and now you're back in the master branch 15:38 oh yeah, 'master' is usually what's targeted for unstable, but of course you have other options 15:38 anyway 15:38 update-patches converts your 'patched' branch commits to quilt patches 15:38 it's all rather seamless and nice 15:39 though it is important to remember a few issues 15:39 1) your commit message is used in the patch name 15:39 e.g. 0001-fix-the-dumb-thing-that-upstream-broke.patch 15:39 2) each commit gets turned into a quilt patch 15:40 the latter means that in your 'patched' branch, it's helpful to sometimes do 'git rebase -i master' to squash commits, etc. 15:40 there are ways to control the q/patch name, and dep-8 headers are preserved, etc. 15:40 so is each patch always a single commit? 15:40 so it's really very nice 15:40 1) is fixed by using Patch-Name: in your commit message ... ah, yes, that 15:40 slangasek: each commit in 'patched' turns into a quilt patch file 15:40 cjwatson: yep 15:41 but it handles refreshing your patches and such 15:41 so that implies that each patch winds up rebased each time? 15:41 to evolve a patch over time, you tend to use rebasey workflows on the 'patched' branch, and then git-dpm merges patched into master 15:41 yes, i think so 15:41 so you rebase, but the history is preserved by way of the tip merge 15:41 yep. it's actually the first example of git rebase that i like :) 15:41 ok 15:41 that's what I was worried about - if someone screws up the rebase, is there a record :) 15:42 s/if someone screws up/if I screw up/ 15:42 oh yes, the 'patched' branch is temporary and local. unlike with git pq, it does not get preserved. git update-patches deletes it 15:42 slangasek: not sure, but i found it difficult to both screw up the rebase *and* get back on the master branch to build your package 15:43 so let's see... 15:43 oh yes 15:43 I mean screw up at a higher level (semantic failures rather than mechanical ones) 15:43 slangasek: not sure actually 15:43 barry: one thing I noticed is *not* to "rebase -i" on the master branch; I've seen quilt patch disapear that way 15:44 Caribou: sure, though sometimes you do want to rebase away a quilt patch. i've used it to squash commits so i have the right number of quilt patches 15:44 slangasek: it's certainly preserved. (git offers many ways to blow off your own foot, but a number of ways to stitch it back on as well.) 15:44 :) 15:45 slangasek: what I find myself doing is amending early and often 15:45 * slangasek nods 15:45 barry: indeed, rebasing while in the "git-dpm checkout-patched" mode is fine 15:45 slangasek: so I do the rebase in a sketchy way, test to see if it works, if it doesn't then try again and git-dpm update-patches --amend, and only push anywhere once I'm happy 15:46 yep, amend is great for working out the final details of a patch 15:46 so, quick example with what sealed the deal for me and git-dpm 15:46 http://anonscm.debian.org/cgit/python-modules/packages/pycurl.git/ 15:46 "test to see if it works" - that involves jumping back out to the master branch so you can do a package build? 15:46 rebasing the master branch is a good way to get very confused, although if you've pushed somewhere you can at least get it back from that, and there's always the reflog too 15:46 slangasek: right, update-patches but don't push 15:46 * slangasek nods 15:47 pycurl has ubuntu deltas which we need to preserve because of cross-pocket dependency constraints debian does not have 15:47 so, i built the debian version, tagged it and uploaded 15:47 then i created an 'ubuntu' branch 15:47 and made the ubuntu deltas there 15:48 i even tested some quilt patches, so `git-dpm checkout-patched` created ubuntu-patched (i.e. against ubuntu branch not master) 15:48 that was nice 15:48 anyway 15:48 once ubuntu version was ready, i made commits to the ubuntu branch, tagged it there with ubuntu/7.19.5-2ubuntu1 and created teh source package for upload to utopic 15:48 then i had a new upstream version 15:48 i did the debian twiddling as normal 15:49 and uploaded 15:49 switched to the ubuntu branch, merged in the master branch changes, updated the delta, and commited to the ubuntu branch 15:49 i was shocked how easy it was 15:49 neither bzr nor svn workflows can touch it 15:50 one last thing 15:50 this is a tool i wrote to import debian package releases into git: 15:50 http://anonscm.debian.org/cgit/users/barry/import-dscs.git/ 15:50 had some nice ipmrovements by tumbleweed 15:51 and while debian-python is still in limited experimental phase, i cringe when i have to go back to svn ;) 15:51 i think that's all i have. any other questions? 