[00:00] This might tie in interestingly with the ideas presented in bug #532055 [00:00] Bug #532055: Trusted credential-management apps are broken and may be doomed [00:01] heh heh [00:02] Does anybody know what is happening with https://bugs.launchpad.net/bugs/529348? It's a trivial fix.. [00:23] is there any documentation for @cachedproperty ? I don't know why, if I use it, I get raise ClosedError("Connection is closed") in the tests [00:36] * mwhudson lunches [00:42] Good morning antipodeans and others [00:48] I was going to say hi to jtv [00:48] but he has no staying power :) [00:51] I think I used to be able to do this: push a branch to launchpad.dev... what am I missing [00:54] ah, it's documented! [00:54] jtv: yeah [00:55] bzr-svn is at least better than cscvs: after some 10 hour long failures, bzr-svn succeeds in 46 minutes: https://code.edge.launchpad.net/~vcs-imports/maxosx/trunk [00:55] w00t === bjf is now known as bjf-afk [02:31] hi stub [02:32] * stub waves groggily and moans something unintelligable [02:41] stub: I could murder a lumberjack... [02:41] cuppa? [02:41] (I've been at it for a few hours already... don't ask what got into me) [02:41] Maybe in a couple of hours... [02:42] I'm still stuffed full of roast beef and belgian fries [02:43] John's cooking? [02:43] need consciousness... [02:43] yer [02:45] Sorry I missed it... I grabbed some delicious "hello sir" from 3/1 to eat while watching a film at home with the missus—her idea—who then decided to skip the film so as to get up early for the temple this morning. [02:45] TIT [02:46] stub: meanwhile, I'm just moving some stuff over to the slave store. [02:59] jtv: I could probably be arsed dragging myself down for a coffee now [03:02] stub: good, because what I just cooked up looks too horrible to swallow [03:02] stub: I can be on a motoecy in a few minutes [03:03] * stub crawls off looking for his pants [03:39] I just love seeing XXX comments with text like "I have no idea what the code below is meant to do." [03:46] wgrant: soyuz code i presume? [03:46] mwhudson: But of course. [03:47] thumper: around to review a branch? [03:48] 1 whole line of changes!! [03:52] mwhudson: yeah [03:52] mwhudson: I have one for you too [03:53] thumper: https://code.edge.launchpad.net/~mwhudson/launchpad/git-http-imports-bug-438929/+merge/20716 [03:53] thumper: ok [03:54] mwhudson: actually, you know what? [03:54] mwhudson: it can wait [03:54] mwhudson: otherwise I'll be proposing against the wrong branch [03:54] mwhudson: just to not show you the db-devel changes [03:54] mwhudson: when db-devel is merged into devel it'll make more sense [03:55] thumper: heh, ok [03:55] thumper: my rc branch landed fine btw [03:55] mwhudson: yours is done [03:55] cool [03:55] mwhudson: which one was that? [03:55] thumper: the one fixing xmlrpc requests in read only mode [03:56] mwhudson: ah, was that what was screwing us over? [03:56] thumper: which screwage do you mean? [03:56] but no, probably not [03:56] mwhudson: about connections hanging around [03:57] thumper: no [03:57] oh [03:57] at least, that seems very unlikely [03:57] so what was the change? [03:57] thumper: it just fixes loads of spurious oopses [03:57] mwhudson: fair enough [03:57] mwhudson: how's your afternoon going? [03:58] thumper: well, i was being fairly unproductive, so i decided to fix a couple of trivial code import bugs [03:58] * thumper nods [03:58] I've been watching the code imports fairly closely [03:58] that git one and https://bugs.edge.launchpad.net/launchpad-code/+bug/513182 [03:58] Bug #513182: +code-imports should show VCS type [03:58] we seem much better now [03:58] cool [03:59] I've been making the description changes get emailed :) [03:59] cool [04:05] mwhudson: what if we renamed +junk to +branch? [04:05] confusing? [04:06] seems a bit random, somehow [04:06] it doesn't really distinguish it from the other, um, branches [04:06] If you change +junk (and I've seen some chatter about that bug), could you leave an invisible redirect for the various immutable docs out there (mostly email archives) [04:07] surely [04:07] Wonderful :) [04:08] I've been trying to think of a meaningful anem [04:08] name [04:08] +own [04:08] +my [04:08] +personal [04:08] all a bit weird [04:08] Weren't there some suggestions in the bug? If not, there were certainly some in -motu and I can dig up