**** BEGIN LOGGING AT Tue Sep 20 02:59:57 2011 Sep 20 04:54:41 wired internet its only letting me use wireless and when i go into a terminal and type in a command my eth0 does not show up Sep 20 04:55:30 what can i type in my terminal to see if my ethernet port is working Sep 20 04:55:57 what can i type in my terminal to see if my ethernet port is working Sep 20 12:56:22 GrueMaster: since you do a lot of testing, have you ever tried unplugging the sd card after boot? (of course when it's not mounted) Sep 20 12:56:49 GrueMaster: or have you tried booting entirely off some other media, and when the system is up, have you tried inserting the sd card? Sep 20 12:56:51 he did some stuff with that on netinstalls iirc Sep 20 12:56:57 and had funny results Sep 20 12:56:57 ok Sep 20 12:57:03 same here Sep 20 12:57:23 hrw opened a bug about sd being inacceisble if he removes/reinsert the sd card Sep 20 12:57:27 well, wait until janimo's drop of the vfat mangling show up in the images Sep 20 12:57:41 no no Sep 20 12:57:43 though wait, you dont use preinstalled, right ? Sep 20 12:57:57 i use preinstalled Sep 20 12:58:08 how can you unplug the rootfs then ? Sep 20 12:58:15 sheer force Sep 20 12:58:18 yeah Sep 20 12:58:25 or abuse of the images ! Sep 20 12:58:31 becasue i moved averyrthing to usb Sep 20 12:58:39 he doesnt use them in their intended environment ! Sep 20 12:58:45 evil guy you :) Sep 20 12:58:46 i use the sd card just for uImage&c Sep 20 12:58:52 :) Sep 20 12:58:56 well Sep 20 12:59:01 i would recommend using a netinst for such setups Sep 20 12:59:26 while it should indeed work to copy the rootfs to another device you never know what mistakes you make doing that Sep 20 12:59:27 but i just found out, that if i boot from another media (no sd card inserted during boot) Sep 20 12:59:31 the sd slot is dead! Sep 20 12:59:47 complain to u-boot or x-loader i guess :) Sep 20 12:59:49 it doesn't recognize any sd insert&c Sep 20 12:59:57 uhm Sep 20 13:00:00 smells very much like either of them Sep 20 13:00:13 that's why i asked GrueMaster, i have to try with some older release Sep 20 13:00:14 did you try with an older u-boot and x-loader ? Sep 20 13:00:15 and on xm now Sep 20 13:00:18 nope Sep 20 13:00:27 just oneiric/panda for now Sep 20 13:00:33 i bet a beer that it works fine with nattys Sep 20 13:00:44 i already own you a beer BTW :) Sep 20 13:00:57 you can win it back now ;) Sep 20 13:01:01 asd :) Sep 20 13:01:03 ok Sep 20 13:01:06 *g* Sep 20 13:01:19 coffee and then back testing older releases... Sep 20 13:22:21 ogra_: its not uboot/xloader - kernel rather Sep 20 13:22:35 system booted, replace card = bug Sep 20 13:22:52 hrw, i think tobin said he doesnt see it when rolling back to an older MLO/u-boot binary Sep 20 13:22:52 hrw: have you tried the same with beagle? Sep 20 13:22:55 load kernel/initrd in uboot, 'bootm', replace card - bug Sep 20 13:23:09 ppisati: would have to dig out old C3 beagle Sep 20 13:23:19 hrw: no prob, i'll do that Sep 20 13:23:32 hrw: another thing that i discovered is that: Sep 20 13:23:51 1) if you boot from another media (load everything in memory, remove sd card and bootm) Sep 20 13:24:05 if you tru later to insert the sd card, the kernel doesn;'t recognize it Sep 20 13:24:18 2) enabling MMC_DEBUG i see what it seems "activity" on mmc Sep 20 13:24:25 even when it's not mounted Sep 20 13:24:32 i wonder if it's polling Sep 20 13:24:38 the slot for some reason Sep 20 13:24:47 * ogra_ thinks the kgeneral bug is that the kernel leaves to much initialization to x-loader Sep 20 13:25:58 we ran into bad stuff using a different x-loader through that (the usbboot one) ... you cant use our kernel *at all* if you dont use MLO/u-boot (doesnt work with generic android bootimages for example because the majority of devices stays unintialized) Sep 20 13:26:34 same would go for kexec Sep 20 13:27:11 (using a binary kernel instead of u-boot.bin) Sep 20 13:58:49 ogra_, you were right, it is lz compressed. It worked for me before as I figured it out after much trial. But I forgot since :( Sep 20 13:59:01 anyway, going forward Sep 20 13:59:11 yeah, i remember banging my head against the same issue a while ago Sep 20 13:59:11 )win 21 Sep 20 13:59:20 I want every boring task on Earth be done by machines Sep 20 13:59:37 which is most of them Sep 20 13:59:50 heh Sep 20 14:12:47 hello Sep 20 15:17:10 ogra_, infinity, slangasek : would any of you be will to participate in the review of : http://revu.ubuntuwire.com/p/libjpeg-turbo ? Aiming for oneiric, universe, thanks! Sep 20 15:28:56 hrw: it works on xm! Sep 20 15:29:45 cool Sep 20 15:29:48 but I ahve pandas Sep 20 15:30:19 :) Sep 20 15:31:56 I just got my pandaboard booting Sep 20 15:32:05 the 11.04 release wouldn't boot properly for me Sep 20 15:51:09 ppisati: I see you are looking at the SD bug I mentioned a while back. Bug 844099 Sep 20 15:51:10 Launchpad bug 844099 in linux-meta-ti-omap4 "System fails to acknowledge changing of SD when rootfs is on a different device." [Undecided,New] https://launchpad.net/bugs/844099 Sep 20 15:56:33 ogra_: Reading the backscroll, what are you talking about on the usbboot not working? It works fine since rsalveti updated the aboot bootloader. Sep 20 16:06:09 GrueMaster: never seen that bug, i was working on hrw bug (that is a duplicate of yours actually) Sep 20 16:09:37 the world needs more arm Sep 20 16:11:28 is the natty process the update manager? Sep 20 16:12:28 brandini: ??? That last sentence made no sense. Sep 20 16:24:48 GrueMaster: what is the natty process? Sep 20 16:25:46 in what? Sep 20 16:26:06 you mean update? Sep 20 16:26:36 brandini: What exactly are you trying to do? Sep 20 16:30:48 GrueMaster, i was referring to the need to fix it ... Sep 20 16:31:20 ogra_: Fix what? It is already fixed. Sep 20 16:31:21 pointing out that way to many things are in the bootloader while they should rather (or additionally) be initialized in the kernel Sep 20 16:31:41 Oh. Sep 20 16:31:43 yes, i wasnt referring to the fix/bug at all Sep 20 16:31:54 just to the fact that it actually needed to happen Sep 20 16:33:23 for example theoretically it should be possible to add a hex header to a vmlinuz that can do kexec and use that instead of u-boot/MLO to chainload a kernel Sep 20 16:33:41 but with the existing design that will never be possible without at least involving MLO Sep 20 16:34:09 (though i suspect many things u-boot initializes wont be done by the kernel either) Sep 20 16:34:51 sigh, the images are still building Sep 20 16:35:31 over 4h already Sep 20 16:35:35 afaik, MLO/aboot just initiallizes the main memory and a few minor other bits (i2c bus). u-boot/kernel does the bulk. Sep 20 16:36:01 well, wasnt the fix in the MLO code of aboot ? Sep 20 16:36:16 now that didnt make sense Sep 20 16:36:24 