[00:33] benji: please don't land that python-pgbouncer branch. [00:33] benji: until the question I raise in the review is resolved. [00:36] benji: pinging here incase you didn't see the email is all. [00:51] WHere did launchpad-buildd move to? Its no longer under lib/canonical/buildd nor can I find it in the source tree [00:57] It was split out [00:57] lp:launchpad-buildd [00:57] Found it, thanks [00:58] NCommander: lib/canonical is also emptyish too [00:58] * NCommander notes the wiki pages really could use an update [00:58] That was known. [01:16] StevenK: so for doing livecd builds on the buildds, I already determined we need a new custom livecd type (raw-livecd). My current idea is to have the CD build infrastructure connected to a devirtualized PPA, and it merely needs to dput a dummy package to spin the livefs's. Does this sound sane? (alternatively, I can have it work in the main archive, but then we have to deal with releasing live images through the queue which will be k [01:16] You got cut off [01:16] 21:16:02 < NCommander> StevenK: so for doing livecd builds on the buildds, I already determined we need a new custom livecd type (raw-livecd). My current idea is to have the CD build infrastructure connected to a devirtualized PPA, and it merely needs to dput a dummy package to spin the livefs's. Does this sound sane? (alternatively, I can have it work in the main archive, but then we have to [01:16] dall with releasing live images through the queue which will be kinda annoying) [01:17] alternatively, I could write a new API to the buildds to build the squashfs's indpendently, but that sounds like a *lot* more pain than its worth. [01:17] NCommander: So, I'm not sure. [01:18] NCommander: doing a dummy package to trigger builds is fugly and unneeded. [01:18] NCommander: where will the [d]debs for the livecd come from ? [01:18] ftpmaster.internal, I guess [01:18] lifeless: ftpmaster.internal. I already managed to spin a squashfs in a PPA [01:18] ok [01:19] so, there are already 3 sorts of builds [01:19] (I needed a helper package so I could run livecd.sh as sudo, but other than that, no headaches (right up until I hosed the uploader :-)) [01:19] recipes, translations, packages [01:19] add a forth type [01:19] recipes? [01:19] recipes run on the buildds [01:19] is this documented somewhere? :-) [01:20] It's probably under buildmaster [01:20] lib/lp/buildmaster, that is [01:21] Actually most of it is in launchpad-buildd :-/ [01:22] NCommander: buildd-manager needs to know about the types of build at least [01:22] (and at least for TRANSLATIONTEMPLATES its in lp/translations; this code is all over the $#!@# place) [01:22] Yeah, ther'es a enum.py which is usually the first place you go so your test cases don't explode with unknown identifer [01:22] SPRecipe is probably in lib/lp/code too [01:23] I've forgotten its identifier [01:23] * NCommander feels like he's about to be in a lot of pain [01:23] Well, duh [01:23] Doing CD builds *properly* via LP is a lot of work. [01:23] And you have LoC constraints [01:24] LoC? [01:24] Lines of Code [01:24] lines of code? [01:24] * NCommander looks at the wiki [01:24] NCommander: New features now have to be LoC neutral [01:24] WTF? [01:25] what does LoC neutral mean? [01:25] NCommander: If you want to add 500 lines of feature code, you have to find 500 lines of something to remove first. [01:25] ... [01:26] rm -r lp-branches/devel/lib/lp/soyuz/tests :-P [01:27] Hah [01:27] NCommander: We have a tool that cjwatson wrote. For example: [01:27] steven@undermined:~/launchpad/lp-branches/devel% loc-contributions 'Steve Kowalik' [01:27] -835 [01:27] You can get LoC waivers if appropriate [01:28] so, let me get this straight. To successfully commit code to Launchpad, I need to step back and delete something else? [01:28] Or you get an exception from lifeless. [01:28] I rather just modify ubuntu-cdimage to build a deb in a PPA and then wget the deb. a *lot* less pain [01:28] or