logs. [04:08] I want to make sure it still makes sense with a team branch [04:08] persia: there were [04:09] persia: discussed in -motu? [04:09] That bug was. [04:09] (someone found +junk offensive, and ranted for a bit, and a few names were thrown out) [04:09] I'll see if I can find the log. [04:10] bug 147407 [04:10] Bug #147407: Junk sounds too harsh [04:10] thumper: thanks for making rockstar do make run_codehosting btw :) [04:10] +user [04:12] Which doesn't work for teams. And the conversation I remember ended being a tutorial on how to stick a branch on people.ubuntu.com [04:12] +misc [04:12] +code [04:13] I think I'd rather have +branch over +coe [04:13] +code that is [04:13] the branch to change this will be relatively simple, and I'd like to get it done [04:14] I'm going to go with +branch [04:14] hi thumper [04:14] and see what happens [04:14] hi poolie [04:14] I'm just trying to walk out the door [04:17] * thumper afk [04:17] Odd. Somebody added LFC.sha256 a year ago, but didn't add the three extra lines to actually get it populated for new files. [07:26] al-maisan, noodles775: Since LP again won't have emailed anybody, have a poke about https://bugs.edge.launchpad.net/bugs/532445 and https://bugs.edge.launchpad.net/bugs/532454 [07:27] wgrant: thanks, looking. [07:50] Hi wgrant! ;) [07:50] Morning henninge. [07:51] wgrant: thanks for the help the other day, I got the buildd to work and command it via rpc. [07:51] henninge: Excellent. [07:51] henninge: I see I forgot the ensurepresent call. I'm glad you worked it out in the end. [07:52] wgrant: yes, I got that from the code ... ;) [07:52] Being able to see the code is handy. [07:52] wgrant: that chroot you pointed me to, is that the same that is used on the production builders [07:52] or will be [07:52] ? [07:52] henninge: It's the same. [07:53] wgrant: Can I adapt it to my purposes? [07:53] henninge: You can now directly get the current URL from https://launchpad.net/api/devel/ubuntu/lucid/i386/chroot_url [07:53] henninge: Please don't. [07:53] Your script should be able to run on the chroot you get from LP. [07:53] Unless we want to start having separate chroots for each job type, which we have so far avoided. [07:53] and would like to continue to avoid. [07:54] You'll see that the recipe build job installs the packages that it needs itself. [07:54] yes, I was trying that, too. [07:55] I get "permission denied" when trying to run apt-get. [07:55] wgrant: what do you mean by "recipe build job" ? [07:55] Sounds like you are running that from inside a script that runs as the user. [07:56] so, what user does the buildd run as? root? [07:56] henninge: See lib/canonical/buildd/buildrecipe [07:56] It runs as buildd, but can sudo. [07:56] That's what I am doing. [07:57] wgrant: there is no sudo in the chroot, though [07:57] henninge: No. You have to sudo from outside. [07:58] wgrant: so should the build job inside the chroot do everything as root? [07:58] henninge: Probably not. [07:58] You need a script running outside the chroot to: [07:58] 1) sudo into the jail and install your dependencies [07:59] 2) jump into the jail as the normal user, and run pottery or whatever it is you want to do. [08:00] wgrant: oh, I think I was not aware that I don't need to be root to call chroot. [08:01] henninge: You do need to be. [08:01] henninge: see the 'su -c' calls in buildrecipe. [08:01] That is run inside the chroot to switch back to the right user. [08:01] looking at that now [08:01] So you end up doing 'sudo chroot su -c "whatever command" buildd' [08:01] oh, forgot about su with all the sudoing ... [08:01] er, of course there's the chroot path in there somewhere. [08:02] Yeah, it's horrifyingly messy. [08:02] yes, shure [08:02] s/h// [08:02] 'shure' makes more sense! [08:02] ;-) [08:02] wgrant: thanks, I feel a lot more enlightened [08:03] henninge: Excellent. [08:03] You may still beat recipes to production.. [08:15] wgrant: I see that buildrecipe installs sudo in the chroot. I uses both sudo -u and su -c [08:15] henninge: Heh, I wonder why. [08:16] * wgrant looks. [08:16] Ah, right, I remember from the sprint now. [08:16] The environment variables. [08:16] wgrant: sudo is nicer to handle with call [08:16] Yeah. [08:17] It was going to be risky and difficult to escape them and stick them in a string for su -c. [08:19] wgrant: also, it is a