well, wasnt the fix in aboot ? Sep 20 16:36:29 that is better :P Sep 20 16:36:29 But I also thought u-boot turns off everything it needs just before context switch to kernel. Sep 20 16:36:44 i dont think it does Sep 20 16:36:53 The fix was in aboot to initialize i2c iirc. Sep 20 16:37:00 yeah Sep 20 16:43:56 GrueMaster: I'm doing an update to 11.04 using the update manager (which was a silly mistake) and I'm watching the process headless from remote trying to monitor when it finishes Sep 20 16:44:51 Ah. I usually just run "sudo apt-get update;sudo apt-get dist-upgrade" to pull the updates. Sep 20 16:44:52 dist upgrade on SD card on an XM ? Sep 20 16:45:16 SD Card :( Sep 20 16:45:26 oh, not a release->release one Sep 20 16:45:44 Doing a full dist upgrade (Maverick->Natty) is a lesson in patience. Sep 20 16:45:46 I'm stopping by microcenter on the way home to pick up a usb->mSata adapter Sep 20 16:45:51 GrueMaster: it is :) Sep 20 16:45:53 i thought you upgrade to oneiric, that would indeed be a waste of time Sep 20 16:46:07 really? Sep 20 16:46:10 yeah Sep 20 16:46:14 10.10 is better than 11.04? Sep 20 16:46:17 shoot :) Sep 20 16:46:38 No, doing a release upgrade is a waste of time. Faster to download a new image. Sep 20 16:46:51 yeah Sep 20 16:46:56 ok, that makes more sense Sep 20 16:47:06 Doing an in-release update is however a good idea. Just time consuming. Sep 20 16:47:12 do I have to do anything special to make the pandaboard boot off a usb device? Sep 20 16:47:42 in-release is fine you just shouldnt try to do something else as well on the system :) Sep 20 16:47:53 heh, yeah :) Sep 20 16:47:58 the load is like 4 Sep 20 16:48:02 on the ac100 i actually dedicate time to it when i dont work Sep 20 16:48:36 If you want to use your existing image, you will need to do it on a different system. What I found works best is to attach both USB & SD to your desktop (not mounted) and use gparted to copy the rootfs partition. Sep 20 16:49:14 Then you need to change the uuid of the rootfs partition on the SD and just reboot. Sep 20 16:49:35 The panda will need the SD for boot partition. Sep 20 16:50:07 ok, that makes perfect sense Sep 20 16:50:16 why wouldnt you just copy qemu-system-statci, chroot and do the dist upgrade that way ? Sep 20 16:50:34 oh, you meant copying the rootfs Sep 20 16:50:35 You can also try our netinstall for oneiric. It will allow you to install to the usb drive directly with different filesystems (EXT4, btrfs, etc), use LVM, cryptfs, etc. Sep 20 16:50:40 ogra_: this is my first time goofing with these types of boards... so beginners luck? Sep 20 16:51:50 The gparted copy method is what I used to create a USB drive with Maverick, Natty, and Oneiric all on one USB drive. Makes SRU testing much easier. Sep 20 16:52:02 I thought this would be a great platform to serve some web stuff up for monitoring and managing my solar/wind power stuff Sep 20 16:52:27 Now that sounds like a fun idea. Sep 20 16:52:44 it is Sep 20 16:53:02 brandini, http://www.grawert.net:81/ Sep 20 16:53:04 ;) Sep 20 16:53:06 Not sure how well everything for that will run under Natty. I only did extensive server testing in Oneiric. Sep 20 16:53:10 I've got the web application built in go (#golang) and it serves things up really nicely Sep 20 16:53:22 though thats runing on an x86 celeron Sep 20 16:53:36 (simply because i had no beagle back then) Sep 20 16:53:48 hey, that's got solar/thermal :) Sep 20 16:54:05 how hot do your panels get when you're running water behind them? Sep 20 16:54:16 oh, it says right on it Sep 20 16:54:22 * ogra_ has a web based room heating control system too, that actually runs off a beagle C4 Sep 20 16:54:52 That's exactly what I'm fleshing out! Sep 20 16:55:02 * brandini introduces himself to ogra_ Sep 20 16:55:06 well, the panel bursted, i havent gotten a replacement yet (condesed water froze inside and blew up one element) Sep 20 16:55:20 it can get up to 130°C Sep 20 16:55:26 sheesh Sep 20 16:55:32 using copper? Sep 20 16:55:39 the panels are supposed to survive up to 220 Sep 20 16:56:04 efficiency drops quickly Sep 20 16:57:04 yeah, to be honest i would rather have half of the collector replaced by power generating panels Sep 20 16:58:00 I just built a single wind generator using some PVC Pipe and I'm quite pleased with it Sep 20 16:58:34 3 19" blades and a 30V DC motor... really works nice Sep 20 16:58:55 cool Sep 20 16:59:09 where you from ogra_? Sep 20 16:59:16 germany Sep 20 16:59:58 I'm from the states in ohio Sep 20 17:11:52 hey hey Sep 20 17:11:55 :) Sep 20 17:12:16 GrueMaster, maybe we need an alarm, loud and red and blinking Sep 20 17:12:20 soo ...binary packages often have arch: any and arch: all components Sep 20 17:12:29 ogra_, right Sep 20 17:12:38 the any components usually get built nativelyy Sep 20 17:12:53 the all components all get built by the x86 builder Sep 20 17:13:15 now the x86 builders are waaaay faster than the armel ones Sep 20 17:13:25 right Sep 20 17:13:28 so rthey finish the package early ... and publish their stuff Sep 20 17:14:07 arm simply doesnt have the new bits yet ... but the existing package has a versioned dep on the arch:all package (which was just updated by x86) Sep 20 17:14:29 so you get uninstallable packages until armel has built Sep 20 17:14:39 this is messy... Sep 20 17:14:42 which in some cases can take a day more Sep 20 17:14:45 but I got it now Sep 20 17:14:50 right Sep 20 17:14:54 and that prevents images from building Sep 20 17:15:10 there are two ways around it ... Sep 20 17:15:15 but you said there are no images since Aug 30th Sep 20 17:15:19 fix soyuz to handle the packages right Sep 20 17:15:38 or work around the whole issue by running a separate mirror (like linaro does) Sep 20 17:15:45 we aim for the fix Sep 20 17:15:52 ogra_, is there a bug for that? because if that's causing lack of images like this, it's kind of critical Sep 20 17:16:26 and you might want to talk to the launchpad stakeholder in Ubuntu to have it fixed Sep 20 17:16:47 i think there is a bug (NCommander should remember the number, i'm not subscribed so i dont have bugmail for it) Sep 20 17:17:10 and its known since quite a while ... even in higher levels ;) Sep 20 17:18:01 Ursinha, its been raised. https://bugs.launchpad.net/launchpad/+bug/34086 Sep 20 17:18:02 Ubuntu bug 34086 in launchpad "removal of arch-all packages while there are arch-specific packages dependent on it results in uninstallable binaries" [Critical,Triaged] Sep 20 17:18:36 Wow, that is very old. Sep 20 17:18:47 indeed Sep 20 17:19:47 but awesome that's already escalated :) Sep 20 17:21:43 after... 