shadeslayer | wgrant: could you also enable arm for https://launchpad.net/~netrunner-arm/+archive/ubuntu/netrunner-odroid-c1-ci-patches and https://launchpad.net/~netrunner-arm/+archive/ubuntu/netrunner-odroid-c1-ci and https://launchpad.net/~netrunner-arm/+archive/ubuntu/netrunner-odroid-c1 please | 00:49 |
---|---|---|
wgrant | shadeslayer: The last was already done. | 01:11 |
wgrant | The others are now too. | 01:12 |
teward | wgrant: is there a suggested limit for ARM builds now? | 01:12 |
teward | other than no more than 6 a week or w/e it is | 01:12 |
teward | (I have 5 nginx packages to upload to the staging PPA that will build ARM) | 01:13 |
wgrant | The suggested limit is that you be reasonable about things. | 01:13 |
wgrant | How long does it take under qemu-user-static? | 01:13 |
wgrant | Does it work? | 01:13 |
wgrant | The limits (well, guidelines really) are in place mostly because qemu-user-static can be very slow. | 01:14 |
teward | judging from prior it's about 45mins each for arm. | 01:14 |
teward | give or take an hour system to system | 01:14 |
teward | but not insane | 01:14 |
wgrant | As long as they work, that sounds fine. | 01:14 |
wgrant | (this'll all be much easier in a couple of months when we have actual ARM VMs) | 01:14 |
teward | indeed | 01:14 |
teward | ... i should probably not be on the computer, there's a really heavy evil storm outside... o.o | 01:15 |
teward | wgrant: i bet there's people that upload things that take eons to build in arm, and that's why the limits exist? | 01:15 |
teward | or people that would *want* to make those kinds of uploads | 01:15 |
wgrant | Right. | 01:16 |
wgrant | The rule is really "don't be stupid about it", but that's not very easy for people to objectively judge about themselves :) | 01:16 |
teward | :P | 01:16 |
wgrant | If something's valuable and not a hideous resource hog, ARM builds are probably fine. | 01:17 |
teward | given that i've made stupid mistakes with the main repos on accident before, and then very quickly rectified, i'm used to judging something insane | 01:17 |
wgrant | Heh | 01:17 |
teward | i failed with an nginx security debdiff that was sponsored a while ago, that was definitely fun xD | 01:17 |
teward | loooong while ago xD | 01:18 |
teward | it caused SEGVs everywhere, and it was Debian's fault... xD | 01:18 |
teward | anyways, thanks, wgrant, for answering :) | 01:18 |
wgrant | np | 01:21 |
sergio-br2 | is it possible to ask arm64 build for PPA? | 03:41 |
wgrant | sergio-br2: qemu-system-aarch64 doesn't work very reliably, so we don't recommend it at the moment. | 03:43 |
wgrant | er | 03:43 |
wgrant | qemu-user-aarch64, rather | 03:43 |
sergio-br2 | why canonical doesn't use real hardware for ARM ? | 03:44 |
sergio-br2 | isn't it easier? | 03:44 |
Peng | like, a rack of cheap tablets? :P | 03:45 |
sergio-br2 | raspberry pi 2 cluster, heh | 03:45 |
sergio-br2 | dunno | 03:45 |
wgrant | It's easier if you can get the hardware, sure. | 03:45 |
wgrant | Server-class, virt-capable ARM hardware has been very difficult to come by until this year. | 03:45 |
wgrant | We have some now, but it's not set up yet. | 03:45 |
sergio-br2 | hardkernel uses a cluster to compile the kernel | 03:46 |
wgrant | Sure | 03:46 |
sergio-br2 | to odroid | 03:46 |
wgrant | But they're not running code from random people as root. | 03:46 |
wgrant | We have lots of ARM hardware for Ubuntu builds, but they can't run VMs efficiently. | 03:46 |
wgrant | They're Cortex-A9s. | 03:46 |
wgrant | hardkernel are probably using virt-capable ODROID-XUs, but they're not exactly server-class :) | 03:47 |
sergio-br2 | hum, VM for what? If it's already arm hardware? | 03:47 |
wgrant | VMs for security. | 03:47 |
sergio-br2 | ah | 03:47 |
wgrant | We don't *particularly* want to allow anyone with an email address to run code as root on bare metal... | 03:47 |
sergio-br2 | so i386 and amd64 uses VM for each build? | 03:48 |
sergio-br2 | dunno how these stuff works | 03:48 |
wgrant | Yes. | 03:49 |
wgrant | Every build runs in a VM, and that VM is destroyed afterwards. | 03:49 |
sergio-br2 | oh | 03:51 |
sergio-br2 | qemu too for them? | 03:51 |
sergio-br2 | canonical uses VM too to every build in the repo? | 03:52 |
wgrant | All virtual PPA builds today run in a 64-bit x86 KVM OpenStack instance. | 03:52 |
wgrant | armhf and arm64 run qemu-user-static on top of those VMs. | 03:52 |
wgrant | Ubuntu's builds are currently on bare metal, but they'll be moved into VMs, just like PPAs, in the coming weeks. | 03:53 |
sergio-br2 | oh | 03:53 |
sergio-br2 | but just a few people has access, so what's the problem? | 03:53 |
sergio-br2 | does debian do it too? | 03:53 |
wgrant | What's what problem? | 03:54 |
sergio-br2 | *about security? | 03:54 |
wgrant | For official Ubuntu builds there are fewer security concerns. But there aren't none. | 03:54 |
sergio-br2 | hum | 03:54 |
wgrant | And there are no obvious benefits to not using VMs for them, so we might as well use VMs for them. | 03:54 |
