=== masACC is now known as maswan [00:25] whats involved in getting backport bugs addressed [00:31] I don't like it when apt-get dist-upgrade says this: [00:31] Fetched 7,012 kB in 25s (275 kB/s) [00:31] Segmentation fault [00:33] is that 16.04? [00:36] karstensrage: Ping someone in https://launchpad.net/~ubuntu-backporters [00:42] anyone? there are only 10? [00:43] the pending members have been pending for quite some time O.O [00:45] broder, maybe you might be interested? [00:59] whats the criteria for backporting new [00:59] stuff === juliank_ is now known as juliank === zequence_ is now known as zequence [14:19] Is there any additonal debug output I can turn on for a video driver? [14:20] It is crashing with cryptic messages like NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context [14:20] and a lot of NVRM: Xid (PCI:0000:07:00): 3, C 00000002 SC 00000003 M 00000104 Data 00000000 [14:21] It is:07:00.0 VGA compatible controller: NVIDIA Corporation G98 [GeForce 8400 GS Rev. 2] (rev a1) [17:14] zigo: LP 1562358 finally filed, along with LP 1562356. Just FYI. [17:15] Launchpad bug 1562358 in python-googleapi (Ubuntu) "python-googleapi is incompatible with oauth2client >= 2.x" [Undecided,New] https://launchpad.net/bugs/1562358 [17:15] Launchpad bug 1562356 in gcalcli (Ubuntu) "gcalcli incompatible with oauth2client >= 2.x" [Undecided,New] https://launchpad.net/bugs/1562356 [19:19] lamont: If the crash was in xenial, apt 1.2.8 fixes a known segfault. It just needs to be synced. It would still make sense to have a full backtrace, if possible. [19:20] Oh wait, no, 1.2.8 fixes an update segfault. [19:20] So a full bt would be *really* useful [19:22] If it's APT that's crashing, otherwise I don't care :) [21:00] Skuggen: usr/include/mysql/mysql/client_plugin.h:103:38: fatal error: mysql/plugin_auth_common.h: No such file or directory [21:00] compilation terminated. [21:00] Skuggen: MySQL's fault, or mythtv's? [21:13] superm1: mythtv seems to hate you. [21:15] infinity: mysql.h includes mysql/client_plugin.h in 5.7, and didn't in 5.6 [21:15] It needs include/mysql on the include path to work [21:16] Skuggen: Is that not provided by pkg-config or similar (or is mythtv not using it?) [21:16] The header structure is a bit of a mess. Devs are working on cleaning it up, but wasn't ready for the 5.7 release :| [21:16] Yeah, I haven't seen that issue outside mythtv so far [21:16] So it's part that the directory structure is weird and part that mythtv does something a bit differently, I think [21:17] Yeah. Looking now. [21:18] I explicitly patched to fix that [21:18] * superm1 shrugs [21:18] superm1: You may have patched the wrong bit. The g++ command sure doesn't have the include. [21:19] * infinity grabs the source. [21:23] I patched the header check and the .pro to add the includes. It did all work correctly in my local test build [21:23] I'm afk atm I'll look when I get to computer [21:30] juliank: gdb apt-get; r dist-upgrade.... and it's working :( [21:30] oh, right .. let's do this on the correct machine [21:30] Hah. [21:31] apt-get dist-upgrade ==> segv. apt-get install gdb ==>runs to completion fine. [21:32] bug inbound against apt with the trqace [21:32] You're lucky the latter didn't fix the former. [21:32] I KNOW [21:32] * lamont was quite concerned actually [21:33] I'd have unpacked a dpkg deb without dpkg or apt's databases being involved, probably. :P [21:33] I see that 1.2.8 is there, I'll try that after I file the bug. [21:33] s/dpkg deb/gdb deb/ [21:33] heh [21:33] you know it's a red letter day, when you get to type Yes, do as I say!, twice. [21:34] I don't think I've had to type that in years. [21:34] I get the prompt all the time, but that's my own stupid fault, and it's saving me from myself. [21:35] oh yes, that