in some other way offset the increased cost of running and maintaining launchpad that the new code brings. [01:29] lifeless: adding one machine of every architecture to the buildd pool increasing relaibility and decreasing live image build time? [01:29] (the livecd builders can be decommisioned and turned into regular buildds if we don't need dedicated machines) [01:29] StevenK: flacoste hands out exceptions too IIRC> [01:29] NCommander: as afraction, how much is the increase there? [01:30] NCommander: how does it decrease live image build time ? [01:30] lifeless: allows parallization of images across flavors, load balancing, and decreases issues if a buildd machine needs to go offline for service [01:31] As it stands, we can build one flavor across all architectures at any given time (ARM has two livebuilders with a rather nasty hack to use both) [01:31] whats the utilisation % of the current livebuilders ? [01:32] 6-8 hours per day on dailies. Can go up to 100% during freezes if large amounts of respins are required [01:32] Spread across 6(?) machines, one per arch one an additional one for armhf [01:32] how many machines are there in the current buildd pools ? [01:32] For LP, or livebuilds? [01:32] two minimium per architecture (I believe we're down a hppa builder, I'll check) [01:32] LP [01:32] * NCommander checks the current numbers) [01:33] hppa is dead to us [01:33] Will Ubuntu want move flavors or more liveimage builds once the latency is reduced ? [01:33] StevenK: don't tell lamont [01:33] lifeless: that is a discussion the tech board must make. Currently we're allowing new live images on all archs aside from ARM [01:33] (the ARM builders barely keep up with the load as is) [01:34] 5 i386, 3 powerpc, 3 amd64, 2 ia64, 2 sparc, 1 hppa, 1 lpia and 12 armel, 7 armhf [01:34] Hell, lpia is more dead to us than hppa [01:34] ARGH, BAD MEMORIES [01:34] NCommander: Worse for me [01:35] StevenK: well yes but lpia wasn't as bad as psb [01:35] Don't use that language with me. [01:35] >;-) [01:35] Crown Beach is so toally dead to me. [01:35] *totally [01:36] StevenK: I recommend high explosives personally. Helps get rid of the burn [01:36] Ok, so we're looking at 16 hours more build time/day across the board [01:36] which is 15% or so on i386, 20% on amd64 [01:36] lifeless: offset by the fact that every architecture will almsot certianly get an additional buildd (once the livebuild pool is decommissioned) [01:36] NCommander: that *is* the additional buildd [01:36] NCommander: 24-8 = 16 [01:37] d'oh [01:37] this will make it easier to argue for more scalable capacity too perhaps. [01:38] I'm inclined to grant an exception with a caveat: Please *try*, after the project, to reduce LoC impact. [01:38] but if you can't, that will be ok. [01:38] lifeless: I will, but I'm a heavy commenter [01:38] SO as long as its only code I need to elimate, I will try [01:38] NCommander: cjwatson is at about -4K lines. [01:38] NCommander: refactoring tests is a big win [01:38] NCommander: there is a huge amount of FAT in the LP code base. [01:39] StevenK: personally, I'd like to add more tests to Soyuz [01:39] * NCommander sucks [01:39] *ducks [01:39] fuck [01:39] Hahaha [01:39] NCommander: lots and lots and lots of places that things are redundant, or wasteful. [01:39] * StevenK quotefiles NCommander [01:39] lifeless: personally, I'd love to see LP not use 300+ queries to upload a package, but that's just me [01:39] NCommander: see above under fat. [01:40] yeah [01:40] Thanks for the LoC exception. [01:40] NCommander: From our OOPS reports, statement counts: 1350 OOPS-1db714aef8b3b55ac9a7a58cc336ebc4 BugTask:+index [01:40] Ow [01:41] * NCommander notes he uses launchpad as proof that postgresql will scale just as well as oracle [01:41] 1054 OOPS-861cfa9611c5fbb82f3b97998ea1dcf1 Archive:+copy-packages [01:41] But Colin is working on that one [01:41] 10000 OOPS-bfc5975ff6ecfe11483e9b28b0f772bf Unknown [01:41] * StevenK blinks [01:41] * StevenK looks up that OOPS [01:41] NCommander: so strictly, we don't insist on before-landing-credit, just that folk keep themselves honest and work on overall shrinking the maintenance burden. [01:42] StevenK: why do I feel like that OOPS hit some sorta query limiter [01:42] NCommander: what I'm saying to you is I accept the argument that the project is itself a burden reduction, but you should still try to address code fat as you do stuff :) [01:42] We don't have a query limiter [01:42] lifeless: I may submit mini-branches on the side to delete cruft on the side [01:42] That's an oddly round number then [01:43] Right [01:43] I'm loading the oops [01:43] Well, trying to [01:43] * StevenK prods neem [01:43] it will take a little [01:43] 10K backtraces. [01:43] * NCommander will see StevenK at UDS still waiting :-) [01:43] NCommander: I wish [01:45] NCommander: for instance: [01:45] https://code.launchpad.net/~cjwatson/launchpad/uefi-ppa-no-unapproved/+merge/119542 [01:45] steven@undermined:~/launchpad/lp-branches/force-ibug-into-line% bzr di -r 14918.. | diffstat -s [01:45] 14 files changed, 519 insertions(+), 753 deletions(-) [01:45] NCommander: -5 LoC, clearer tests, and bug fix. [01:46] That's one of mine. I fixed IBug so it no longer read like a bad romance novel and made the ZCML clearer. [01:47] HAhaha [01:47] lifeless: Just got a ISE from neem [01:47] \o/ [01:48] "Please contact the server administrator, [no address given]" [01:50] Hah, and the error log is useless [01:50] Thanks for being so helpful, WSGI [01:50] StevenK: it should have oopsed [01:51] StevenK: it may still be processing though. [01:51] I have no idea, I got a generic ISE page [01:51] The error log just says "Premature end of script headers" [01:54] yay. Not. [01:55] lifeless: So, I guess it didn't OOPS? [01:55] I guess not. [01:55] the oops occurs when the thing propofates up [01:55] no script headers means the app didn't send any before apache flicked it off [01:56] wgrant: Can haz review? https://code.launchpad.net/~stevenk/launchpad/no-oops-request-daily-build/+merge/119654 [01:58] StevenK: Busy now, sorry. [01:58] Possibly this afternoon [02:33] lifeless: Why more OOPS summaries? [02:34] StevenK: Why indeed? [02:34] StevenK: I have nil idea. [02:34] Ask -ops [03:50] wgrant: Are you unbusy now? [03:51] StevenK: Not really [03:51] :-( [03:52] Busy benchmarking paranoia [04:23] lifeless: AH, come here [04:37] StevenK: ? [04:39] lifeless: http://pastebin.ubuntu.com/1148122/ [04:40] lifeless: Not sure why mailman is there, given lib/mailman. [04:41] lifeless: But the main reason I grabbed you is because the last time I tried to rip subvertpy out of sourcecode, things went bang. [04:45] trlololol https://plus.google.com/109613336078485786201/posts/1mC8Wm37W2g [04:45] oh sorry. right. [04:45] StevenK: lib/mailman is built code [04:45] StevenK: (yes, our build system writes into lib/) [04:45] StevenK: you can't remove mailman like that. [04:45] StevenK: :( [04:46] StevenK: but testresources should be fine. [04:49] lifeless: So I put subvertpy into download-cache, add it into versions.cfg, remove it from sourcedeps.conf, and make, and then nothing can import subvertpy [04:49] What steps have you taken t debug this ? [04:49] None, because I'm not sure what steps to take. [04:50] ok [04:50] so when something can't be imported [04:50] it may mean its not on the python path [04:50] or it may mean that something is broken in the thing [04:50] is there an egg? [04:50] It's a tarball [04:50] is the egg on the path when you look at sys.path under bin/py ? [04:50] StevenK: tarball is whats distributed. [04:51] eggs are what live in eggs/ [04:51] eggs/subvertpy-0.8.10-py2.7-linux-x86_64.egg exists [04:51] Oh, wait, that's the old one [04:51] So, no, no egg [04:51] well [04:51] tried running bin/buildout ? [04:51] make clean && make should do that, no? [04:52] have you removed lib/subvertpy ? [04:52] I didn't know that was there. Sob. [04:52] I