wgrant | VMs have several benefits and minimal drawbacks. | 03:55 |
sergio-br2 | very interesting | 04:07 |
=== anthonyf is now known as Guest97499 | ||
leitao | is launchpad down? I am not able to create a bug.... | 14:05 |
leitao | I got ' (Error ID: OOPS-fad1e185aee71ff2d33ce6044b50c7b7) ' when creating a bug. | 14:05 |
ubot5 | https://oops.canonical.com/?oopsid=OOPS-fad1e185aee71ff2d33ce6044b50c7b7 | 14:05 |
dobey | no. at least, i can view merge proposals just fine | 14:16 |
jafo | How do I decrypt the message PGP message Launchpad sends to confirm my key? I am getting "gpg: CRC error". Is launchpad sending invalid PGP data? I saved off the message as text, and did "gpg -d 0.txt", also tried "gpg --armor -d 0.txt". | 16:34 |
jafo | http://paste.debian.net/241524/ | 16:35 |
jafo | Aside: It sure would be nice if my PPA page said that I didn't have a key in my account and uploads would say they succeeded but would silently fail. Or the PPA page would say "Hey, an upload was tried, but it failed." Because searching I found something saying it could take hours to generate a PGP key for the PPA, so I waited all day yesterday for it to "catch up". | 16:39 |
dobey | does your e-mail client not have pgp support? | 16:44 |
jafo | Correct. | 16:44 |
dobey | afaik, launchpad sends valid mime content in the e-mail | 16:45 |
jafo | Oh, it is quoted printable. I can convert it using Python. | 16:49 |
jafo | It looks like it is just PGP data I can save and decrypt, but it isn't. | 16:50 |
dobey | right | 16:50 |
jafo | Yay, at least now I get an e-mail about why it failed! | 16:52 |
dobey | yeah, lp can't really give you an e-mail if you don't sign the package with a key that's connected to your account. the uploads are anonymous ftp, and it has no way to connect the upload to a specific account in a reliable way | 16:53 |
dobey | it might be good if it sent an e-mail to the owner of the ppa in such cases though | 16:54 |
jafo | dobey: I'm telling dput that it is for "ppa:username/package". It doesn't include that information in the FTP upload? | 16:57 |
jafo | Yay, I have my package up there now! | 16:58 |
dobey | jafo: it does, but you could be trying to upload to someone else's ppa, or a team-owned ppa. | 16:59 |
jafo | Oh, it is building for precise. How do I tell it to build for Precise *AND* trusty? | 16:59 |
dobey | this is why i said it might be good to notify the owner of the ppa, rather than the uploader, in such question | 16:59 |
jafo | dobey: Gotcha. | 16:59 |
dobey | you have to upload a different source for each series | 16:59 |
dobey | or copy to other series | 16:59 |
jafo | dobey: How do you copy to another series. I tried changing the changelog and uploading again, but it rejected that because of a file with an existing name but different contents. | 17:04 |
dobey | you can't re-upload the orig.tar.gz with different contents | 17:04 |
dobey | there's UI on the web page for the ppa to do a copy | 17:04 |
jafo | I guess I just don't understand it. I select "Copy packages" select the source, select a destination of "This PPA" and "Series: Trusty", and it fails: "Copied from libRETS. Target series: Trusty. librets 1.6.1-1realgo in precise (same version already building in the destination archive for Precise)." | 17:08 |
dobey | were you trying to copy the binaries, or did you select to rebuild? | 17:11 |
jafo | dobey: I don't have binaries yet. | 17:19 |
jafo | Yeah, I selected "rebuild the copied sources". | 17:20 |
dobey | ok, i'm not sure | 17:22 |
jafo | http://box.jafo.ca/launchpad.jpg | 17:23 |
jafo | That's the page I submitted. | 17:23 |
dobey | ok. i'm not sure what's wrong | 17:25 |
jafo | I sure do miss having a good friend on the launchpad team... :-/ :-) | 17:28 |
jafo | Maybe it will work when the package builds... I've also tried pushing up a "1" suffix package name with trusty specified, and that was accepted. Not ideal, but workable. | 17:29 |
dobey | is the code hosted in a bzr branch in launchpad? or mirrorable into a bzr branch on launchpad? | 17:30 |
jafo | The code is not hosted in a bzr branch. I guess it is possible to mirror it to launchpad, it is hosted on github. I don't control the upstream, I'm just packaging it. | 17:32 |
dobey | when i package things in PPAs, i've started just maintaining a branch which only contains the contents of the debian/ dir, and using source imports and source package recipes, so that it's mostly all automated | 17:36 |
dobey | pretty nice, and makes having different builds for different series mostly trivial | 17:37 |
=== dpm is now known as dpm-afk | ||
jafo | I've got a github repo that has the debian dir. Might submit it upstream once it stabilizes. | 17:43 |
jafo | Sounds like what you are doing might be a good way to go for maintaining the PPA. | 17:44 |
jafo | Is there something that documents making that happen? | 17:44 |
dobey | there are docs on the launchpad help page for source package recipes, yeah | 18:12 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!