5 years Sep 20 17:22:44 rsalveti, better late than even later Sep 20 17:22:47 :) Sep 20 17:22:52 well, its a heavyweight ... pushing it uphill needs many people apparently :) Sep 20 17:23:12 sure, just surprised it took so long Sep 20 17:23:18 took 5 years to get the crowd together ;) Sep 20 17:23:42 ogra_, hehe, it's escalated, and people are apparently wrapping up the derived distros feature (which took some time to get completed) Sep 20 17:24:19 rsalveti, launchpad touches lots of aspects of ubuntu, it's hard to fix everything relevant in a reasonable time, I believe Sep 20 17:25:18 yup, probably Sep 20 17:25:27 but I also want the derived distro working properly ;-) Sep 20 17:25:41 that reminds me I need to ping some folks at launchpad Sep 20 17:28:36 GrueMaster, janimo, does mx5 need flash-kernel.conf ? Sep 20 17:29:09 * Ursinha adds #ubuntu-arm to the list of channels-to-lurk-in Sep 20 17:29:49 Aha. That's the problem. On mx5, boot.scr, and uI* is on the second partition. Sep 20 17:30:21 yeah, that shoudl get a flash-kernel.conf then :) Sep 20 17:30:27 so you can set it Sep 20 17:30:57 but i wonder if the code uses it yet Sep 20 17:31:00 Trying to run it after manually fixing flash-kernel.conf. Sep 20 17:31:14 We'll see if oem-config still crashes. Sep 20 17:31:15 yeah, that should tell Sep 20 17:50:03 GrueMaster, I am working on that now Sep 20 17:50:15 cool Sep 20 17:50:53 ogra_, uboot is not on the vfat partition but kernel and initrd are so presumably needs some flash-kernel.conf Sep 20 17:51:01 yeah Sep 20 17:52:35 I wonder why resizing rootfs on mx5 is done in a blink Sep 20 17:52:45 IIRc on omap it had a progess bar and took a while Sep 20 17:52:56 and it didnt fail ? Sep 20 17:52:57 maybe it doesn ot happen for some reason - bug to be hunted Sep 20 17:53:05 ogra no error msg, just Done Sep 20 17:53:11 It may not be resizing the correct partition. Sep 20 17:53:15 It is suspicious probably does not happen Sep 20 17:53:16 and your partition has the expected size ? Sep 20 17:53:31 GrueMaster, ++ Sep 20 17:53:35 will check after Sep 20 17:53:51 still I;d expect e2fsresize to complain if you ask it to resize a vfat Sep 20 17:53:54 GrueMaster, i bet i added a variable for that too ;) Sep 20 17:54:03 janimo: Small card? Sep 20 17:54:10 infinity, default mx5 Sep 20 17:54:12 janimo: On my 1GB card, resize on my Panda is literally instant. Sep 20 17:54:14 janimo, well, it would, in the jasper log Sep 20 17:54:17 That was changed as part of the ext4 change. Sep 20 17:54:18 resize is supposed to happen from 2g to 4g Sep 20 17:54:59 init: mounted-proc main process (385) terminated with status 1 Sep 20 17:55:06 mountall: Event failed Sep 20 17:55:11 I wish I knew why these happen Sep 20 17:55:20 no plymouth i guess Sep 20 17:55:32 for the mountall bit Sep 20 17:55:49 no idea about the upstart one Sep 20 17:55:50 ogra_, when is the next round of respins? Sep 20 17:55:59 and only for mx5 for now is omap done for b2 ? Sep 20 17:56:04 janimo, well, if you ask GrueMaster i guess he doesnt want one :) Sep 20 17:56:12 great, I agree Sep 20 17:56:13 janimo, thats not how it works Sep 20 17:56:36 if you upload jasper we need to re-test all jasper using images Sep 20 17:56:45 at least up to the jasper part Sep 20 17:57:04 yep. Sep 20 17:57:11 ok, there is a new jasper needed for mx5 of course Sep 20 17:57:11 we dont cherry pick arches for builds during milestones Sep 20 17:57:18 yes Sep 20 17:57:30 * ogra_ wasnt expecting todays images to be final actually Sep 20 17:57:30 Although I am still pulling, so if a respin is required I am ok. Sep 20 17:57:54 we also need to fix the missing slideshow still Sep 20 17:58:00 post beta though Sep 20 17:58:10 and the missing ti icon Sep 20 17:58:16 ogra_, when are the jasper hooks executed? Sep 20 17:58:20 for the mlabel copying Sep 20 17:58:32 Do we have packages for the ti icon to install? Sep 20 17:58:42 one in local-premount, one in local-bottom Sep 20 17:59:06 GrueMaster, no, its a jasper thing, but the handling of the favorites completely changed, so it doesnt show anymore Sep 20 17:59:27 GrueMaster, iirc persia had the bug assigned to move that into packages Sep 20 17:59:56 I meant when you click on the icon, is there packages in the ppa? Sep 20 18:00:04 it is there, you just dont see it :) Sep 20 18:00:11 and no, there is no package atm Sep 20 18:00:22 i'll add that by release time (if TI doesnt) Sep 20 18:01:05 Yea! Fixing /etc/flash-kernel.conf fixed the oem-config crash. Sep 20 18:01:12 awesome ! Sep 20 18:01:29 And I have unity. Sep 20 18:01:35 wow Sep 20 18:01:40 This is on yesterday's image though. Sep 20 18:01:50 Well, unity-2d. Sep 20 18:02:32 And the install icon is still there. sigh. Sep 20 18:02:57 ubiquity ? Sep 20 18:03:08 well, oem-config didnt finish Sep 20 18:03:54 did you actually see it removing packages ? Sep 20 18:05:05 I wasn't paying attention. Too many monitors to see all the details on all systems. But I have seen this before. Sep 20 18:06:06 interesting. When ubuntu-bug pops up to report a detected crash, it blanks the screen and asks for sudo access. Sep 20 18:06:23 bug in gksudo i think Sep 20 18:06:32 thats supposed to be transparent Sep 20 18:06:41 (teh black bg you see) Sep 20 18:07:03 poke mvo about it, i did it several times, he said he doesnt need a bug, it would go away anyway Sep 20 18:07:25 if it didnt go away yet, he might want to know about it :) Sep 20 18:10:35 It appears that the resize worked, but the system thinks the image is using 3.92G on my 4G SD. Will try again with a 16G SD later today. Sep 20 18:10:49 weird Sep 20 18:10:53 well ... Sep 20 18:11:00 * ogra_ is off to find some dinner Sep 20 18:11:17 Does it think there's free space anywhere? Sep 20 18:11:17 Cause 4G cards aren't 4G... Sep 20 18:15:49 No, it is a resize issue. du-sh / says it is only using 1.4G Sep 20 18:17:35 df -h shows / as 1.5G with 1.4G used. Gparted shows mmcblk0p3 as 3.95G Sep 20 18:17:52 jasper.log ? Sep 20 18:18:43 n/a Sep 20 18:20:28 Found it in /run. Tried to resize mmcblk0p2. Fail. Sep 20 18:20:44 no further info ? Sep 20 18:20:44 Right, because that partition number is hardcoded in jasper. Sep 20 18:20:53 Of course, this is yesterday's image. Prior to jasper fix for ext4. Sep 20 18:20:57 And p3 is the one you're using. Sep 20 18:22:01 its not hardcoded Sep 20 18:22:41 i wish i could have hardcoded it ... but certain people at linaro wanted to use jasper, so it has that weird detectuion code that parses root= for finding the disk Sep 20 18:22:46 ROOTPART="${ROOTDEV}2" Sep 20 18:22:50 ^-- That's not hardcoded? Sep 20 18:23:32 hmm, that should be dreived from $DISK Sep 20 18:24:36 ROOTDEV="/dev/${DISK}${SEP}" Sep 20 18:24:38 and it is Sep 20 18:24:43 Dude. Sep 20 18:24:44 The 2. Sep 20 18:24:52 The partition NUMBER is what's hadcoded. Sep 20 18:24:54 hard* Sep 20 18:25:00 yeah, sorry Sep 20 18:25:03 And mx5 is using a different layout. Sep 20 18:25:08 Kinda indicates structured fail. Sep 20 