note your pastebin is missing the remove of lib/testresources [04:52] Right [04:52] I'll do that too [04:52] (this is why removing sourcedeps is such a great idea :)) [04:53] I noticed mustache quietly died from sourcedeps [04:53] I nuked. [04:53] I think. [04:53] Therefore I delete. [04:53] Well [04:53] Um [04:53] It was better in sourcedeps than where it is now :) [04:54] lifeless: Deleted lib/subvertpy and no dice. Still no egg [04:54] no, it was rick that did it. [04:54] StevenK: ok, and you ran bin/buildout ? [04:55] StevenK: (don't answer by talking about make :)) [04:55] StevenK: does 'python -c "import subvertpy"' work for you ? [04:55] Just ran bin/buildout and still no egg [04:55] StevenK: ok, and is subvertpy listed in setup.py ? [04:56] StevenK: Did you define it as a dep? [04:56] Yaeh [04:56] in setup.py [04:56] Ah [04:56] Let me do that [04:56] Now, you may say 'why?' and thats a good question. But until we get the *bzr deps* into setup.py, buildout won't traverse through them to find subvertpy. [04:58] lifeless: Right, so indeed, putting it in setup.py was the answer. [04:58] cool. [04:58] I would add a comment explaining why. [04:59] But this branch is at -3/+3 :-( === almaisan-away is now known as al-maisan === al-maisan is now known as almaisan-away [07:41] good morning === almaisan-away is now known as al-maisan [08:14] How do I get a local LP instance to accept a 3.0 (quilt) source package? [08:20] There's a table [08:21] NCommander: Have a look for InitializeDistroSeries, it copies stuff into a table that ends with or contains 'Format' [08:22] * NCommander just patched soyuz-sampledata-setupo to give me a precise [08:22] Hm, didn't I already fix that? [08:22] You go natty->oneiric and forgot precise [08:22] Apparently not [08:22] Well [08:22] I have onerous :) [08:22] From before the name was announced [08:22] So I haven't updated it in a while. [08:22] ('Onerous Ocelot', SeriesStatus.FUTURE, '11.10'), [08:27] * NCommander remembers how much 'fun' it is to setup LP from scratch [08:29] NCommander: Impossible. The brain suppresses bad memories, which is exactly what we are counting on. [08:31] * NCommander misses dak sometimes [08:31] s/sometimes/g [08:34] seems soyuz dev really wants a SMTP server setup [08:34] File "/usr/lib/python2.7/smtplib.py", line 284, in _get_socket [08:34] return socket.create_connection((port, host), timeout) [08:34] File "/usr/lib/python2.7/socket.py", line 571, in create_connection [08:34] raise err [08:34] error: [Errno 111] Connection refused [08:34] I get the same issue, ignore it [08:35] StevenK: then my GPG key doesn't import [08:35] Are you using utilities/make-lp-user? [08:36] documentation for soyuz sampledata says it should add my key to ppa-user [08:36] It imports my SSH key just fine [08:37] there [08:40] * NCommander does a test build [08:42] * NCommander discovers Launchpad rejects uploads to security if the archive is in DEVELOPMENT, and goes to modify the changelog to try again [08:46] * NCommander groans [08:46] deb http://archive.launchpad.dev/ubuntu precise main [08:46] What do I need to fix that? :-/ [08:47] nm found the magic UI page with the value I had to change [08:49] * NCommander has the basic framework in launchpad-buildd to build live images, and is just making sure he didn't break the world when I wasn't looking [08:58] NCommander: I do wonder how you're going to implement for-project [09:02] StevenK: flavor argument passed into the xmlrpc for launchpad-buildd [09:02] * NCommander already has that done and working [09:03] I meant inside LP [09:04] StevenK: new API in all likelyhood controlled by a celeberity [09:04] Ew, celebrity [09:04] StevenK: I wanted to use dummy packages, much easier :-P [09:06] NCommander: And then Colin will drive to your house and stab you. [09:07] But basic design idea I had with ubuntu-cdimage uses launchpadlib to tell LP to kick a build for a given architecture/flavor/series/etc, get the librarian