bit hard to try out the installing because the sources.list in the chroot points to ftpmaster.internal [08:19] I'll just have to overwrite that [08:20] henninge: You can pass a list of sources.list entries in the build call. [08:21] Assuming that you're using the normal chroot infrastructure code (unpack, override sources.list, upgrade). [08:21] hm, I have not yet seen the "overide sources.list" part [08:22] Add to the extra_args of your build call: "'archives': ['deb http://blah/ubuntu blah blah blah', '..']" [08:22] I see [08:27] ok, found it [08:27] wgrant: I am not using the DebianBuildManager btw. Maybe I should. [08:27] Where are you looking? (/me wants to follow along as much as I can) [08:28] henninge: I think you should. [08:28] Let me look. [08:28] I created it, but that was almost two months ago now so blaaaah. [08:28] Yeah, you should probably use it. [08:28] noodles775: DebianBuildManager.doSourcesList [08:28] henninge: ta. [08:29] noodles775: it calls slavebin/override-sources-list [08:29] wgrant: ok, let me look into that [08:29] henninge: If you want to see how normal builds work, look at the build logs on production. [08:30] wgrant: where do I see them [08:30] ? [08:31] https://edge.launchpad.net/ubuntu/+builds?build_text=&build_state=built [08:31] wgrant: cheers [08:33] oh yeah, very interesting [08:35] good monring [08:37] henninge: What's still to do with the slave and master template code? [08:37] wgrant: what do you mean? [08:37] adeuring: moin [08:38] henninge: Well, what are you doing to it? [08:38] moin henninge [08:38] I thought it was mostly working. [08:38] wgrant: well, obviously it's not ... [08:38] Heh, yeah, I've just looked at the stuff in the tree and seen that it doesn't do much. [08:39] It's obviously never been run like I am trying to run it now. [08:40] Right. [08:40] I have added the call to the actual template generation code but getting it to be run is my challange now. [08:40] wgrant: but I think I am getting there [08:40] The call to that code seems to already be there. [08:40] Although the filename is wrong. [08:41] you mean "generate...py"? [08:41] And you can save an awful lot of code by using DebianBuildManager. [08:41] Yeah, that. [08:41] that's just a dummy, too. [08:42] wgrant: oh no, sorry [08:42] I already landed that part ... ;-) [08:42] It doesn't look like it will actually check out the branch, though. [08:42] And it imports from lp.*, so it won't work in the slave. [08:43] wgrant: I added stuff to debian/rules [08:43] so the importing works [08:43] henninge: Oh. [08:43] That seems a bit ugly. But I guess the whole slave thing is pretty ugly. [08:44] My thinking ... [08:44] ;) [08:44] henninge: Is there a reason for that file to live in lp.translations at all? [08:45] wgrant: I was trying to mimic the way tachandler was included. [08:46] it seemed to replicate the structure from the canonical tree, so I did the same for the lp tree [08:46] wgrant: the code is only needed in the buildd, so it could live in canonical/buildd . [08:46] henninge: tachandler is copied in because it's used on both sides. [08:47] So I would just put your file in lib/canonical/buildd and be done with it. [08:47] I guess that's true. [08:48] wgrant: but it seems nicer for all translation code to live lp.translations ... [08:48] * wgrant is very tempted to split up lib/canonical/buildd's contents into subdirectories. [08:49] Into what for example? [08:49] good idea [08:49] per app? [08:49] translations, recipe, binary [08:49] Right. [08:49] Per build type, yeah. [08:49] And a general one. [08:49] Since the dir is getting pretty big. [08:49] And there is obvious separation. [08:49] +1 [08:50] +1 [08:50] * henninge has to relocate, back in 20 min max. [09:13] Morning [09:15] * henninge is back [09:20] mrevell, good morning [09:20] Good morning jml [09:33] I'm having some trouble with codehosting on my dev machine... I "make run" and "make run_codehosting." I push a branch to lp://dev/... I make changes to the branch, and when I push again, I'm told the branches have diverged. [09:33] Anyone know what I'm doing wrong? === jtv1 is now known as jtv [09:39] jtv: is make run_codehosting new? I just do make_run [09:39] used to, I mean [09:39] henninge: I believe it's new, yes [09:40] killall -9 firefox [09:40] ahhhh [09:41] I should put that in cron.hourly [09:44] stub: