[00:31] <adam_g> hey, im trying to bisect for a bug that was seemingly introduced with v3.3. according to tags in the ubuntu-precise repo, the testing window seems to be: Ubuntu-lts-3.4.0-3.7 and Ubuntu-3.2.0-9.16, but 'bisect start' says 'Bisecting: a merge base must be tested' and starts me at Linux 3.2.  i'm assuming i've done something wrong, are these the points to start, or have i chosen incorrect tags?
[00:50] <bjf[afk]> adam_g: take a look at https://wiki.ubuntu.com/Kernel/KernelBisection
[00:53] <adam_g> bjf[afk]: i've followed that so far
[00:54] <adam_g> AFAICS, it doesn't realy cover my issue
[01:05] <scientes> where is the gitweb for this? http://kernel.ubuntu.com/git-repos/ubuntu/ubuntu-precise.git
[01:12] <jk-> scientes: http://kernel.ubuntu.com/git/ubuntu/ubuntu-precise.git/
[01:14] <scientes> I just got a not permitted (EPERM right) from prctl(PR_SET_CHILD_SUBREAPER) instead of the expected EINVAL
[01:14] <scientes> (36)
[06:46] <ppisati> moin
[07:00] <cooloney> ppisati: morning, man
[07:24] <smb> morning .*
[07:25]  * RAOF matches that regex!
[07:25] <RAOF> Good morning smb :)
[07:26] <cooloney> smb: morning
[07:26] <smb> :) 
[07:27] <smb> RAOF, Maybe you will tell me someone will frantically work on my acceleration failure with my radeon driven Dell... ;-P
[07:27] <RAOF> smb: Nobody magically picked that up while I was away? Bad team! No biscuit!
[07:28] <smb> Nope, all hiding to wait for someone else stepping up. :) 
[07:28] <lifeless> RAOF: wb, and congrats.
[07:28] <RAOF> lifeless: Thanks!
[07:28] <RAOF> Hah. I see that launchpad now sends oops emails when failing to copy packages :)
[07:29] <lifeless> cjwatsons work
[07:29] <smb> RAOF, And yeah well, hm, congrats... I guess your effort was the easier... But anyways :)
[07:29] <RAOF> Yup. Not nearly as big as Sam's effort!
[07:31] <RAOF> lifeless: So, the oops email doesn't exactly make it clear - do I try again? Will LP finish the copy later, after it's its hissy fit has ended?
[07:49] <lifeless> RAOF: ask cjwatson :P
[09:48] <apw> cooloney, so i guess we need to see kdepim done for panda es there as thats the only one which built
[09:49] <cooloney> apw: yeah, it will finish soon. from the launchpad log, it should be about 4:30 hours
[09:49] <apw> cooloney, does your machine have swap configures, the one which fails with out of memory
[09:50] <cooloney> apw: oh, no. i did setup that for both imx6 and pandaES
[09:51] <cooloney> apw: i will try that later
[09:52] <apw> cooloney, are you saying you didn't have swap on either?
[09:53] <cooloney> apw: yeah, there is no swap on my imx6 and pandaES right now
[09:54] <cooloney> apw: i'm not sure about our PandaES buildd, does it have swap?
[09:54] <apw> cooloney, i would imagine they would have to have swap to build anything of any size
[09:55] <cooloney> apw: ok, after finishing building kdepim, i will add a swap and try to build gcc-4.6 and webkit then
[09:56] <smb> cooloney, How much memory have these things? I mean main memory?
[09:57] <apw> infinity, i assume the arm builds have swap and i assume it is on the spinning media
[09:59] <cooloney> smb: for both of them it's 1G memory
[09:59] <apw> i do not see firefox building in 1G of ram, without swap
[10:00] <smb> Ok, not as bad as older Arms but still not that much compared to other things
[10:00] <smb> apw, One would be surprised
[10:00] <smb> That actually could be the reason for the failures, couldn't it?
[10:02] <cooloney> for firefox, i think it should be some package dependency issue,
[10:02] <cooloney> i failed to build firefox on my x86 host with sbuild precise armel schroot.
[10:03] <cooloney> the error is the same. 