18:25:09 Or, so I understand. Sep 20 18:25:25 yeah, thats definitely a bug Sep 20 18:25:34 the code should just parse root= Sep 20 18:26:01 mx5 uses a non-fs part1 for u-boot, fat for part2 (kernel, initrd), and rootfs on part 3. Sep 20 18:26:09 right Sep 20 18:26:35 but that doesnt need to be hardcoded at all given we set root= at image build time Sep 20 18:26:40 Ironically, if there's a UUID in play, things work correctly later... Sep 20 18:26:46 yes Sep 20 18:26:47 But only partially correctly. Sep 20 18:26:58 So, yeah, that whole mess needs a bit of a clean up. Sep 20 18:27:05 a bit ? Sep 20 18:27:06 I should have looked closer where I was in there for the FSTYPE thing. :/ Sep 20 18:27:06 :) Sep 20 18:27:17 it needs a rewrite Sep 20 18:27:26 thats why we had discussions about it in dublin Sep 20 18:30:29 This is where my comment on us getting bogged down in last weeks meeting applies. Sep 20 18:30:46 yup Sep 20 18:31:15 Well, I could fix most of this root-finding mess in jasper right now, but the diff might be a bit unpleasant for a freeze. Sep 20 18:31:26 Not faulting anyone, just that we get tied up with long, difficult tasks, and the simple stuff falls through the cracks. Sep 20 18:31:31 infinity, do it post beta Sep 20 18:32:03 ogra_: Sure, this just means mx5 gets to be broken for beta. *shrug* Sep 20 18:32:04 jasper is initial prototype code that happened to work right ... it was never intended to stay in production in that state for that long Sep 20 18:32:49 infinity, just add a three liner that checks /proc/cpuinfo and resets the partition number then Sep 20 18:33:12 another hack for a few days wont harm us ;) Sep 20 18:33:28 Ew. Sep 20 18:33:29 No. Sep 20 18:33:36 I'm either going to fix it right, or postpone it. Sep 20 18:33:45 The latter sounding like a better option. Sep 20 18:33:48 well, postpone means no mx5 at all Sep 20 18:33:49 Though, we can stage a fix now. Sep 20 18:33:56 "at all"? Sep 20 18:34:07 this is the critical milestone Sep 20 18:34:07 Looks like it now fails due to fstab. Sep 20 18:34:10 It's a community image, there's no quality requirement. Sep 20 18:34:20 well Sep 20 18:34:24 It's not like it was going to be on releases.ubuntu.com. Sep 20 18:34:26 we produce it... Sep 20 18:34:29 It'll always be on cdimage. Sep 20 18:34:55 i actually have some quality requirement for the work i ship ... at least on the surface :) Sep 20 18:35:34 infinity, point is that we agreed with the release team that images that arent ready by beta1 (actually) wont be releasable Sep 20 18:35:36 I do too. I'm just saying that non-release images have no such requirement. Sep 20 18:35:51 we talked kate into b2 Sep 20 18:35:53 Yeah, so we don't "release" it. We weren't anyway. Sep 20 18:36:00 Sort of my point. :P Sep 20 18:36:06 if we dont make it i wont step up again for it Sep 20 18:36:14 so we will miss Sep 20 18:36:27 Wait. Okay, are we talking past each other, or... Sep 20 18:36:36 Did you actually expect ac100/mx5 to be on releases.ubuntu.com? Sep 20 18:36:50 we were supposed to release it (opposed to having a daily on cdimage that gets wiped with the first P build) Sep 20 18:36:53 Where, y'know, half our images never go. Sep 20 18:37:01 no Sep 20 18:37:04 Oh, it can be "released" on cdimage. Sep 20 18:37:13 That's a whole different quality criteria, though. Sep 20 18:37:22 i expect the tested and QA^ approved release to be on cdimage in the r5eleases subdir of the respective image Sep 20 18:37:24 And I think people get muddied up about that sometimes. Sep 20 18:37:34 we only rarely have images on releases.u.c Sep 20 18:37:53 all armel images but one live on cdimage Sep 20 18:38:11 we had one release where we actually dumped them on r.u.c Sep 20 18:38:47 anyway, moot point, do you see a fix thats not a horrid diff and that can work around it ? Sep 20 18:39:12 Well, I'd have to see exactly where it's failing to come up with a "hack" instead of a fix. Sep 20 18:39:23 irregardless of the current arguements regarding release, this image is fail. It actually is mucking up the partitions now to the point where it won't boot. Sep 20 18:39:24 But... Does every preinstalled image have a sane root= on the command line? Sep 20 18:39:34 up to now they did Sep 20 18:39:35 And is it always a device? Or sometimes a UUID? Sep 20 18:39:45 always a device (see jasper code) Sep 20 18:39:55 jasper expects to not see a UUID Sep 20 18:39:58 Well, the jasper code assumes it might not be a device. :P Sep 20 18:40:04 (because it actually creates the UUID) Sep 20 18:40:18 On first boot, it is a device. jasper rewrites boot.scr to use UUID. Sep 20 18:40:23 right Sep 20 18:40:24 if echo "$root" | grep -q '^UUID='; then Sep 20 18:40:24 VOLID=$root Sep 20 18:40:24 else Sep 20 18:40:24 if echo $root| grep -q ^/;then Sep 20 18:40:24 ROOTPART=$(basename $root) Sep 20 18:40:26 fi Sep 20 18:40:35 If it's never a UUID, that code's broken. Sep 20 18:40:47 you look at the wrong script i think Sep 20 18:41:04 Look at premount Sep 20 18:41:53 there is UUID stuff, but thats unrelated to resize code Sep 20 18:42:59 Right, well same ROOTDEV2 issues there. Sep 20 18:43:06 And both can be solved with the same small amount of code. Sep 20 18:43:07 right Sep 20 18:43:16 then go for it Sep 20 18:43:17 But the above that I pasted is also an issue. :P Sep 20 18:43:27 It'll be writing the fstab incorrectly on mx5. Sep 20 18:43:27 janimo needs an upload anyway Sep 20 18:43:28 So. Sep 20 18:43:29 Yeah. Sep 20 18:43:41 I'll get on this in 30ish minutes, I'm starving. :) Sep 20 18:43:50 * ogra_ finishes his dinner Sep 20 18:43:51 infinity, I'll upload my changes now Sep 20 18:43:56 janimo, wait Sep 20 18:44:07 janimo: Or just commit them, if you don't need an upload/rebuild cycle. Sep 20 18:44:14 just commit them so we can get along with one upload Sep 20 18:44:16 ogra_, not uploading just committing Sep 20 18:44:18 sure Sep 20 18:44:20 ah Sep 20 18:44:22 janimo: (And I suspect rebuilding is useless to you without fixing the rootdev stuff) Sep 20 18:44:50 well, we have a good set for testing atm (beyond mx5 indeed) Sep 20 18:45:23 janimo: Looks good. Should fix all of our problems. Sep 20 18:45:26 I changed rootpart , uboot_part and boot.scr contents for mx5 Sep 20 18:45:38 GrueMaster, what was the flash-config change about? Sep 20 18:45:43 oh, rootpart too ? Sep 20 18:46:10 I tried with my changes in a modified initrd and it still crashed at the end albeit differently. When I came back to the screen it had restarted ubiquity Sep 20 18:46:18 just like I saw it done for omap previously Sep 20 18:46:42 but var/crash is empty this time so some joy Sep 20 18:46:54 ogra_, well rootpart to be part 3 and vfat part 2 Sep 20 18:47:00 yes Sep 20 18:47:03 as they were hardcoded to 1 and 2 for omap Sep 20 18:47:06 but see above Sep 20 