I believe you said tests were going back to a single store... do we have a bug for that? [09:45] launchpad is, not just tests [09:45] single replication set [09:45] How are lp_* going to be replicated to c-i-p? [09:46] I'm not counting that as part of launchpad ;) [09:46] stub: oh, so still replicated... Using the slave store will require some extra commits in tests then. [09:46] Yup. [09:48] didrocks, hi [09:48] hey jml [09:48] jml: sorry, but yesterday, it was like hell :) [09:48] didrocks, understood [09:48] jml: I pulled your branch and get scared by the amount of change :) === noodles785 is now known as noodles775 [09:50] didrocks, wgrant makes a very interesting objection -- if we have an API for adding, say, SSH keys, then a malicious application that gained your approval could add a key, then use that key to get your branch data even after you revoked privileges from the application [09:51] jml: I think modification of that belongs in the trusted client. Not the cleanest solution, but just about anything else is unsafe. [09:51] jml: GC already does that manually (upload a ssh key for you). I'm not sure not exposing that will prevent this usage [09:52] didrocks: People should soon be educated to not give their password to such applications. [09:52] And that will stop working soon. [09:52] I hope. [09:52] wgrant: right :) [09:53] that's just too much hope on the user side I'm afraid against a promise for a "wonderful application making your life better (and spamming you, destroying your data and so on…)" [09:54] but that's just my opinion :) [09:57] didrocks, anyway, I'm not sure how to proceed. The work we've done already is good & useful and should be merged. [09:58] jml: right, but unfortunately, that won't save the "I don't achieve to upload my gpg/ssh key or uploading wrong one, and so on…" [10:01] didrocks, hmm. [10:04] adeuring, I notice that the bugs with patches view includes fix released bugs [10:04] jml: yes [10:04] adeuring, is this deliberate? [10:05] yes [10:05] ...but I can't remember the exact reason. Probably because "fix released" is intersting for upstream [10:05] kfogel might remember better ;) [10:05] ok. I'll ask him. [10:07] Doesn't that just mean it's going to get longer and longer and become less and less useful? [10:07] wgrant: I think DebianBuildManager has too much specific stuff that I don't need. Any reason doSourcesList, doUpdateChroot, doReapProcesses should not be in BuildManager instead (and their confgis moved to allmanagers) ? [10:07] henninge: What specific stuff don't you want? [10:08] wgrant: what's ogre model btw ;-) [10:08] henninge: It's not used any more. The master used to pass in a component rather than a full sources.list [10:08] ogre model refers to the layering of the components. [10:08] main includes just main. restricted includes restricted and main. [10:09] universe includes universe and main. [10:09] multiverse includes multiverse, universe, restricted and main. [10:09] ah, I see. [10:09] wgrant: changes file handling for example I don't need [10:10] gatherResults will work differently for me, too. [10:10] Yeah, so just gatherResults, _parseChangesFile and getChangesFilename? [10:10] wgrant, henninge: the FSM for DebianBuildManager is also a lot more extensive than we need, isn't it? [10:10] jtv: yes, that's another thing [10:11] jtv: Why? [10:11] I think you need everything there. [10:11] wgrant: our job is simpler... no need to install build dependencies, for one [10:12] DebianBuildManager doesn't handle build-dependencies AFAIK. === daniloff is now known as danilos [10:13] Then who installs them before building a package? [10:13] jtv: sbuild in the binary case, or the recipe build script in the recipe case. [10:13] oh, it's further down in the hierarchy [10:13] Yes. [10:14] So is DebianBuildManager not limited to building Debian packages then? [10:14] The branches we work with don't necessarily have a debian/ dir etc. [10:14] jtv: no, it doesn't look that way [10:16] wgrant: Yes, you are right with that list (gatherResults...) [10:16] Looks like [10:16] henninge: Why not just override gatherResults? [10:16] Yes, that does seem to make sense [10:16] wgrant: yes, the obvious next step [10:17] You already need to do similar things on the master side, where we've made assumptions in superclasses that you have a build. [10:17] So you