[10:03] <cooloney> but for webkit building error, i guess it might be related to memory or sway
[10:03] <cooloney> swap
[10:04] <cooloney> apw and smb, if we just compare the sucessful case 'kdepim', imx6 is about 3 and half hours, pandaES should be 4 and half
[10:06] <apw> thats not a huge improvement really is it compared to what was claimed
[10:07] <smb> One hour seems not too bad, though
[10:13] <cooloney> apw: probably need more testing. i just have one board, maybe rtg can help to test on his side.
[11:43] <cooloney> apw and smb, on my pandaES. it took 5hours to build kdepim
[11:43] <apw> cooloney, ok, we probabally need to get swap on, and then do like 3 of each
[11:44] <apw> so we can see how variable it is
[11:44] <cooloney> apw: ok, i will try that
[11:48] <cooloney> apw: and do you know how to resize a partition with some command line tool in ubuntu. i want to add a new swap partition and shrink my root partition 
[11:49] <apw> cooloney, hmmm i don't think i do, the installer does it, but i am not sure how
[11:49] <apw> though google knows all normally
[11:49] <cooloney> apw: installer is using gparted
[11:50] <cooloney> ok, maybe it should be resize2fs
[11:50] <smb> I think gparted would use that under the hoods
[11:50] <smb> But only offline (whole disk not mounted)
[11:53] <cooloney> smb: yeah, i can boot from SD card and mount the hard drive for resize2fs.
[11:54] <smb> cooloney, Though resize2fs is not enough, you want the paritioning to be adjusted as well
[11:54] <cooloney> smb: right, i plan to use fdisk after that to setup the swap 
[11:55] <smb> cooloney, Just saying, why not take the whole disk to some machine that runs gparted and do it from there.
[11:57] <cooloney> smb: aha, my bad. that's a better method. thx
[12:01] <smb> I try to use the simplest approach :) Otherwise it is too easy to mess up
[12:31] <mwk_> the ubuntu provided ec2 images include kernels that are not compiled with CONFIG_HW_RANDOM_VIRTIO.  does ec2 support the virtio hwrng?
[12:37] <smb> mwk_, I have not checked but virtio would sound to me like something belonging to kvm not xen
[12:40] <mwk_> smb, it does doesn't it -- do you know of an equivalent way of getting good entropy into an ec2/xen vm?
[12:41]  * henrix -> lunch
[12:47] <smb> mwk_, hm, not having thought too much about that. probably also depends on hvm vs. pvm (which most ec2 instance types are). Thought /dev/urandom would be producing enough randomness though sometimes take a bit to build up entropy
[12:49] <mwk_> my application is specifically for key generation -- i would prefer hwrng if at all possible -- at the moment my only real alternative is the cryptographically secure prng in openssl
[13:17] <smb> mwk_, I don't believe there will be better options. Even if it would be possible to use virtio-rng, that would require some support from the qemu used. And ec2 is using Xen 3.x which would not have any newer features. And from all I can find /dev/urandom sounds like the only option. But you could probably ask on the xen-user (or xen-devel) mailing list to get a better answer.
[13:30] <mwk_> smb, thx for your feedback
[13:32] <rtg> ogasawara, while reading https://wiki.ubuntu.com/KernelTeam/ReleaseStatus/Quantal I noticed the first bullet item is misspelled; 'Overall Statusbj'
[13:33] <rtg> which I guess _I_ can fix since its a wiki. doh!
[13:33] <ogasawara> rtg: hrm, thanks.  must have been some copy and paste error on my part.  I can fix it.
[13:34] <rtg> ogasawara, take it away...
[13:36] <ogasawara> rtg: welcome back btw
[13:36] <ogasawara> rtg: I assume you're drowning in your inbox
[13:42] <rtg> ogasawara, yep, 600+ including SPAM
[14:14] <infinity> apw: The buildd have 30G (which is overkill) swap partitions on the same spindle as the system, yeah.
[14:15] <apw> cooloney, ^^
[14:53] <cooloney> infinity: is it a swap file or a swap patition? i just allocated 2G partition for building test
[14:54] <cooloney> 2G swap
[14:54] <infinity> cooloney: Yeah, 2G won't be enough.
[14:54] <ogra_> what is he building ? 
[14:54] <ogra_> libO ?
[14:54] <infinity> cooloney: And it's a swap partition, but file wouldn't impact things awfully if you had to.
[14:55] <cooloney> infinity: ok, i will setup a 30G swap partition.