prompt is the machine going, no seriously, think about this one before you say yes, multiple times. [21:35] if you ever determine that you should really type it, it's time to file a bug. [21:35] It's triggered by removing an Essential package, IIRC. Not a whole lot of logic behind it. [21:35] And yes, absolutely a distro bug if it happens without you explicitly removing something you shouldn't. [21:37] yep. removing essential is the whole and sole cause of that [21:39] * infinity upgrades his IBM PowerStation from 14.04 to 16.04 for the lolz. [21:40] * lamont attaches the trace from 1.2.8 [21:40] juliank: over to you sir, lp #1562402 [21:40] Launchpad bug 1562402 in apt (Ubuntu) "segv in apt 1.2.7 with dist-upgrade" [Undecided,New] https://launchpad.net/bugs/1562402 [21:41] Man, this disk (and the SAS controller it's attached to) were wildly overspecced for this machine. [21:41] Oh no [21:42] Two different backtraces :/ [21:42] I do believe this is the only 15k RPM disk in my house. [21:42] infinity: is it time for me to start shifting the home infra to xenial? [21:42] But I guess this is what happens when you ask IBM to make an "affordable workstation". [21:42] lamont: Can you generate a bt with dbgsyms? [21:42] maybe [21:42] lamont: That's what I'm doing right now. I think we still have one or two upgrade bugs you might want to wait on, though. [21:43] * lamont hasn't bothered with dbgsyms in ages... process? [21:43] lamont: Definitely have one that fubars upgrades in GUIs. [21:43] lamont: I'm seeing now how serverish upgrades go. [21:43] infinity: I see [21:43] * lamont will let infinity embrace the suck on his behalf [21:44] Right now, all I can report is that server upgrades are fast. If you have a disk that can actually keep up with your CPU. [21:44] Or, in this case, a disk that might be faster than my CPU. :P [21:44] lamont: and/or run it in valgrind [21:45] (preferably and) [21:45] lamont: dbgsyms == install libc6-dbg and install the matching lib and apt ddebs from http://ddebs.ubuntu.com/pool/main/a/apt/ [21:45] sigh. installed valgrind (and the most recent maas, because of huh), and then ran apt-get dist-upgrade... and it's running [21:46] lamont: Yes, you can add ddebs.u.c to sources.list and 'apt-get install apt-dbgsym' too, if apt is working. [21:46] Oh. Or valgrind broke you. \o/ [21:46] :( [21:46] yeah it could have been that the maas ppa install tickled it into happiness, too [21:46] Yay for perturbing databases until it works. :( [21:46] but valgrind and I are about to have some fun [21:46] infinity: no yay [21:46] * lamont loves hard failurs [21:46] Saving /var/lib/dpkg, /var/cache/apt, and /var/lib/apt might have been nice. [21:47] If you have a time machine. [21:47] DO YOU HAVE A TIME MACHINE?! [21:47] "dammit, I wish I'd created that lvmsnapshot that I didn;t think to do this time" [21:47] ah, but I do have a backup of yesterday that includes /var/lib/dpkg [21:47] Indeed you would. [21:48] If you know what you've installed since, it's pretty safe to restore the old dpkg bits. Well, safe-ish. [21:48] And just reinstall overtop again. [21:48] After debugging. :P [21:48] sigh [21:48] backup excludes for /var: lib/dpkg/ [21:49] lamont: /var/backups [21:49] I'll rewind the maas bits, maybe [21:49] oh! [21:50] lamont: Do you perhaps have /var/backups/dpkg.status.0{,1.gz,2.gz} [21:50] It's not the entirety of /var/lib/dpkg, IIRC, but it's the databases apt cares about. [21:50] * juliank has that on Debian [21:50] but not sure why [21:50] yeah [21:50] do I care about more than dpkg.status? [21:50] Possibly apt.extended_states.0 too, depending on where the crash was. [21:51] That should be it. You can just pass -o Dir::State::Status=/var/backups/dpkg.status.0 to APT I'd think [21:52] Segmentation fault [21:52] \o/ [21:52] Yay! Let's go valgrinding that bastard. [21:53] * lamont installs dbg syms [21:53] W: Failed to fetch http://ddebs.ubuntu.com/ubuntu/dists/xenial-proposed/main/binary-amd64/Packages 404 Not Found [21:53] W: Failed to fetch http://ddebs.ubuntu.com/ubuntu/dists/xenial-proposed/main/binary-amd64/Packages 404 Not Found [21:53] infinity: sigh.. what does that ddebs line look like? [21:54] lamont: No ubuntu. [21:54] ta [21:54] Just ddebs.u.c/ $series $comp $comp [21:54] We really should put an ubuntu symlink in there to / [21:54] infinity: oh, thanks for the 1.2.8 sync. [21:54] Maybe I'll do that right now. [21:55] juliank: I caught you talking about it in backscroll, changelog looked sane, so you're welcome. [21:55] http://paste.ubuntu.com/15514548/ <-- with dbgsyms [21:55] infinity: yes we should! it's been bugging me SO many times [21:55] stgraber: Done. [21:56] * infinity tests. [21:56] I don't think it's mentioned in the changelog, but I made the flaky tests retry, so I hope it passes CI on all architectures now, even the always-failed armhf [21:56] http://paste.ubuntu.com/15514567/ <-- juliank [21:56] looking forward to 1.2.9 :D [21:56] yay [21:56] Looks like that did the trick. [21:56] lamont: Works with ubuntu now too. :P [21:57] * lamont hugs infinity [21:57] Err. [21:57] hmm, seems like those dbgsym do not match your installed version :( [21:57] Except for apt yelling at me about it. [21:57] oh, yeha [21:57] Oh, but that's true regardless of the path. [21:57] It just needs fixing still. [21:58] Anyway, that's fairly weird stuff [21:58] Or I need to ship the key in the distro. Derp. [21:58] ddebs could likely be useful to add commented out in the default sources.list these days. [21:58] juliank: that's 1.2.7 syms with 1.2.8 apt. :( [21:58] * juliank thought so [21:58] So get 1.2.8 syms? [21:58] seems the new ones are not available yet [21:59] They're available in LP. [21:59] https://launchpad.net/ubuntu/+source/apt/1.2.8/+build/9403755 [21:59] http://paste.ubuntu.com/15514599/ <-- full valgrind [21:59] infinity: reinstalled 1.2.7 [21:59] That works too. But I suspect he wants both if they're different. [21:59] http://paste.ubuntu.com/15514613/ [22:00] juliank: ^^ 1.2.7 bt [22:00] lamont: reinstall libapt-pkg5.0=1.2.7 [22:00] lamont: Also, dude, learn to autoremove. [22:00] meh [22:00] You only pulled in the apt binaries, not the library [22:00] infinity: I do.. this is me not perturbing the db [22:00] infinity: Now both are the same... [22:00] Oh well, sorry, library is not downgraded yet. [22:01] ii libapt-pkg5.0:amd64 1.2.7 amd64 package management runtime library [22:01] weird [22:01] dpkg -l | grep 1.2.8 yields only zlib1g [22:02] infinity: lp #1543683 has me being generous in my keeping of kernels [22:02] Launchpad bug 1543683 in linux (Ubuntu Xenial) "Fails to detect (second) display" [Medium,Fix released] https://launchpad.net/bugs/1543683 [22:03] To be fair, I have no idea what's going on. It seems that we are iterating to a broken dependency [22:03] (in both cases) [22:03] and sometime between now and mondya, I need to boot a kernel so that I can definitively say that the "fix released" there is an outright lie [22:04] juliank: there is one small lie in the status file, that I don't think should matter... Package: perl5 Depends: perl, because I haven't rebuilt a binary yet to depend on perl instead of perl5 [22:04] mentioned in the interests of full disclosure [22:05] * lamont tries one other thing [22:08] interesting... if I drop the maas ppas, then all is happy and right in the world again [22:08] * lamont bisects [22:08] * infinity still resents that he needs to do this after do-release-upgrade: [22:08] dpkg -l | awk '/^rc/ {print $2}' | xargs dpkg -P [22:08] amusingly, /var/backups is in my