might as wlel make it easy for yourself and just override the two methods. [10:17] correct [10:17] (the one to do the actual building, and gatherResults) [10:17] wgrant, jtv: yes, I will do that. [10:18] thanks [10:18] It does mean that DebianBuildManager is a total misnomer [10:18] (what's this Ogre thing btw? Do we need that at all?) [10:19] jtv: It doesn't. You're building in a Debian chroot. [10:19] You don't care about that. We should remove ogre support. [10:19] ahh, Debian only in _that_ sense [10:19] It hasn't been used in nearly three years. [10:19] So... what build managers do we have that are not debian build managers in that sense? [10:19] * jml rebooting [10:19] None at this point. [10:21] But there could quite conceivably be some later. [10:28] stub: Do you know why there is a LFC.sha256 column, but it's not actually being populated? [10:29] because soyuz team decided they needed it, added the column, but didn't need it enough to make the librarian actually store it when files are uploaded. [10:29] lol [10:29] Odd, since it was about 3 lines extra to do it. [10:30] Yup. Also about 3 lines to drop the column again... [10:30] But we'll need the column soon. [10:30] We should be using more than MD5 in the indices. [10:30] I've heard that for maybe a year now ;) [10:30] This is Soyuz. Time is different here. [10:31] my clock says it's 1984 [10:31] So three line fix to the librarian to store the sha256. A quick and ugly data migration script to fill in the values for the existing files. A DB patch to set the column to NOT NULL. [10:31] Party like its 1999 [10:32] Quick, ugly, and several-week-taking? [10:32] didrocks, so what are we going to do? [10:32] wgrant: Its only a few terrabytes - shouldn't take too long. [10:33] Was SHA1 there from the start? [10:33] bigjools, "It was a bright cold day in April, and the clocks were striking thirteen." [10:33] I think it was, yes. [10:34] So it's not been done before :( [10:34] jml: we need to talk! [10:34] jml: that's a good question… :/ [10:34] Its pretty trivial. I can hack up the migration if nobody else wants to. [10:34] bigjools, we do. [10:34] bigjools, lemme book a time [10:41] bigjools, 12:30 ok? or we can talk now until 11 [10:41] jml: no 12:30 is bad, I have a hospital appt at 1 [10:41] so now is good [10:41] ok. I'll skype you. [10:53] henninge: you may have to pretend that the dir with your packages is an apt cache... something like -o Dir::Cache=/tmp/extra-packages [10:54] jtv: let me try that out [10:54] was I stopping you? [10:55] henninge, jtv: Alternatively, upload them to a local or real PPA. [10:55] What custom packages are there? [10:56] wgrant: no, it's launchpad-buildd [10:56] wgrant: I am using pbuilder to try it out and want to install it in there [10:57] actually, pbuilder might be able to do it, too [10:57] henninge: Ahh. [10:57] pbuilder login [10:57] exactly [11:06] henninge: does that solve it? [11:06] Morning, all. === al-maisan is now known as al-maisan-lunch [11:06] jtv: havn't gotten that far, got distracted. Read it on the wiki page next week ... ;-) === al-maisan-lunch is now known as al-maisan [11:07] henninge: I'm also looking at ways to do this [11:07] henninge: ah, you are capturing all this on a wiki page? Great! [11:07] noodles775: not yet but we just decided to do it === Guest4641 is now known as joey [11:10] henninge: haven't found anything better than adding a "file:" URL into the slave's /etc/apt/sources.lists.d [11:10] hi joey! [11:10] hi jtv [11:10] jtv: henninge: Why can't you just dpkg -i it? [11:10] wgrant: need to satisfy the deps as well [11:10] wgrant: I want dependencies to be pulled automatically [11:11] jtv: dpkg -i blah.deb; then 'apt-get -f install' [11:11] eh [11:11] I guess that'd work too :) [11:11] I thought I described this on the wiki page. [11:11] * jtv facepalms [11:11] wgrant: which wiki page is that? [11:11] There's so much text out there, sometimes it's like not having any [11:11] wgrant: which wiki page [11:11] jtv: HowToUseSoyuzLovally [11:11] uh-oh [11:12] s/v/c/ [11:12] Lovely [11:12] ;) [11:12] Or Run or whatever it is. [11:12] Heh. [11:12] Use, yes [11:13] wgrant: I thought henninge was going to document something like, extending the build-system for translations. [11:13] noodles775: that's next [11:13] Great. [11:13] But first, "running a local build slave" [11:14]