15:51 I haven't myself tried doing this with anything that has an Ubuntu delta, so I'm pleased to find out that that generally seems to work. It will be interesting to see how robust that is across non-trivial changes 15:51 oh, i did recommend d-python team switch to git-dpm, but we don't have an eta yet for mass conversion 15:52 yep 15:52 I'm not sure how the merge workflow would work (we wouldn't want it to accidentally serialise the patched branch onto master, for instance) 15:52 how about a tool to import udd branches into git-dpm? ;) 15:52 slangasek: frankly, i would just recommend making our lives simple and import-dscs but that does lose history 15:53 One thing I'd add, if you're converting non-trivial repository history into git (patch helpers or not), I can thoroughly recommend http://www.catb.org/~esr/reposurgeon/ 15:53 for debian-python, i really don't care about the svn history ;) 15:53 *preserving 15:53 It's basically a domain-specific language for repository conversions 15:53 barry: right; I'm rather attached to my detailed histories 15:53 actually, maybe that's more the case for other people's packages than my own, since I write changelog entries ;-) 15:53 I converted a ton of my history using it and have been very happy with the results I managed to get, such as the Debian openssh packaging that has been through cvs (with ill-advised vendor branch) -> svn -> bzr -> git 15:53 cjwatson: yes, have you actually run reposurgeon? esr says it needs a "beefy machine and lots of time" (that's for the emacs repo, which has crazy multi-vcs gobbledegook) 15:54 * slangasek bookmarks reposurgeon 15:54 yes I have. It took a while for grub2 I think but not so bad that I wasn't able to iterate a number of times on it 15:54 cjwatson: cool. i have upstreams that i want to convert to git eventually and reposurgeon is tops on my list for that 15:55 so for openssh I had a config file that looked like http://paste.ubuntu.com/8574593/ 15:55 specifically this allowed me to stitch the packaging history into the same commit graph as the upstream history, which is awesome 15:56 i've followed the discussion on emacs-devel. seems esr has done an impressive amount of work on reposurgeon 15:56 cool 15:56 * barry hands the mic back to slangasek 15:56 ok, 4 minutes left 15:56 any questions for barry? 15:56 I think that the only sane way to do this involves stitching in upstream history (assuming you have it), but that's really hard to do without a DSL 15:57 ("can we have this in Launchpad tomorrow") 15:57 barry: what're the opinions looking like in the rest of the d-python team? 15:57 cjwatson: i've only had limited feedback. ScottK was +1 ;) 15:57 no -1 15:58 there are some edge questions, such as wither upstream's repo should be remoted in, whether source full repos shoudl be used, that kind of thing 15:58 no one advocating for sticking with svn or using git pq 15:58 some people don't want to use pristine-tar 15:58 which i don't agree with 15:58 i.e. if upstream uses tarball releases, we should use pristine-tar workflows 15:59 and all pypi packages are still tarball based 15:59 git-dpm at least makes that easy to do 15:59 yep! 15:59 those aren't edge questions, those are lunatic fringe questions ;-P 15:59 though i want a --uscan option :) 15:59 slangasek: :D 15:59 (git-dpm import-new-upstream -p UPSTREAM-COMMIT --ptc) 15:59 "should source ful repos be used" - yes, always 15:59 slangasek: yep 15:59 remoting in upstream's repo can be a bit trickier if it's non-git 16:00 iirc you had some trouble with the hg setup in six 16:00 yep 16:00 though i don't recall the details 16:00 I think it required starting by cloning from upstream with git-hg and then remoting in the Debian branch, which isn't ideal 16:00 i'm personally not a big fan of remoting in upstream, but i ack that it can make cherrypicking fixes a little easier 16:01 it also lets you use git blame/log/etc. 16:01 i'd rather just grab the patch from an upstream clone or github 16:01 yeah 16:01 the *main* thing is - don't send irc and email notifications to d-python team for upstream commits! 16:01 haha 16:01 yeah, has that been fixed yet? :) 16:01 i think so 16:01 anyway, thanks for the talk, I'm really glad to see others using this 16:02 it's a life changer frankly 16:02 cjwatson: barry: seeing you using it convinced me to use it for my two projects 16:02 and i don't even hate git anymore :) 16:03 thanks, barry :) 16:03 #endmeeting Generated by MeetBot 0.1.5 (http://wiki.ubuntu.com/meetingology)