URL for the image file when done, wget it, continue as normal [09:07] * NCommander *loves* we have three different livefs image builders [09:07] live-build/livecd.sh/ubuntu-defaults-image [09:08] here's what the arguments look like for a LiveImage build: http://paste.ubuntu.com/1148377/ [09:09] StevenK: if celeberities are bad, what should I key squashfs's builds off of? We don't want people willy-nilly being able to smack to respin API [09:11] * NCommander could connect it to the distro driver (~ubuntu-drives) [09:11] *drivers [09:11] Probably ~ubuntu-cdimage in the short term [09:12] I was personally thinking ubuntu-release [09:12] or a new team, since I though hardcoding ubuntu stuff in LP was a bad thing [09:13] though then again, ubuntu itself is a Celebrity ... [09:41] wgrant: busy? [10:06] lifeless: Hi [10:06] hi [10:10] lifeless: You pinged? [10:11] I did [10:11] cgoldbergs thing; I'm interviewing folk for 3 of the 4 hours I'm working tomorrow - half day [10:11] I'd really like it to get a paranoid once over. [10:11] Ah. I might look tomorrow, if you're busy [10:12] I'm running in place. [10:12] (yes, am busy, or wouldn't be asking) [10:14] k [10:23] wgrant: StevenK what dispatches things like TranslationTemplate builds? [10:24] builddmaster [10:24] * NCommander has a fairly decent grasp on how jobs go from LP->slaves now, but isn't quite sure what triggers the jobs (aka, how we geting */model/*buildbehavior.py) [10:24] NCommander: not sure why you talk about a celebrity above. [10:25] NCommander: any permissions you need should be distro centric, to support derived distros (e.g. linaro). [10:25] lifeless: building a livefs will require building on devirtualized builders, and not connected to any package [10:25] and? [10:26] lifeless: what team should it be connected to. There's isn't exactly a nice place where to hang it (first though was ~ubuntu-drivers but that team is already overloaded as is) [10:26] Unless I add a new Image Builder or Release Team thingy somewhere in a distroseries [10:26] any published distro driver would be my starting point; creating distros is a privileged operation. [10:27] don't overthink this. [10:27] lifeless: right, but the driver team currently has 60+ members in it and is being used for fun things like UDS managers [10:27] you don't need to limit it to 'just the people that have to do it', what you need to do is to 'exclude folk that might abuse or attack it' [10:27] these are wildly different problems. [10:27] * NCommander is about halfway getting launchpad-buildd to understand livebuilds, and suspect I'll need help when we get to the plumbing side in LP [10:28] If a UDS manager spins an extraneous livecd build, who cares. Tell them off. Problem solved. [10:28] Good point [10:28] lifeless: thanks for your help [10:29] lifeless is just bored because he can't play CS:GO. [10:29] * NCommander personally recommends civ 5 :-) [10:29] I was playing that a few nights ago. Bloody hell, it's *enormous* [10:29] That game is a soul-sucker. [10:30] Works amazingly well under wine though. Was pretty suprised when the demo just went [10:33] StevenK: helping and being bored are not mutually exclusive ;) [10:35] * NCommander notes his end goal is to have a nice webpage that completely replaces nusaken with all image building tasks being done on LP [10:36] vs. the arcane voodoo required now [10:36] *nusakan [10:37] so flavors can click a button and presto; respun images [10:48] * NCommander grumble [10:48] *s [10:48] my builder went to an ABORT state and I can't figure out where the log file is [10:55] NCommander: There's no build log for an ABORTed builder [10:55] NCommander: Check /var/log/launchpad-buildd/default.log on the slasve [10:59] wgrant: thanks. So my build succeeded and uploaded but running the test cases for launchpad-buildd sent it straight into OMG broken mode [10:59] mgz: did your patch make it to prod? [11:02] mgz: ah, looks like it is in 15805 waiting for the