=== Jack87 is now known as Jack87|Away [04:03] helo [04:03] hello* [06:41] Does qemu-system-arm work *at all* on lucid? [06:51] Hm, same behaviour on sid [07:30] I give up, qemu-system-arm won't work for me on lucid, and the only non-lucid host I have, is a shitty little Atom netbook (where it works, but unbearably slowly) [07:30] I will just reflash the TF101 again with ubuntu so that I have something to actually do compiles on === Jack87|Away is now known as Jack87 === Jack87 is now known as Jack87|Away === doko_ is now known as doko === Quintasan_ is now known as Quintasan === brandini_ is now known as brandini === sniperjo__ is now known as sniperjo [17:36] Grr.. the Oneiric kernel froze again.. (pandaboard with 2 DVB-T tuners atached) [19:26] janimo: Hey! Can you confirm the bump is correct on linux-ac100? [19:26] New package: linux-ac100 (universe) [2.6.38-1.3 → 2.6.38-1000.1] [19:26] Daviey, yes [19:26] janimo: What is the reasoning? [19:26] I am about to upload the linux-meta-ac100 so it is in universe by tomorrow [19:27] Daviey, I started with a new packaing using git [19:27] and followed a tutorial and existing ubuntu/linaro numbering convention [19:27] probably not needed in this case but I played it safe [19:27] janimo: I don't follow? Where does it state jump to 1000? [19:27] being my first time packaging a kernel by lots of trial and error eve [19:28] Daviey, it doe snot, the linaro template I started with uses 1000.0 [19:28] and all linaro kernels use such large abi numbers afaik [19:28] omap 1200, landing team ones 1400 etc [19:29] seems odd, but ok [19:29] infinity: ^^ [19:30] Daviey, tbh it seemed odd to me too, but if I were to question or dig into everything that seems odd to me in kernel packaging it would take me a lot more to get something ready [19:31] not one of the tutorials were 100% correct or covered the various cases, jcrigby's one is the closest and best yet [19:32] Daviey, at leats this package is now maintained in git and more or less in sync with how the kernel team and linaro manage their kernels [19:32] That works for me. [19:32] janimo: heh, it's a black art that i don't understand either... The changelog generation blows me :) [19:32] I even plan to upload it to the kernel repos so it is not only on my disk [19:33] Daviey, I thought I understood everything about the changelog, when an innocent change in it (well version number addition _before_ the last one) tripped the ABI checker [19:33] Would it surprise you to know that most of the kernel team doesn't know how the packages are built either? ;) [19:33] A lot of that stuff is from BenC and I fiddling years ago. [19:33] at which point I deferred work to next morning being an unexpected error after I thought I had ironed out all issues and built the package a few times already [19:33] so I try to touch as little as possible [19:33] The fact that it's survived is a testament to opaque code that no one wants to touch. [19:33] heh [19:33] and learn more as I upload new revisions [19:34] infinity, that would not surprise me in the least actually [19:34] janimo: To fully understand it, you have to first understand Ben Collins. After rooming with him at many conferences, I don't recommend it. [19:35] and also it does not scale well [19:35] would need many many conferences or very large rooms [19:35] (That's not to say that I have anything against the guy, we're great friends, just warning that insanity may ensue) [19:36] infinity: I think you might want to room with lag. [19:36] I don't need to corrupt any more Canonical employees, past, future, or present. [19:37] For the good of the company, I should probably always room alone. :P [19:38] heh [19:38] infinity, how selfless, you are. [19:38] infinity: I was more worried he'd corrupt you. :) [19:38] Seems unlikely. [19:38] But I'm always willing to be surprised. [19:39] jcrigby: LINARO: always build debug packages [19:40] jcrigby: ^-- From your recent uploads. Don't we have that turned off for distro kernels (to avoid duplication of massive packages between -dbg and .ddeb)? [19:40] crap [19:41] that was a change for linaro ppa [19:41] I figured. [19:41] because we wanted the debug stuff [19:41] infinity, ok educate me [19:41] Because PPAs don't do ddebs, yes. Longstanding feature request for that in LP. :P [19:41] so my fix didn't fix because it still goes to ddeb [19:42] it just always does it [19:42] Oh, in which case, these are fine for the primary archive. ;) [19:42] right [19:42] And you still need a better fix for your PPA, I guess. [19:42] so two wrongs make a right [19:42] yes, I need to not do the rename to ddeb [19:42] in ppa [19:43] so the debug stuff will just stick around like a normal deb [19:43] You can cheat and use archive_purpose to conditionally rename. [19:43] I'm a bit shocked the default templates don't do that. [19:43] you can tell me what the current stuff does [19:44] it looks in / for some file [19:44] let me look [19:44] /CurrentlyBuilding [19:44] yes [19:44] and uses that to decide if it is a local developer build or an archive build [19:45] if that file exists it sets fullbuild or somesuch [19:45] Kay, but if you check the value of it, you can get more in-depth. [19:45] then later do_debug get set or not based on