[14:56] <cooloney> ogra_: i'm trying to build several packages like kdepim, firefox, chromium-browser, webkit and gcc-4.6 on imx6
[14:56] <ogra_> ah
[14:58] <cooloney> infinity: did you open my google doc about the build fail
[14:59] <cooloney> ogra_: imx6 has 4 cores, we wanna test to see whether it is worth to replace panda buildd with it.
[15:00] <infinity> cooloney: I will later, I'm in meetings this morning.
[15:00] <ogra_> cooloney, i know, i have one (that i just brocked and recovered)
[15:00] <ogra_> *bricked
[15:03] <Kano> hi apw , could you enable aufs with your next kernel upload based on 3.5.2?
[15:05] <apw> we wouldn't be planning on enabling aufs unless we have a usecase for it, its one more thing to support
[15:05] <cooloney> ogra_: did you try to build some packages on it before?
[15:05] <ogra_> cooloney, only the kernel ... 
[15:05] <ogra_> which isnt really representative 
[15:05] <ogra_> but it was twice as fast as a panda doing it :)
[15:05] <Kano> apw: aufs is much older and better tested than overlayfs. if you use overlayfs in casper by default it does not hurt you
[15:06] <apw> Kano, if its not in my kernel i don't have to security support it for 18m
[15:09] <dileks> apw: built linux-next with overlayfs.v14 but didnt run any tests so far - so compile-tested
[15:10] <cooloney> ogra_: it just took about 42 minutes for me to build a kernel on imx6
[15:11] <ogra_> cooloney, i build the package in exactly 100min
[15:12] <cooloney> ogra_: probably, you use chroot, right? i just build it natively
[15:12] <ogra_> i built natively 
[15:12] <ogra_> the mx6 kernel package from your mx6 branch
[15:13] <ogra_> (including the udebs, headers etc)
[15:13] <Kano> ogra_: mx6?
[15:13] <ogra_> yep
[15:13] <ogra_> freescale sabrelite
[15:14] <apw> cooloney, we also need to do some 48hour soak tests to confirm they are stable over long perious of use, building all the time
[15:14] <cooloney> ogra_: or really, 100 min?
[15:15] <cooloney> apw: yeah, but i still faced some lockup on my board. 
[15:15] <ogra_> cooloney, i did five builds in a row (until i bricked the board) ... all with the same result ... 
[15:15] <apw> cooloney, well if thats the outcome, they lock up if used continuiously, then that is the result of a 48hr test, FAIL
[15:15] <cooloney> ogra_: did you meet some lockup when you building the kernel. like
[15:15] <apw> cooloney, and we know they arn't suitable
[15:16] <ogra_> cooloney, nope, it ran stable (but using the shipped uImage atm, you might remember i didnt get it to boot your kernel)
[15:16] <cooloney> apw: right. i didn't see such issue on the board in Portland.
[15:16] <cooloney> apw: the board i'm testing is borrowed from ericm|ubuntu, 
[15:17] <cooloney> rtg: could you please also do some package building on your imx6 board?
[15:17] <apw> cooloney, you likely didn't test for such a long period either over there
[15:17] <rtg> cooloney, I've done a few, but I have yet to setup an sbuild environment.
[15:18] <ogra_> apw, i tested across 48h but it didnt have constant load 
[15:18] <cooloney> apw: building webkit will take like 48 hours, i think. 
[15:18] <ogra_> (the five kernel package builds randomly over two days )
[15:18] <cooloney> apw:  i will try that soon
[15:19]  * cooloney needs go to sleep.
[15:28]  * ogasawara back in 20
[15:28] <infinity> ogra_: Testing with the Freescale kernel is, of course, not useful, since it's not what we'd be running in production.
[15:29] <bjf> ogasawara: bug 1013270 is the video bug QA were referring to in the hangout
[15:29] <ubot2`> Launchpad bug 1013270 in linux "if no monitor is connected, Nouveau fails to initialise" [High,Triaged] https://launchpad.net/bugs/1013270
[15:29] <ogra_> infinity, well, once io found a way to make u-boot not explode with devicetree issues i will hgappily try out our kernel :P
[15:30] <rtg> jsalisbury, I uploaded a fix for bug #1032779, so you could take it off the top 10 (if Nick agrees that it fixes his problem)
[15:30] <ubot2`> Launchpad bug 1032779 in linux-firmware "Missing firmware for wireless chipset on pandaboard" [Medium,Fix released] https://launchpad.net/bugs/1032779
[15:30] <ogra_> having to recover from brickage wasnt really a pleasantly motivating experience though, so i moved to something else for the interim ;)
[15:32] <jsalisbury> rtg, ack
[15:59]  * rtg starts ripping apart a server for CPU upgrade. back in a bit....