next NDT [11:02] exceptions.ValueError: Slave is not BUILDING when told build is complete - *grumble* [11:08] NCommander: Ah, yeah, running the test suite on a machine with a live launchpad-buildd is a good way to die. [11:09] * NCommander doesn't really feel like setting up a second machine at the moment to run as a slave :-/ [12:05] NCommander: VMs :) [13:32] deryck: ping for standup [13:32] rick_h_, coming now, sorry === al-maisan is now known as almaisan-away === slank` is now known as slank === slank is now known as Guest68178 [14:17] anyone have a link handy for using the people.canonical space? [14:18] I think I just sent a test .png file to the middle of no where [14:21] rick_h_: jpds might know [14:22] thanks czajkowski [14:23] np [15:16] yay for stable wifi at the wife's shop finally. [15:21] yay [16:06] sinzui: what happens when you click sub someone else https://answers.launchpad.net/launchpad/+question/205885 [16:07] czajkowski: js failed to init on the page I think [16:09] is a 404 (js-less) alright [16:11] sinzui: odd it's happening on https://answers.launchpad.net/launchpad/+question/205854 also [16:11] czajkowski: because is a js error [16:11] nods [16:11] just randomly happened [16:11] js fails every where, not on one page [16:11] well, the page not existing isn't a js error... if there's no non-js mode, there shouldn't be a href pointing to lies [16:12] czajkowski: The subscribe someone else is shared with bugs and mp reviews [16:13] since js is broken, we can see that answers does not support that via a html view [16:13] The error in the page related to comments. maybe jcsackett can understand the error on questions [16:15] * sinzui adds a comment to a question to see if the js starts to work [16:16] sinzui: with a comment it should. [16:16] yep [16:16] yea, there's no .boardComment and the JS doesn't allow for that to not exist so it's dying [16:16] looks like a goof in the cl setup--it assumes (and our test provides) a boardcomment. [16:16] var container = Y.one('.boardComment').get('parentNode'); [16:16] sinzui: i can whomp up a fix for that today. [16:16] assume it must exist [16:16] czajkowski: subscribe js on answers is broken when there is not comments [16:16] I added a comment, reloaded the page and js worked [16:17] oh most odd [16:17] thanks [16:17] I will report the bug. The js just needs to return early if there is no work to do [16:17] sinzui, rick_h_: looks like the cl can just not exist if it doesn't find boardcomment, since there's nothing for it to do in that situation anyway. agree/disagree? [16:18] jcsackett: yea assuming there's no way to ajax add comments [16:18] jcsackett: this is true for answers, Mp and bugs start with comments I think [16:18] jcsackett: that would blow up if a CL wasn't there for it to get added to [16:19] rick_h_: there's no interaction like that to worry about; there is an ajax add comment form, but when you add a comment on the page you can't immediately hide it anyway. that may be a defect, but it's a separate one as it's long been the case. [16:19] jcsackett: This has been broken for weeks. I think subscribing before commenting is rare [16:19] sinzui: i concur. and re: bug's having comments off the bat, that's correct, but i don't believe they are shown as boardComment--the first "comment" is the description. [16:20] ah, correct [16:20] jcsackett: you can report the bug and optionally choose to fix it. "disclosure information-type regression" [16:21] sinzui: ok, i can create a fix for this pretty easily. [18:28] jcsackett: since there re no reviewers today, I pre-emptively ask you to review a branch. I am not asking you to review it now [18:43] sinzui: no problem. [20:11] sinzui: r=me. [20:19] thank you jcsackett [21:01] lifeless: you might be interested in: https://code.launchpad.net/~jameinel/launchpad/loggerhead-clear-cache/+merge/119793 [21:02] calling '.clear_cache()' on all the versioned file objects has a pretty big impact on the 