18:47:25 infinity just wanted to start working on a general fix for that Sep 20 18:47:44 no problem Sep 20 18:47:50 oh, geez ! i'm sooo proud ! Sep 20 18:47:53 general fix being? Sep 20 18:48:00 janimo: I manually changed /etc/flash-kernel.conf to show UBOOT=/dev/mmcblk0p2. Sep 20 18:48:07 (its my cats first bday today and he just brought me his first mouse) Sep 20 18:48:09 I had a version that autodetected vfat/root using sfdisk Sep 20 18:48:21 Your jasper mods should fix this automatically. Sep 20 18:48:33 ogra_, cats do not know the customs. Unclear about who needs a present Sep 20 18:48:37 janimo, well, talk to infinity so you guys dont duplicate work then Sep 20 18:48:38 GrueMaster, good Sep 20 18:48:44 haha Sep 20 18:49:08 ogra_, I did not do the smartsfdisk based approach so I introduce as little new code as possible Sep 20 18:49:12 even if it looked better Sep 20 18:49:19 he still doesnt get that he can eat his toy though Sep 20 18:49:23 no need for grepping cpuinfo for instance Sep 20 18:49:39 how do you know you are on mx5 ? Sep 20 18:49:51 has MX53 in the hardware line Sep 20 18:50:02 oh you mean without cpuinfo Sep 20 18:50:10 you just said you dont look at cpuinfo Sep 20 18:50:14 yeah Sep 20 18:50:24 well for the partitions just look at where sfdisk finds vfat and linux Sep 20 18:50:28 and operate on those devices Sep 20 18:50:45 yeah, that at least works for all images with vfat Sep 20 18:50:46 but for the boot.scr content subarch is still needed Sep 20 18:51:11 so not entirely streamlined without special casings Sep 20 18:52:27 well, that sounds like you already implemented what adam planned to Sep 20 18:52:36 GrueMaster, but my fixes to jasper need testing with omap just in case I blundered something Sep 20 18:52:36 or at least something similar Sep 20 18:52:38 Someone going to upload this and trigger a respin, or do I call mx5 a cosmic failure? Sep 20 18:52:51 not yet :) Sep 20 18:52:52 ogra_, similar, not very elegant, for the same reasons - freeze being close Sep 20 18:52:58 yeah Sep 20 18:53:12 ogra_, I still need to add the mlabel thingie Sep 20 18:53:28 That can wait post-beta Sep 20 18:53:30 so when is that hook/ copy_exec executed? Sep 20 18:53:34 GrueMaster, fine by me Sep 20 18:54:15 yeah Sep 20 18:54:22 leave that for post muilestone Sep 20 18:54:44 janimo, during update-initramfs Sep 20 18:54:53 call update-initramfs -v ;) Sep 20 18:55:02 that shows you every copy_exec it does Sep 20 18:55:08 ogra_, ah so after install Sep 20 18:55:09 including the list of libs Sep 20 18:55:40 we call update-initramfs several times during the build too Sep 20 18:55:56 well, once at least, more depends on the image type Sep 20 19:03:31 janimo: We already know rootpart from root=, the sfdisk thing feels like overkill. Sep 20 19:03:38 infinity, ok Sep 20 19:03:49 but vfat needs to be detected too Sep 20 19:03:58 anyway, for now the case with MX53 works Sep 20 19:04:15 Is that in another commit? (the vfat thing) Sep 20 19:04:17 * janimo tries again with logging to the screen Sep 20 19:04:22 infinity, I commited all I have Sep 20 19:04:25 UBOOT_PART Sep 20 19:04:51 2 or 3 commits I think Sep 20 19:05:14 Ahh, that's not autodetected, though, just hardcoded for the subarch. Check. Sep 20 19:05:25 Should be on rev 165 now. Sep 20 19:05:57 Yeah. Sep 20 19:05:59 er, 167 Sep 20 19:06:10 165..167 are his changes. :P Sep 20 19:06:28 But yeah. I just understood the mentioning of "have to detect vfat" as meaning he'd done so. :) Sep 20 19:06:59 well, if its sufficient to get mx5 through Sep 20 19:07:04 Sec. Sep 20 19:07:14 So, I still need to know if we are respinning with this. Sep 20 19:07:20 We will. Sep 20 19:07:28 yep Sep 20 19:08:03 We'll have to respin server and kubuntu/kubuntu-mobile as well. All images except netboot & core. Sep 20 19:08:07 However, if we're sure that we'll always have a correct root=/dev/blah, then we can scrap that whole bit in _setup. Sep 20 19:08:25 ROOTPART="${ROOTDEV}2" Sep 20 19:08:26 #iMX53 has rootfs on part 3 Sep 20 19:08:26 grep MX53 /proc/cpuinfo >/dev/null && ROOTPART="${ROOTDEV}"3 Sep 20 19:08:30 well, we control what goes into the images Sep 20 19:08:33 ^-- Unnecessary because of: Sep 20 19:08:40 if echo "$root" | grep -q '^UUID='; then Sep 20 19:08:40 VOLID=$root Sep 20 19:08:40 else Sep 20 19:08:40 if echo $root| grep -q ^/;then Sep 20 19:08:40 ROOTPART=$(basename $root) Sep 20 19:08:53 we could as well make jasper read /etc/subarch and wipe it afterwards Sep 20 19:09:30 where is that? Sep 20 19:09:36 adding a full AI to have disk detection seems overkill Sep 20 19:09:38 jasper_setup Sep 20 19:09:56 No, the /etc/subarch part. Sep 20 19:10:01 Nowhere. Sep 20 19:10:05 He was saying we could create it. Sep 20 19:10:17 right Sep 20 19:10:22 So another added failure point. Lets not. Sep 20 19:10:28 to remove all subarch detection code Sep 20 19:11:10 Yeah, not worth it for now. Sep 20 19:11:13 well, its 20 lines of subarch detection by parsing /proc/cupinfo ... vs SUBARCH=$(cat /etc/subarch) Sep 20 19:11:26 heh, no, surely not for now Sep 20 19:11:32 i'm just thinking aloud Sep 20 19:11:34 janimo: If your changes work, they seem low impact for now, I just hate committing hacks unless we also commit to fix them later. :/ Sep 20 19:11:54 infinity, well, jasper is a lost case ... somewhat Sep 20 19:11:55 The right thing to do is to make jasper detect the correct partitions. This way, it will work with other platforms and also with linaro images on existing platforms. Sep 20 19:12:07 ogra_: Indeed. Sep 20 19:12:20 it is a hack consisting of hacks Sep 20 19:12:26 GrueMaster: Well, yes. Sep 20 19:13:00 GrueMaster, linaro has no interest in using or helping with jasper Sep 20 19:13:08 they pretty clearly told me so Sep 20 19:13:22 Only because it breaks their current model. Sep 20 19:13:36 they had no model when jasper stzarted Sep 20 19:13:38 They have a master partition on P1. Sep 20 19:13:41 also because they presumable wish to keep their sanity Sep 20 19:13:49 and asac worked quite hard to try to convince people Sep 20 19:14:01 GrueMaster, for mx5 I copied their part layout Sep 20 19:14:06 or uboot would not work Sep 20 19:14:18 janimo, jasper isnt a bad thing ... its just that we are still using the proof of concept prototype Sep 20 19:14:39 I wonder if the same layout could be used for omap/omap4. Sep 20 19:14:42 Well, jasper shouldn't exist. Sep 20 19:14:43 ogra_, the idea is not bad, but the hacks and the maintainability are Sep 20 19:14:59 The growroot concept should be rolled back into ubiquity, and we'd be done with it. Sep 20 19:15:03 infinity, what would do the resizing then ? Sep 20 19:15:34 That's the only thing it does different from ubiquity. It resizes in place instead of copying. Sep 20 19:15:39 Other than that, we WANT ubiquity. Sep 