[16:22]  * smb -> gone
[16:23] <bjf> jsalisbury: bug 1013270 should be on the hot list
[16:23] <ubot2`> Launchpad bug 1013270 in linux "if no monitor is connected, Nouveau fails to initialise" [High,Triaged] https://launchpad.net/bugs/1013270
[16:36] <bjf> ogasawara, jsalisbury, ^ is blocking _all_ desktop qa tests using nvidia hw for all series post Lucid, including mainline kernels
[16:39] <ogasawara> bjf: so it would seem a simple workaround until we resolve it is to make sure a monitor is connected?
[16:40] <ogasawara> bjf: or am I asking too much
[16:41] <jsalisbury> bjf, ack
[16:46] <bjf> hggdh: i'd like to continue any further discussions here so I don't have to repeat myself :-)
[16:46] <hggdh> bjf: OK
[16:49] <hggdh> ogasawara: we will need to get local to hook a monitor, there is no provision for monitors in the rack. There is also a dongle update for the KVM that, when installed, will solve the issue (we hope)
[16:49] <hggdh> ogasawara: this dongle will have connections for pretty much all video outputs
[16:49] <ogasawara> hggdh: is there an eta for that new dongle?
[16:50] <hggdh> ogasawara: there should be, retoaded would know. I do not.
[17:12] <ogasawara> hggdh: and just so I have it clear in my head, when exhibiting this issue, the system in question is just connected to a kvm swtich, but no monitor is attached to that switch correct?
[17:13] <hggdh> ogasawara: the switch does have a monitor
[17:27] <ogasawara> hggdh: so I wanted to post a comment to the bug, but I'm still a little confused as to where the "if no monitor is connected..." part comes into play then.
[17:28]  * ppisati bails out
[17:30] <hggdh> ogasawara: the machines connect to a KVM (in this case using a DVI<-> VGA converter; the KVM has a local monitor that allows one to connect and use a system; it also allows one to remotely connect to the KVM, and select a system to work on
[17:32] <hggdh> amd firepro V5900 prints out a few EDID errors and is done; system can be used from the console. nVidia keeps on printing EDID errors, making console usage impossible (server images, 3.5.0-10, now installing desktop)
[17:45] <ogasawara> hggdh: so a monitor *is* connected (albeit via a kvm switch) when you're seeing these repeated EDID errors.
[17:46] <jwi> hggdh: maybe try https://patchwork.kernel.org/patch/1301341/ ?
[17:54] <rtg> apw, what is going on with 'UBUNTU: SAUCE: khubd -- switch USB product/manufacturer/serial handling to RCU' in Quantal ? Is it still relevant ? it caused a minor rebase conflict for 3.5.2 which is what caused me to look at it.
[17:55] <hggdh> ogasawara: indeed
[17:56] <ogasawara> hggdh: so the bug report then is completely misleading with the title and description being "if no monitor is connected..."
[17:57] <hggdh> ogasawara: I agree
[17:59]  * rtg -> lunch
[18:34] <apw> rtg, it was a performance enhancement, i can try and do the rebase if you like
[18:35] <rtg> apw, oh, the resolution is easy. I just wondered if it was making its way upstream
[18:35] <apw> rtg, that one is on my list to re-test and confirm it still provides benefit
[18:35] <apw> and to resumbit if so
[18:36] <rtg> apw, I'll just leave it in Quantal for now
[18:36] <apw> thanks
[18:54]  * ogasawara lunch
[19:07] <retoaded> ogasawara, hggdh: no eta on the new KVM CIMs (dongles) yet. the rfq hasn't been completed yet; once it is completed it will be submitted for PO then ordered. That should all be done in time for me to head up right after Labor Day weekend to start installing them.
[19:22] <ogasawara> retoaded: ack, thanks
[20:14]  * rtg -> EOD