full_build [19:46] infinity, jcrigby a meta package should just need a changelog bump and debian/rules will DTRT on build? [19:46] janimo: Depends on the meta source. But that's how most work. [19:47] ok, this is ogra's ac100 meta. Should work as it has no magic besides the debian/ dir in it [19:47] jcrigby: You can see in build logs that Purpose is actually set to a value in that file. [19:47] ok so packages are ok except for crappy short changelog [19:47] jcrigby: (Check the sbuild call for '--purpose=' [19:48] jcrigby: For instance for primary archive, it's --purpose=PRIMARY [19:48] ahh [19:48] infinity, linux-meta-ac100 uploaded, fingers crossed, etc [19:48] jcrigby: For PPAs, it should be PPA, but double-check one of your build logs. :P [19:48] wow, I need to hang out with infinity for a week of intense training [19:48] ok, I'll do that [19:48] infinity, is this documented anywhere? [19:48] or just in package masters heads? [19:48] jcrigby: *mumbles something about the launchpad source code* [19:49] :) [19:49] jcrigby: I'd be willing to bet that about 5 people know how this works, and 2 don't work here anymore. [19:54] jcrigby / janimo : Kernels all accepted, after some very generous review. May $diety have mercy on your souls if they all break. :P [19:55] * infinity really goes to find food this time. [19:56] infinity, thanks [19:57] jcrigby: Even for subarches like this where we're not building official images, it would be swell if, next cycle, you keep slightly more in sync. Diffs that large a 2 weeks before RC are a bit scary. [19:58] s/a 2/2/ [19:59] I agree [20:00] I need to make the ubuntu upload part of my normal workflow even though we (linaro) consume kernels from ppa [20:04] Certainly until we hit kernel freeze, anyway. [20:04] Then you need to get picky and potentially fork. [20:29] infinity, are you back from lunch [20:29] ? [20:30] jcrigby: I keep forgetting to start lunch. [20:30] jcrigby: 'sup? [20:31] so did I understand that I can look inside /currentlywhatever to figure out if I am ppa or normal or ... [20:31] jcrigby: Yup. [20:32] jcrigby: "Purpose: PPA" or "Purpose: PRIMARY" for PPA versus primary archive. There are a few other possible values (PARTNER, etc), but those are probably the only two you care about. [20:32] jcrigby: And no /CurrentlyBuilding at all can generally be assumed to be "user machine, or non-LP buildd". [20:32] ok those two are all I need, thanks [20:32] right [20:32] thanks, I now need to go fix debug for ppa again [20:33] but conditionaly so i don't break primary [20:33] If and when we get around to merging the archive sbuild and the buildd sbuild, this might become more easily testable. [20:33] sure [20:34] And someone should probably hack up some CurrentlyBuilding support into pbuilder for similar "pretend I'm a buildd" testing reasons. [20:34] infinity: I started doing that for sbuild [20:34] The upshot of that is that it would de-facto document it all. [20:35] Daviey: You're not the only one. :P [20:35] infinity: CurrentlyBuilding support requires to know the component, which means it is network bound - or you have to delcare it at spawn, i guess? [20:35] Daviey: On the buildds, it's declared when you start the job. [20:36] Daviey: --component=foo --purpose=bar [20:36] yeah, but launchpad is lucky enough to know the compoent. [20:36] :) [20:36] * Daviey looks up what a compoent is [20:36] Daviey: That's more than enough to get people to be able to match the archive behaviour for reproducibility. [20:37] Daviey: I'd argue that if you don't know what component your source package is in, you probably shouldn't be uploading it. :P [20:37] infinity: Incidently, i came acorss a package which built fine in pbuilder and sbuild, but failed on the buildd's. Grabbing the LP chroot, and throwing it into pbuilder allowed me to debug it [20:38] infinity: Well, my usecase was automated builds.. so thar! :) [20:38] Daviey: Yeah, for doing automated rebuilds or something, you could still cheat with the apt Sources.gz and grep-dctrl or some such. [20:38] Daviey: It wouldn't be up-to-the-minute accurate, but close enough. [20:39] But you'd want that sort of logic in a wrapper, not in pbuilder/sbuild themselves. [20:39] pbuilder and sbuild should just trust user input. [20:39] Not be clever. [20:40] Oh aye.. that is what i was doing [20:41] infinity: Incidently, i'd like to steal LP's Architecture: $target handling.. Need to see if/how that is exposed via API. [20:42] Hrm? [20:43] Not sure I know which Arch handling you're referring to. [20:44] You mean the part where it selects arch_indep for i386, and passes '-A' to sbuild? [20:47] If so, that's exposed via the API. See is_nominated_arch_indep at https://api.launchpad.net/devel/ubuntu/oneiric/i386 [20:47] Daviey: Was that what you were driving at? ---^ [20:49] * infinity lunches for real this time. Honest. [20:59] infinity: Well that yes, but also processing of linux-any, and the others [20:59] Although, not really an sbuild issue itself. [21:00] for each package [21:03] Daviey: http://paste.ubuntu.com/463055/ <-- The patch that was applied to lp-buildd's sbuild to handle linux-any. [21:03] Though, I'd assume the distro sbuild does the right thing anyway. [21:05] ah, nice === lopi is now known as Lopi