20 19:15:48 why would we Sep 20 19:15:56 we *use* ubiquity Sep 20 19:15:59 Why wouldn't we? Sep 20 19:16:12 we only dont use the partman bits Sep 20 19:16:30 and pkgsel Sep 20 19:16:32 Sure, but if it was all in one place, you could use one installer to install in-place OR to another drive. Sep 20 19:16:35 This debate really should be brought up at UDS. Too late for major changes now. Sep 20 19:16:43 And yes, not helpful now. Sep 20 19:16:44 haha Sep 20 19:16:48 janimo: If your hacks DTRT, upload. Sep 20 19:17:22 infinity, fix the copying speed and i'll happily be with you on live images Sep 20 19:17:41 ogra_: The current way to install to external drives on a Panda are either netboot (that's fine), or install with jasper and then cp -a to another partition. That second case is just wrong. :P Sep 20 19:17:47 * GrueMaster toddles off to play with the netinstall images while waiting for respins. Sep 20 19:17:56 preinstalled images mainly exist because the linaro guysw complained to me it takes to long to install Sep 20 19:18:06 that was way before they started to do their own Sep 20 19:18:08 ogra_: And this has nothing to do with copying speed. I'm saying the growroot (install in-place) option should be added to d-i proper. Sep 20 19:18:21 actually its all amitkÄs fault ! Sep 20 19:18:25 ogra_: So you could EITHER copy, OR do the in-place thing we do now. Sep 20 19:18:37 hmm Sep 20 19:18:46 yeah, that sounds sane Sep 20 19:18:49 infinity, I am still testing a bit the fs resize Sep 20 19:18:54 janimo: Mmkay. Sep 20 19:18:59 infinity: Growroot needs to happen before the rootfs is mounted. EXT4 doesn't allow grow while mounted. Sep 20 19:19:13 i think it does Sep 20 19:19:17 ext3 doesnt Sep 20 19:19:22 janimo: It's not how I would have solved it, but it's definitely low-impact your way, and I'd rather push something in and have working images than rewrite things right now, so... Sep 20 19:19:28 and i think its a lot faster when mounted Sep 20 19:19:32 No, I tested it during my EXT4 testing. Sep 20 19:19:47 infinity, out of curiosity how would you have done it? Sep 20 19:19:53 It actually produces an error. Even when mounted RO. Sep 20 19:20:23 janimo: Well, as above, I'd remove some redundant code, I'd use root= instead of sfdisk, I'd remove more redundant code... And more... Sep 20 19:20:28 janimo, i guess a udeb that brings that functionallity (using parted or partman) Sep 20 19:20:41 janimo: But my changes have a regression potential if things don't work exactly as I'm told they do. :P Sep 20 19:20:46 janimo: Yours should Just Work, for now. Sep 20 19:20:54 which then can be used by d-i and ubiquity/oem-config Sep 20 19:21:18 infinity, ah indeed. Seeing I removed some code for the VFAT formatting because it was the right thing, introduced a regression (label) I thought I resist the urge to clean up things Sep 20 19:21:51 janimo, just look at the jasper buglist, 90% of jasper wouldnt exist anymore if they were all fixed Sep 20 19:22:23 it would mainly boild down to growroot Sep 20 19:22:55 u-boot setup shoudl be done by flash-kernel-installer buit that needs a re-write to accept being run in non d-i envs Sep 20 19:23:18 and all the other setup bits should go elsewhere too Sep 20 19:23:42 we have a re-occuring jasper rewrite spec since we have preinstalled Sep 20 19:23:50 * ogra_ looks at infinity ... Sep 20 19:23:51 Heh. Sep 20 19:24:05 we should wprobabls have a jasper slaying spec this time :) Sep 20 19:24:11 *probably Sep 20 19:24:14 I'm inclined to agree. Sep 20 19:24:23 add it to the spec page ;) Sep 20 19:24:48 and dont let me implement it ... you might end up with jasper-ng :P Sep 20 19:25:04 Chris Jones shouldn't be near installers. Sep 20 19:25:12 We'd end up with multiple installer windows. Sep 20 19:25:16 heh Sep 20 19:26:46 ogra_, I remember, I took the notes at the meeting and filed most of those bugs in Dublin Sep 20 19:27:08 * ogra_ has a slight prob with chris' nick since it took him quite a while to make out that the guy who asked about "angie" from canonical IS actually meant chris :) Sep 20 19:27:32 infinity, ogra, uploaded jasper Sep 20 19:27:32 janimo, yeah, they are still there :D Sep 20 19:28:22 ogra_, closed the VFAT reformat one Sep 20 19:28:32 heh, cool Sep 20 19:28:56 i think you own a WI from my plate as well (or was it infinity) for the ext4 changes in jasper Sep 20 19:29:22 (i'll make sure to re-assign when i close it) Sep 20 19:32:34 janimo: Accepted. Sep 20 19:32:38 infinity, thanks Sep 20 19:34:15 * ogra_ calls it a day Sep 20 19:34:33 'Night. Sep 20 19:34:36 night Sep 20 19:35:03 ogra_: I hope you don't mind I shared that link with my wife because I wanted to share what I'm trying to accomplish Sep 20 19:35:23 ogra_, bye Sep 20 19:35:56 hmm, resize does not happen apparently. mount fails and the script quits Sep 20 19:36:06 mount: can't read '/etc/fstab': No such file or directory Sep 20 19:36:33 triggered by mount ${ROOTPART} /root >>${LOG} 2>&1 apparently Sep 20 19:37:24 Is that you stepping through manually, or? Sep 20 19:37:55 Cause at that point, on "proper" install media, /etc/fstab exists (albeit empty), and mount should be fine. Sep 20 19:38:29 infinity, no, running yesteday's image patched with a modified initrd to test my jasper changes Sep 20 19:38:46 I'll wait for proper images could be an artifact of something I did Sep 20 19:38:54 anyway resize is the last step in that file Sep 20 19:39:06 so the rootfs stays at 2G Sep 20 19:39:16 maybe that is why it crashes, no space for installed packages Sep 20 19:39:25 I test it on the mx53 Sep 20 19:40:31 janimo: yesterday's image worked for me once I modified /etc/flash-kernel.conf. Sep 20 19:40:46 GrueMaster, resized ext3 even? Sep 20 19:40:55 May be something with the ext4 stuff. Sep 20 19:40:59 maybe because ARM boards love you Sep 20 19:41:13 you are much closer to them than I am Sep 20 19:41:17 It can't have resized correctly. Sep 20 19:41:19 It didn't resize properly (rewrote the partitions but that was it). Sep 20 19:41:21 you cultivate a proper relationship with them Sep 20 19:41:25 ok Sep 20 19:41:35 so part resize is good, fs resize not Sep 20 19:41:46 Yea, I'm all touchy feely with them. Sep 20 19:42:26 This was yesterday's image prior to ext4 changes and everything else. Sep 20 19:43:04 Yeah, partition resizing would have always worked, because it just expanded "the Linux partition", it didn't use numbers. Sep 20 19:44:40 That would fail pretty entertainingly if anyone tried to use jasper in an image with multiple partitions. Sep 20 19:44:54 Thankfully, no one would do such a thing. :P Sep 20 19:47:58 ogra_: hehe :) I still think anybody thinking of installing ubuntu on ARM devices needs some introduction to the real world :-p Sep 20 19:49:03 ogra_: (as opposed to pre-installed