/srv/irclogs.ubuntu.com/2012/11/12/#ubuntu-arm.txt

=== Ursinha-afk is now known as Ursinha
ptland here I am!! \o/04:12
ptlhey guys04:13
ptland gals04:13
ptlanyone been using Ubuntu on CuBox?04:13
ptlok... what about on raspberry pi?04:13
ptlUG802 maybe?04:14
ptlok, so what about Nexus7? anyone?04:14
ngharon7.  works good04:15
ptlgotta get mine in one week tho]04:15
ptlso far I've only ubuntu running on CuBox04:16
ptlUG802 will get by here any day04:16
lilstevieptl, from the topic "If you have a Pi, try #raspbian ! "04:16
ngharoworkin on minimal ubuntu build process for n704:16
ptllilstevie: I was just trying to start a thread. I don't have a raspberry pi...04:16
ngharolive-build rox, just a lot of docs to read04:17
lilstevie"start a thread"?04:17
ptlisn't the build process already in place at canonical?04:17
ngharoi couldnt find configs or anything re: build process other than how to create the images04:17
ptlah...04:18
ngharowhich is fairly trivial04:18
ptlwill you make it public? homepage and such04:18
lilstevieptl, canonical does not support the cpu architecture that the raspberry pi uses04:18
ngharohttps://github.com/ngharo/ubuntu-nexus7-minimal-livebuild here's the repo.  I'll be fleshing it out more on tuesday04:19
ptllilstevie: isn't the kernel unifying the ARM architectures right now?04:19
ngharothis will dump out a tarball, which you'd need to massage a bit for n704:19
* ngharo is going to tweak makefile to do the massaging soon04:19
ptlngharo: thanks. It will not have any use to me now, until I get my hands on the Nexus7. Do you think it might work in other devices?04:19
ptllike, not tablets04:20
lilstevieptl, that doesn't matter, even once the kernel unification is done, the raspberry pi is a different cpu architecture to what canonical support04:20
ngharoyeah probably, it should work on any armhf arch device04:20
ptlhumm... do you know if canonical supports the Marvell Armada?04:20
ngharojust would need to remove the n7 packages from the config04:20
ptlthat's what my cubox uses04:20
lilstevieptl, ubuntu-arm is built on armv7+ with hardfloat support. the raspberry pi is armv604:20
ptlarmv7l04:21
ptlah04:21
ptlI never got the gist of these ARM types and variants04:21
ptlI'll wait for my second CuBox and maybe make it a highly available cluster using drbd04:22
lilstevieptl, the difference is the evolution of the processor, they work differently, mostly they are backwards compatible, but older versions cannot run stuff for newer versions due to lacking support for the instructions04:22
ptllilstevie: that's why I did not buy a Pi! :) CuBox, yeah. I thought about buying a D2plug but it was too expensive04:24
lilsteviethere are many reasons not to buy a pi04:25
lilsteviearmv6 is just one of them04:25
ptldoes the ubuntu on nexus 7 supports multi-touch?04:26
lilstevieafaik it is limited04:27
lilstevieon the tf201 I get multitouch, but how it can be used is still fairly limited04:28
ptlwhat is a tf201?04:28
ptlah04:28
ptltransformer prime04:28
ptldid you install ubuntu on it?04:28
lilstevieyes04:29
ptlhow it works compared to the one in nexus7? better? worse? or more or less of the same?04:40
lilstevieI don't have a nexus 704:41
ptlthought you had04:43
=== zz_chihchun is now known as chihchun
ptlngharo: you have a "configure" script but no config.ac / Makefile.am ?05:12
ptl*configure.ac05:13
dholbachgood morning07:47
hrwptl: configure.ac may also reside in separate subdir10:15
ogra_hrw, so that missing dri support seems to be quite an issue in the armsoc driver, any idea how to get around that ? (see the build log)10:19
hrwogra_: plan to work on it after vacations10:21
hrwlooks like configure(.ac) needs fixing10:21
hrwI was supposed to be on streets of Palamos enjoying last sun. instead sitting with laptop in a building ':D10:22
ogra_well, even if i force DRI=yes in there ist doesnt build any armsoc_dri.so (which it looks for according to Xorg.0.log)10:22
ogra_oh. get out then !10:22
hrwogra_: first have to reply to interesting email10:23
hrwogra_: chromium os also lacks dri driver10:28
hrwogra_: can you check is there dri driver for omap boards?10:28
ogra_do you have the same message in your Xorg.0.log there ?10:28
ogra_there is a dri2 one but that doesnt land in the search path the driver uses10:28
ogra_also did you note that there is a link in the source to ../xf68-driver-omap or so ?10:29
hrwarmsoc is a continuation/fork of omap/armsoc one10:30
ogra_yep, i know10:30