backup [22:09] lamont: Could the maas PPA have a broken control file somewhere that dpkg somehow missed? [22:09] I have been doing rm /var/lib/apt/lists/*; apt-get update each pass [22:09] infinity: Would you rather do apt purge ?config-files [22:10] juliank: I didn't know that was a thing. [22:10] File descriptor 3 (pipe:[5854305]) leaked on vgs invocation. Parent PID 3149: /usr/sbin/grub-probe [22:10] infinity: It isn't [22:10] stupid grub-probe [22:10] juliank: But it's not the shell pipe that I resent, it's that I can't upgrade-with-purge via the release-upgrader. ie: it's our bug. :P [22:10] infinity: It's an aptitude thing right now... [22:10] But I definitely want patterns in APT soon-ish [22:12] lamont: As to the parallel topic, a serverish do-release-upgrade -d seemed to DTRT. Packages all installed, postinsts all succeeded, the only unexpected removal was aptitude (which I almost consider a feature, but I suppose I should look into). [22:12] heh [22:12] Rebooting now to be sure it actually worked... [22:12] yeah, they'll squawk about that [22:13] And it boots! [22:13] All systemdish and kernel 4.4ish. [22:14] lamont: YOu might want to run with Debug::pkgPackageManager=yes and see which package causes that (it should be the last printed in a line starting with SmartUnPack) [22:14] then you can see if there's anything odd about that [22:16] We need to retry the APT test on armhf [22:16] \o/ [22:17] dist-upgrade with everything but http://ppa.launchpad.net/maas-maintainers/experimental3/ubuntu enabled, ran fine. uncommented that one, and *splat* [22:17] Erm. Reboot worked okay except that my / is read-only now. WTF, systemd. [22:17] SmartUnPack maas-region-controller-min:amd64 (replace version 2.0.0~alpha3+bzr4810-0ubuntu1 with Segmentation fault [22:17] how appropriate [22:18] Replace it with a SEGV indeed. [22:18] juliank: Retried. [22:18] infinity: thx [22:18] http://paste.ubuntu.com/15514818/ [22:19] lamont: What's the apt-cache policy output for that? [22:19] http://paste.ubuntu.com/15514824/ <-- almost pasted before you asked [22:21] lamont: apt-cache showpkg would be good to have too [22:22] http://paste.ubuntu.com/15514844/ <-- showpkg juliank [22:22] That's show, not showpkg. [22:22] -kohlrabi(root) 381 : apt-cache showpkg maas-region-controller-min > zz [22:22] might be, but iz showpkg. [22:22] infinity: BTW: The failed test tries to detect that progress reporting works (start at 0, pulses somewhere in the middle, and stops at 100). It tests a 800k file download, at speed 1600/i where i is in [1..10] [22:23] So it actually loads the 800 KB file at 160 KB/s [22:23] lamont: showpkg looks more like http://paste.ubuntu.com/15514856/ [22:24] oh haha [22:25] http://paste.ubuntu.com/15514869/ <-- actually different and correct this time [22:26] But if it crashes after the "replaces ... with" part, it's actually worse than I though [22:27] Well, it makes sense, though [22:28] maybe it's a regression from bug #1550741 [22:28] bug 1550741 in kmod (Ubuntu) "Upgrade failed - unauthenticated package (module-init-tools)" [Critical,Fix released] https://launchpad.net/bugs/1550741 [22:28] hmm, typo [22:30] That is: The problem now appears to be that the cache has an invalid version selected for installation. [22:30] Which might be a regression from http://anonscm.debian.org/cgit/apt/apt.git/commit/?id=0390edd5452b081f8efcf412f96d535a1d959457 [22:31] \o/ [22:31] juliank: and reproduced on another machine! [22:31] juliank: point me at an ssh key for you? [22:32] Second or third one of https://launchpad.net/~juliank/+sshkeys [22:35] http://paste.ubuntu.com/15514948/ [22:36] | fa1b75a1-152b-4479-a80e-9b8edaea6caa | ubuntu-released/ubuntu-xenial-16.04-beta1-amd64-server-20160223.1-disk1.img | ACTIVE | | [22:38] hmm, where do I