images, that is) Sep 20 19:54:50 amitk: Well, depends on how you define "ARM devices" too. Sep 20 19:55:01 amitk: I think most of the madness we do with flash is, well, madness. Sep 20 19:55:17 amitk: But with devices that can actually use external storage, things (finally) change a bit. Sep 20 19:55:25 flash? Sep 20 19:56:30 brandini: Flash, SD, MMC, nvram, whatever you want to call it. "slow, unreliable storage that belongs in watches, not desktop computers". Sep 20 20:00:01 :) Sep 20 20:00:20 the choice to rely/depend on that for booting seems odd to me Sep 20 20:00:22 infinity: right, and when I complained a year ago, the only HW we had access to had flash storage and slow usb storage. Over one year later, we still don't have enough devices/boards in Linaro that can do SATA Sep 20 20:03:08 so I'm going to start updating openbsd's armish arch to support the pandaboard Sep 20 20:03:23 and the deeper I dig the more I wonder what I've gotten myself into Sep 20 20:04:13 brandini: Pain. Sep 20 20:04:21 Sure seems like it Sep 20 20:04:50 the current install guide for openbsd doesn't start actually talking about the install until about half way through the document Sep 20 20:48:48 infinity: Ouch (from the builders status page): armel 18 283 jobs (22 hours) Sep 20 20:49:10 GrueMaster: That't the rebuild test. And a huge improvement over when it said "12 weeks". Sep 20 20:49:22 ah Sep 20 21:15:17 infinity: Has the jasper update been published and accepted? No sense respinning without it. Sep 20 21:16:16 GrueMaster: Yes. Sep 20 21:16:53 ok. I knew we were waiting on ubiquity. Wanted to make sure jasper was also updated. Sep 20 21:17:13 Yeah, jasper builds in a matter of seconds. ;) Sep 20 21:17:39 Doesn't necessarily mean it is accepted during a freeze. Sep 20 21:17:52 (I accpted it) Sep 20 21:18:08 Ah. Sep 20 21:18:09 13:31 < infinity> janimo: Accepted. Sep 20 21:18:09 13:31 < janimo> infinity, thanks Sep 20 21:19:07 I was monitoring #u-release. Thought it would get an honorable mention there. Sep 20 21:19:13 But no matter. Sep 20 21:20:03 Nah, I reviewed it and slipped it under the radar somewhat intentionally. Sep 20 21:20:22 Not that it's a "secret" (obviously), just didn't feel the need for fuss, since we were waiting on ubiquity. ;) Sep 20 21:21:29 No problem. I just wanted to make sure. When I'm not keeping everyone on their toes, things slip by (like daily builds). Sep 20 23:30:27 Ok, I've procured my external usb enclosure Sep 20 23:31:13 it's shiny and black, but that wasn't part of the instructions :) Sep 20 23:32:00 brandini: Still seems important. Sep 20 23:32:24 it is Sep 20 23:32:46 man, I'm kinda bummed that I'm going to power off my pandaboard half way through this update from 10.10 to 11.04 Sep 20 23:32:50 :( Sep 20 23:51:24 is there something different I need to do when I install the 11.04 preinstalled image to a CF and SSD? Sep 20 23:54:20 With that image, you just dd to the SD card same as 10.10 and boot. After going through the oem-config and getting to a working image (before installing updates), shut down. Then you can transfer the rootfs to the USB drive using your desktop system. Sep 20 23:56:28 ok Sep 20 23:56:40 I should use 11.04 release or daily? Sep 20 23:57:35 If you want to start with Oneiric (11.10) on your USB drive, you can do a netboot install directly to it, but it takes a little bit of working. Sep 20 23:58:56 would I always have to netboot or just for the install? Sep 20 23:59:14 Essentially, there is a bug in the netboot install that fails to repartition the SD properly for normal booting. But it is easy to work around. See bug 806751 for info. Sep 20 23:59:15 Launchpad bug 806751 in debian-installer "Boot partition on SD is too small on omap/omap4" [Medium,New] https://launchpad.net/bugs/806751 Sep 20 23:59:27 Just the install. Sep 20 23:59:56 This will pull the latest packages. Once the system is installed, you can just pull updates as normal. Sep 21 00:04:35 The only problem is that it pulls only from ports.ubuntu.com (no mirrors). You can get around it if you have a local mirror, but that takes a large amount of drive space. Sep 21 00:06:26 maybe I'll stick with trying the headless daily Sep 21 00:07:23 s/headless/server/ Sep 21 00:07:35 server? Sep 21 00:07:52 I only see netbook and headless Sep 21 00:08:19 Not for oneiric. Sep 21 00:08:34 We're now more in line with the rest of the distro, with desktop and server. Sep 21 00:08:52 http://cdimage.ubuntu.com/ubuntu-server/daily-preinstalled/ Sep 21 00:09:07 Oneiric changes the names. Netbook>Desktop (daily-preinstalled). headless>Server Sep 21 00:10:20 brandini: If you want the netboot installer, it is at a different location. http://ports.ubuntu.com/dists/oneiric/main/installer-armel/current/images/omap4/netboot/ Sep 21 00:14:49 naw, I think I'll stick with server Sep 21 00:14:58 how come cdimage is so slow? Sep 21 00:58:52 infinity: Does this certificate thing invalidate the current images that were just rebuilt? Sep 21 00:59:03 For armel? Sep 21 01:01:23 GrueMaster: Almost certainly. Sep 21 01:01:23 GrueMaster: I think we ship that package on pretty much every image. Sep 21 01:01:23 Task: standard Sep 21 01:01:23 Yeah. Sep 21 01:01:23 Every image except core. Sep 21 01:02:20 * GrueMaster mutters a few choice words that get stuck in the irc filters. Sep 21 01:03:10 Well, may as well fix jasper (yet again). FS resize is failing. Testing server images now. Sep 21 01:05:58 It's failing on the server images? Sep 21 01:06:55 yes. It runs way too quickly (1-2 seconds) on a 16G SD. Sep 21 01:07:18 Waiting for the rest of the install to see what it is doing (or not doing as the case may be). Sep 21 01:08:03 heh. This is a definite indicator of fail: /usr/sbin/oem-config-firstboot: line 98: cannot create temp file for here-document: No space left on device Sep 21 01:08:11 16G SD. Sep 21 01:08:31 Not promising... Sep 21 01:09:01 Considering installed we use ~1.9G max... Sep 21 01:10:43 Not entirely sure how the changes would have made it fail. A log would be nice. Sep 21 01:11:56 Working on it as fast as these turbo systems will go. Sep 21 01:14:04 What is the kernel cmdline to break in init during initrd? Sep 21 01:16:52 Hate to say "I told you so", but this is kind of why I was against ext4 switch this late in the game. Sep 21 01:23:41 infinity: What is the proper way to break in init prior to premount? I can't see what is going on with jasper. Sep 21 01:27:05 We can switch back to ext3 with almost no effort. I can do it right now, if you think the failure is the filesystem. Sep 21 01:27:06 But if it's not the filesystem, you get to blame all the mx5 changes, and not me. :P Sep 21 01:27:06 And the init break would be something like "break=top" Sep 21 01:27:06 Or similar. Sep 21 01:27:06 Depending on