ogra_i was just wondering if anything from that link is used during build10:30
brendandanyone else notice that network-manager-applet doesn't work properly on the first boot of the nexus 7?11:08
ogra-cb_brendand, yep, we need a rebot after setting the hostname, there is a bug open somewheer iirc11:10
ogra-cb_*reboot11:10
kulveis there a way to flash a bigger rootfs to nexus 7 with fastboot? Everything bigger than ~700M seems to just hang the fastboot (I guess it doesn't fit to memory or something)11:11
ogra-cb_kulve, there is the -S option buut it is slightly unreliable which is why we stopped using it11:15
kulveogra-cb_: -S option to what..?11:15
ogra-cb_ffastboot flash11:15
kulvehmm.. Is that documented somewhere? "fastboot --help" doesn't show it11:17
ogra-cb_it does here11:19
ogra-cb_http://paste.ubuntu.com/1352881/11:20
ogra-cb_see the last line11:20
kulveah, true. I was checking in a different Android env assuming that it would the the same..11:21
kulvedo you happen to know what that does in practice..? I don't understand what the word "sparse" means in that context..11:23
ogra-cb_it chops file it flashes into pieces of the size you define and re-assembles it during write11:27
ogra-cb_so you can define chunks that fit in ram11:27
kulveogra-cb_: thanks. Sounds exactly what I need (even if it wouldn't work 100%). I somehow thought that the option was for something else11:50
ogra-cb_marvin24, hmm, do you remember if we pulled zram separately into the ac100 tree ?12:14
ogra-cb_i'm looking at 3.4 mainline atm (for chromebook) and zram has a depends: X86 in Kconfig here12:15
marvin24ogra-cb_: no, I cherry picked it12:16
ogra-cb_ah12:16
ogra-cb_thx12:16
marvin24in 3.4, it was disabled for !x8612:16
marvin24which was fixed in 3.5 or 3.6 or so12:17
ogra-cb_right, just dropping the dep breaks it12:17
marvin24ogra-cb_: nivdia made a new kernel release12:18
marvin24this has nothing to do with the video drivers12:18
marvin24just a kernel service release12:18
ogra-cb_yeap, i know but i assume the drivers will follow12:18
ogra-cb_since they are currentlly heavily broken12:19
xnoxogra_: ogra-cb: looking at bug 771736 what exactly does it need in the initramfs?13:14
ubot2Launchpad bug 771736 in nilfs-tools (Ubuntu Oneiric) "please include nilfs2 in hook-functions" [Medium,Confirmed] https://launchpad.net/bugs/77173613:14
xnoxthe gc daemon?13:14
ogra_yeah13:14
ogra_there was an initramfs snippet already somewhere13:14
ogra_also the module needs to be pulled into the initrd13:15
ogra_i dont think we have that yet13:15
xnoxI see the snippet to pull module, don't see one for any daemons....13:16
ogra_xnox, i can imagine that cleanerd cant be restarted safely once it runs, so it is omitted for the initrd13:34
ogra_(indeed stateful re-exec in upstart should fix that)13:35
xnoxok. let me try stuff then.13:38
ogra_stuff is good :)13:39
* ogra_ tries more coffee-stuff 13:39
lilstevielol13:39
* xnox remembers about the cheese and toast stuff13:40
ogra_victorp, !14:01
victorpogra_, !?14:07
=== dmart_ is now known as dmart
=== karim is now known as Guest78743
brendandheh, with that 'Disable "glx"' tweak the shaded keyboard comes up fairly quickly15:28
ogra_well, it eats one of your CPUs then though15:28
ogra_software rendering doesnt come for free sadly :)15:29
brendandogra_, oh right. it *is* only sw rendering15:30
brendandogra_, what needs to happen for hw rendering?15:31
ogra_brendand, i assume onboard needs to use gles instead of gl functions15:31
ogra_havent looked much at the code yet15:31
brendandogra_ - oh it's onboards problem specifically?15:32
ogra_well, its a prob of app using plain GL instead of the GLES subset15:32
ogra_if you use any GL functions that arent in the GLES subset it will try to use SW renerind15:33
diablotinHello15:35
diablotinSomeone in there ?15:36
ogra-cbno, just 165 people idling here15:36
ogra-cb:)15:36
diablotinGreat !15:36
diablotinI'm looking for information about Ubuntu for Android15:37
diablotinI've bought the PadFone from Asus and I would love to have ubuntu running on it when docked !15:38
tassadarogra-cb: you've said that I really should not flash different boot image on my nexus7 every time I wanna switch android<->ubuntu, because of the mmc load15:38
tassadarbut the /data partition must get several times higher io load, no? Is the boot partition somehow different?15:38
=== NekoXP is now known as Neko
=== Neko is now known as NekoXP
ogra-cbtassadar, flashing means you always write to the same place (i.e. dd), there is no filesystem layer15:40