get that nova tool? [22:38] juliank: see also /query [22:38] that [22:38] s python-novaclient [22:43] Oh, you can't be serious. [22:44] lamont: Okay, one glaring effin' regression from upstart/mountall->systemd that I can't believe I'm the first to run into. [22:44] oh? [22:44] do I even want to know? [22:44] lamont: It fails to remount / rw if there's no (OTHERWISE COMPLETELY POINTLESS) entry for / in fstab. [22:44] lol [22:44] how's that readnoly thing going? [22:45] Well, fine now that I added an fstab entry. :P [22:45] Stupid bug is stupid. [22:46] And completely counter to the "empty /etc" goal, so I have no idea WTF. [22:46] infinity: you know what's hilarious? "Myspace for dummies", published 2008 [22:47] infinity, do you pull from unstable or just testing? for xenial? [22:47] juliank: amuslingly? apt-get install maas fix0rs everything [22:47] karstensrage: Unstable. [22:47] sweet [22:47] its up there [22:47] karstensrage: Does "it" have a name? [22:48] note also that the automatic pulling from unstable stopped for xenial a while ago [22:48] one sec [22:48] lamont: Yeah, he's trying to get something NEW in. [22:48] lamont: Which would be easier with a package name. [22:48] *hint, hint* [22:48] heh [22:48] infinity: what, can't you just do a for loop checking for new packages? :p [22:49] * lamont has lost faith [22:49] lamont: We effectively do just that when autosyncing, but not past DIF. [22:49] https://packages.debian.org/unstable/libpam-ufpidentity [22:50] karstensrage: FWIW, source packages are the real interest, so https://packages.qa.debian.org/libpam-ufpidentity would be the more relevant URL for such requests (happens that binary and source match this time, though). [22:50] karstensrage: And synced. [22:51] thank you so much [22:52] karstensrage: And https://launchpad.net/ubuntu/+source/libpam-ufpidentity to follow along at home. [22:52] * lamont wanders off for a bit [22:58] "Bash script for wrapping cron jobs to prevent excess email sendin" ... Because that extra 'g' would have been too much. [22:59] Hah. :D [22:59] Thanks for reviewing cronic! [22:59] Unit193: "Review" is a strong word for what I do with direct Debian syncs. [23:00] Still, thanks. [23:00] But now all I can think of is some redneck worryin' about too much email sendin' on his dialup. [23:01] Ahaha. :D [23:01] * Unit193 is from Ohio, but city parts! [23:03] lamont: infinity: Issue caused by the commit I linked. [23:03] So: Yay, it's mvo's fault! [23:04] juliank: Man, that mvo guy sure is a jerk. [23:04] Oh, wait, no. The other thing. [23:04] ohio has city parts? [23:04] lamont: Cleveland? [23:05] I guess so [23:05] lamont: Population of 400k. Doesn't qualify as a "city" to me, but it does to most people. :P [23:05] infinity: for you, it's not a city until it's big enough to have an underbelly? [23:05] Welp, I don't know what I live in then, it certainly isn't a 'city' though if CLE isn't. >_> [23:06] On, and Columbus is 800k. [23:06] That's approaching a real city. [23:07] Unit193: tbf, the post office serving my house is for a town that might be 4500 now, and the next closest town has about a 150k population [23:08] maas-region-controller-min:amd64 Depends on maas-common [ amd64 ] < 2.0.0~alpha3+bzr4810-0ubuntu1 -> 2.0.0~alpha4+bzr4843-0ubuntu1~xenial2 > ( net ) (= 2.0.0~alpha4+bzr4837-0ubuntu1~xenial1) can't be satisfied! [23:08] lamont: But you no longer get your Internet via Pringles can, so you're moving up in the world. [23:08] lamont: ...The metro area has 124,475. >_> [23:09] infinity: but I hate paying for my T1 [23:10] which, tbf, is about 1/7 the bandwidth of the radio that's still on the tower, just not used by me [23:10] lamont: I thought you finally got cheap residential cable out there? [23:10] hahaha [23:10] cable comes to about 2 miles away [23:10] Oh. :P [23:10] 2 miles is pretty close! [23:10] the DSLAM is 17000 feet -- tariff is 15000... and "T1" is a total misnomer for a pair of bonded hdsl lines [23:11] and that is why I have an office in town [23:11] Oh, it's not actual ISDN/T1 old skool tech? Just DSL sold at T1 speeds? [23:11] I could have DSL if the firestation were 1/2 mile closer [23:12] not entirely sure... "4-wire T1" [23:12] but I'm given to understand that it's hdsl pretending to be a T1... 