which step you want to break into. Sep 21 01:27:18 break=premount, I would assume. Sep 21 01:27:27 Unless that happens after premount. Sep 21 01:27:37 * infinity hasn't hacked initramfs-tools in a while. Sep 21 01:27:39 You can just put "break" and it stops just before the pivot_root Sep 21 01:27:50 twb`: That's too early. Sep 21 01:27:52 Err, late. Sep 21 01:27:57 Brain bleeding. Sep 21 01:27:59 If you grep for maybe_break over /usr/share/initramfs-tools you can see all the break points Sep 21 01:28:27 Yeah. Just did. :P Sep 21 01:28:35 infinity: I hack the shit out of casper to make it work on diskless thin clients -- I know all about initramfs :P Sep 21 01:28:44 GrueMaster: You want break=premount Sep 21 01:28:54 got it. Trying. Sep 21 01:34:40 export ROOTDEV=/dev/mmcblk0p Sep 21 01:34:44 Looks wrong Sep 21 01:36:09 Nah, that's right, just unneeded now with jani's changes. Sep 21 01:36:21 Since rootpart used to be rootdev+n Sep 21 01:36:59 I can't see where it is failing. Sep 21 01:37:40 If ROOTPART is correct, then the only place it really could fail is just because resize2fs doesn't work on ext4 (which we know it does, but could be a bug or something) Sep 21 01:37:52 Or just a grumpy SD. My testing worked on one, and not another. Sep 21 01:37:59 PS: I hate flash memory. Sep 21 01:38:27 Let me grab a copy of this image and flash it. Sep 21 01:38:32 fails on both beagleXM and panda. different cards (4 now). Sep 21 01:39:58 I'm perfectly willing to believe there's a bug with resize2fs and ext4 on arm (or in general), just going to look here quickly as well and see if I can see how or why it fails. Sep 21 01:40:17 Like I said, switching back is easy (one small diff on antimony), now that jasper itself doesn't hardcode fs types. Sep 21 01:41:59 I don't think that is the problem. Sep 21 01:43:04 resize works fine manually from init. Sep 21 01:43:46 Running the script, or just typing resize2fs manually? Sep 21 01:44:04 running resize2fs manually. Sep 21 01:44:32 Also, is there a sane way to do kernel command line editing on these things, or do I have to mangle the image? Sep 21 01:44:47 Well, I guess, mangle boot.whatever on the card. Sep 21 01:45:50 That's what I do. You could interrupt u-boot also and do it manually. Sep 21 01:46:21 The filesystem on /dev/mmcblk0p2 is now 15589376 blocks long. Sep 21 01:46:25 I call that a success. Sep 21 01:46:40 And ext4? Sep 21 01:46:57 In that case, I don't think you get to blame me. :P Sep 21 01:47:05 But I'll see if I can figure out WTF *is* going wrong. Sep 21 01:48:02 exit Sep 21 01:48:10 grr. Sep 21 01:48:34 Ok, it seems to be booting through after manually massaging the partitions. Sep 21 01:48:44 jasper-growroot has issues. Sep 21 01:50:49 * infinity goes to meet a woman, get married, and have a few kids, while he waits for this image to flash. Sep 21 01:52:15 infinity: didn't realize flashing images is taking that long now ;) Sep 21 01:52:44 Doh!. Looks like jasper-growroot is mounting the filesystem before resizing. Sep 21 01:54:20 It would be nice if these derived variables were also in the log. Not a single log entry has the variable output. Sep 21 01:55:25 GrueMaster: It mounts and the umounts, doesn't it? Sep 21 01:56:38 Yes, but who knows if the umount actually happens before it tries to check the drive. Sep 21 01:56:49 flash-kernel has timing issues with this. Sep 21 02:02:21 I wonder if this is begging for an icky sync solution and we're just racing... Sep 21 02:02:38 probably. Sep 21 02:03:17 * GrueMaster is approaching the 12 hour mark w/o glasses. Everything is getting real fuzzy. Sep 21 02:04:00 Oh, ffs, that image doesn't fit on a 1G card. Sep 21 02:04:08 * infinity tries his sketchy 16G card that fails half the time. Sep 21 02:05:44 This script makes me sad... Sep 21 02:05:55 How was $MYECHO deemed less typing than "echo "? Sep 21 02:06:13 MYECHO feeds plymouth Sep 21 02:06:35 Oh, I missed the if plymouth bit. Sep 21 02:08:12 what is this frigging initrd format this week? I have stripped the uboot crap off, but non of my unpack scripts work. Sep 21 02:09:13 Same as it's been for the last several years... Sep 21 02:09:39 A cpio archive, compressed with the compression flavour of the month. :P Sep 21 02:10:01 gzip still, looks like. Sep 21 02:10:31 gzip: initrd.img: not in gzip format Sep 21 02:10:31 lzcat: Decoder error Sep 21 02:10:37 Those are the two I work with the most. Sep 21 02:14:19 xz? (from mkinitramfs) Sep 21 02:14:40 Really? Mine claim to be gzipped. Sep 21 02:15:09 I'm talking about the one on the boot image. Sep 21 02:15:32 Stripped first 64 bytes, nothing. Sep 21 02:17:33 Ahh, lzma. Sep 21 02:17:42 Well, the ones from live-build are. Sep 21 02:17:54 Of course, your lzcat still says no to that. Sep 21 02:18:06 ARRRGGGG Sep 21 02:18:18 file just comes back as data. Sep 21 02:20:04 I'm leaving the jasper debug to you. My head really hurts. No glasses doesn't help. Sep 21 02:28:16 I hate flash media. So. Much. Sep 21 02:30:50 Wait... Sep 21 02:31:19 bad geometry: block count 897024 exceeds size of device (897008 blocks) Sep 21 02:31:24 That seems like a bad thing. Sep 21 02:33:25 How are we 16 off? Sep 21 02:34:39 nand flash itself isn't that bad, it's just the dancing that you have to go through when there's no FTL Sep 21 02:35:08 twb`: nand is pretty bad, I have a ton of dead and dying cards. Sep 21 02:35:26 Shrug Sep 21 02:35:36 I could say the same about HDD or optical media Sep 21 02:35:51 Hard drives die a tad more slowly. :P Sep 21 02:35:57 (Or, rather, after a lot more use) Sep 21 02:36:13 As a write-once (or write-infrequently) medium, nand is great. Sep 21 02:38:00 Meh. Sep 21 02:38:27 That's partly just that HDDs are more mature technology Sep 21 02:39:11 As long as I get 3-5 years of life out of a unit, I'm happy. Sep 21 02:44:16 GrueMaster: I'm now wondering if this is a bug we've had for ages, and we only trigger it sometimes... Sep 21 02:44:39 GrueMaster: I think we're occasionally creating an image that's one block short. Sep 21 02:45:02 What, sync? I hit it quite often. Just kept getting told it is only affecting me. Sep 21 02:45:19 That bug too. Sep 21 02:45:56 The 1 block short thing causes the mount and umount to fail. Sep 21 02:46:05 Cause it's a bad FS. Sep 21 02:46:09 But resizing works. Sep 21 02:46:58 And reflashing over the same SD will occasionally work fine. Sep 21 02:48:07 * GrueMaster sometimes wishes he could have a developer stand next to him and sucker-punch him for every bug that flies by. Sep 21 02:48:39 That sounds like a job for an automaton, not a developer :P Sep 21 02:48:44 You have a live-in one, don't you? Sep 21 02:49:06 Only when he's awake. Sep 21 02:49:36 And he rarely enters my office anymore. guess he is learning. :P **** ENDING LOGGING AT Wed Sep 21 02:59:57 2011