'spider loggerhead' peak memory. [21:03] does this impact performance? was the cache getting used before ? [21:04] lifeless: every request gets a new bzr_branchd [21:04] so no [21:04] cool then [21:04] (we could try to re-use them, but that is much hairier than not :) [22:22] see you guys on the flip side. [23:35] so how do files leave librarian? [23:35] * NCommander is trying to figure out how old squashfs's will poof out of librarian [23:37] NCommander: They are garbage collected when they are not referenced. [23:37] StevenK: what defines a reference? (or is there a minimium age before they get GC'ed?) [23:37] Mostly I want squashfs's to stick around for a day or two then I don't care if they get deleted [23:38] NCommander: So, stuff in the the librarian is an LFA and a correspending LFC (LibraryFileAlias, and LibraryFileContent) [23:38] NCommander: Things like BugAttachment link to LFA. [23:38] Sounds like squashfs's need a LFC when they're uploaded that points to the most recent squashfs [23:39] If the attachment is deleted, the row in BugAttachment is dropped, and the LFA now has nothing pointing to it [23:39] And it will get GC'd after a stay of execution, which is a week or something [23:40] NCommander: Your custom upload will create the LFA and LFC. Don't worry about that. [23:40] eh, in that case, its mostly then I don't care. I just need to be able to find the latest squashfs in librarian, download it, and let it get executed whenever its time rolls around [23:40] StevenK: er? I though non-package uploads just go straight into librarian and bypass the queue. I can't have a custom upload without an assiocated source package, no? [23:41] (aka, how translations are handled currently) [23:42] NCommander: Just because they're directly ACCEPTED does not mean there isn't a PackageUpload for them. [23:43] so launchpadlib call to trigger a build -> live-build job runs on buildds -> live-image upload to Soyuz -> PackageUpload? [23:43] * NCommander was thinking the system worked trigger->buildd job->librarian [23:44] * NCommander does have launchpad-buildd at the point where it now can do a live-image build [23:44] I [23:44] Sigh [23:45] I'm trying to remember if the translations stuff has a PackageUploadCustom associated with it [23:45] raw transitations do, which then via some automagic process get queued up by buildd manager to be executed. THere seems to be *very* little code in soyuz about it [23:45] The processed translations, I'm not so sure [23:46] NCommander: The revelant code would be in buildmaster [23:46] there's also very little code there :-/ [23:46] * NCommander has heard of trying to find a needle in a needlestack before ... [23:46] Perhaps archiveuploader [23:47] * NCommander thinks he needs to run a translations job and watch to see where the files go [23:47] If they go into soyuz, its easier because then I can simply iterate on an archive for the proper custom upload type (per docstories) [23:48] NCommander: I think you need to write a LEP [23:49] There's a partial one for this usecase [23:49] https://dev.launchpad.net/Soyuz/Specs/BuilddGeneralisation [23:50] you want a new build type? [23:50] I implemented a new build-type. live-image [23:50] rick_h_: Just found some internet, if you're available for a call sometime. Otherwise I'll send you an email. [23:50] WHich partially works [23:51] I think its stuck in an infinite loop ATM though [23:51] it's not an easy task [23:51] bigjools: well, my build-type works, the trick is getting ways to trigger the builds via launchpadlib, and getting those files somewhere useful [23:52] * NCommander originally was just going to have a source package + custom upload [23:52] wgrant: http://pastebin.ubuntu.com/1149798/ [23:54] NCommander: Why are you trying to use a packageupload? [23:54] It's not a packageupload [23:54] And I'm not sure there's any reason for it to be. [23:54] wgrant: I didn't think it was, StevenK was the one who brought it up [23:59] :(