2-pair to the house, and then magically 1-pair to get to the terminus in the computer room [23:14] http://goo.gl/BTcjCE <-- infinity [23:14] that's the box in the comp room [23:15] the next step down was 128kbps ISDN for about 2/3 the price of the T1 [23:15] amusingly, it's not uncommon for me to see 180KB/s from the T1 [23:15] That sounds like real old skool analog tdm T1. [23:16] yeah - it's more the 2-pair from the ped to the demark that leads me to believe it's playing games [23:16] But maybe their magic ... Yeah. [23:16] Almost fixed [23:16] lamont: It must be weird living in the past. [23:16] totally [23:17] lamont: Did you have to brush up on WTF Frame Relay is? :P [23:17] also, screw you. :p [23:17] I don't think they're actually using frame relay, either. [23:18] Well, Frame or ATM would seem like the protocols of choice. [23:18] Perhaps with a PPPo in front of it if they hate their customers. [23:18] the funniest part? they deliver me 3 voice circuits from that adtran, which then go into a telephony card [23:18] BECAUSE THE ISP DOESN'T DO VOIP [23:19] oh, the actual service for me: their adtran is configured on a 1918 network talking to my router and passing me a /29 (and sometimes, even their tier 2 is most confused how that even works) [23:20] so from my perspective, it's just "ip route default via 192.168...." [23:20] and, of course, they advertise my /24 from the swamp for me, at no additional charge [23:22] anyrate, afk [23:22] juliank: Are you going to tag a 1.2.9 with this fix "soonish", or should I just yoink the commit for an ubuntu1? [23:23] infinity: I'll do some more testing and upload this tomorrow I think [23:23] juliank: WFM. [23:23] juliank: I'm in no rush, lamont got unstuck, and I'm guessing it's a bit of a corner case. Just want it in xenial sometime soonish. [23:26] https://github.com/julian-klode/apt/commit/71f2ab088f8e077ba8375e87ee9bb595655f954a [23:26] That should fix it [23:28] I should also come up with a test case, though [23:31] infinity: I'm not sure if removing maas-region-controller-min and installing maas-region-controller is the right choice, though, but it's what APT < 1.2.7 did and does not crash [23:32] I see, maas-region-controller-min is not up to date enough [23:32] (4837 and the rest 4843) [23:33] Now I need to build a test case where one package needs to be removed to allow another package to be installed. [23:33] Let's see if I can do that [23:46] juliank: maas-region-controller-min got renamed to maas-region-api [23:48] juliank: ^^ [23:48] Yeah, saw that. [23:48] so before was maas-region-controller{,-min} and now is maas-region-{api,controller} [23:49] It unfortunately works in a minimized example test case [23:49] of course it does. :( [23:49] SCIENCE! [23:56] ● nosferatu [23:56] State: running [23:56] Jobs: 0 queued [23:56] Failed: 0 units [23:56] ● leviathan [23:56] State: running [23:56] Jobs: 1783695312 queued [23:56] Failed: 1783695312 units [23:56] I WONDER IF THAT MIGHT BE AN ENDIAN BUG. [23:57] Ooo, even better on s390x. [23:57] ● z13-028 [23:57] State: running [23:57] Jobs: 3804834690 queued [23:57] Failed: 3804834690 units [23:57] SO MANY JOBS! [23:58] xnox: ^-- Some lolz for you. nosferatu = amd64, leviathan = powerpc, z13-028 = s390x. [23:59] systemd, you never cease to amaze.