ogra-cbwhile using a rootfs you have an additional layer that cares for wear levelling15:40
ogra-cbdiablotin, i dont think there are any public versions of U4A yet, its still being worked on15:41
diablotinogra-cb: even an alpha ? beta ?15:41
ogra-cbnothing public i know of15:42
diablotinBut is U4A gonna be downloadable or we'll have to buy it with the phone ?15:42
diablotinOk15:42
tassadarhmm, right, thanks15:43
tassadardatasheet of kingston emmc memory chips (the ones used in nexus 7) says that these chips have static wear-leveling15:53
ogra-cball MMCs have that15:54
ogra-cbrobclark, do you happen to know anything about the armsoc driver ?15:55
* ogra-cb wonders what to do to make it not complain about dri.h not being compilable15:55
tassadarand "the internet" tells me that eMMC is NAND chip with controller, which handles the wear leveling and bad block management, among other things15:56
tassadardoes dd bypass that?15:56
ogra-cbno, its HW, nothing you can bypass15:56
robclarkogra-cb, well, armsoc was branched off the omap driver.. I think hrw made some patches to fwd port it to latest xorg..15:57
ogra-cbright15:58
ogra-cband ir works quite well hwen copzing the GLES/EGL and mali libs15:58
tassadarso it is not using the same cell over and over again, and each cell should endure 1k-3k overwrites15:58
ogra-cbbut it still complains about dri.h during build and i.e. unity doesnt seem to be accelerated15:59
tassadarthat gives me the impression it is not that bad15:59
ogra-cbit will shorten the lifetime of the MMC nontheless and its really dangerous, i wouldnt use flashing if i dont have to15:59
ogra-cbif you crash in the middle your device turns into a brick and you have to completely reinstall16:00
tassadaryeah, I'd like to use the kexec, but I can't get it working by myself - it is simply too low-level16:00
tassadarI tried applying the kexec-hardboot patches, but I've got stuck on reseting of the CPU via watchdog16:01
tassadarthe patches were made for different CPU, and well, there is no datasheet for tegra 3 available16:02
ogra-cbrobclark, https://launchpadlibrarian.net/122595562/buildlog_ubuntu-raring-armhf.xf86-video-armsoc_0.4.0-0ubuntu4_BUILDING.txt.gz see the dri.h complaint ? i can force DRI=on but dont get different results in behavior of the binary16:02
janimoogra_, so what is the best method to put up a notice at package install and have it ignored at image build time? I remember debconf notice (or note?) being discouraged16:06
ogra-cbwell, note is discouraged, but probably the best you can use16:07
ogra-cbhave a look at the sun java package16:07
robclarkogra-cb, hmm, /usr/include/xorg/dri.h looks like dri1 stuff.. I suspect armsoc shouldn't be looking for that16:08
ogra-cbjanimo, during all aem image builds we export FLASH_KERNEL_SKIP=true ... so only show the note/message/whatever if thats not set16:09
robclarkhmm, that said, xf86-video-omap does too..  I'm not entirely sure why you see that issue but I don't.. maybe try building xf86-video-omap the same way?16:09
ogra-cbrobclark, ah, awesome, so i can just rip that out16:09
ogra-cbthst the info i was missing :)16:09
robclarkI think so..  at least that is what I'd try first16:09
ogra-cbwell, it already works fine for es2gears or glmark ... its just unity thats very slow16:11
ogra-cbthough hrw doesnt use the alternatives system, might be that the current driver ends up with a mishmash of mesa and mali16:11
hrwhi16:11
robclarkyeah, sounds a bit like it16:12
hrwogra-cb: I have crhomium-opengles package which handle egl for mesa stuff16:12
ogra-cbhrw, oh !16:12
hrwbut it requires binaries16:12
hrwand I do not think that libmali* has license for re-distribution16:12
ogra-cbthe loibs we manually copy i guess16:12
ogra-cbyeah16:13
* ogra-cb was fearing that16:13
hrwogra-cb: http://tygrysek.juszkiewicz.com.pl/~hrw/ubuntu/chromebook/16:14
hrwogra-cb: fetch tar.xz, add libmali and build16:14
* ogra-cb will try that later today16:15
hrwogra-cb: cgpt, kernel signing and kernel is what left on my list16:16
hrwI gave up on xf86-input-cmt - too many deps16:16
ogra-cband really no need for it16:16
hrwyep16:17
ogra-cbi have my touchpad working exactly as on cros here16:17
ogra-cbit feels a bit delayed in reaction, but i blame X here or unity16:17
ogra-cbbut two finger scrolling, two three and four finger taps etc, all work16:18
ogra-cbjanimo, bug 1068182 was just pointed out to me16:26
ubot2Launchpad bug 1068182 in linux-ac100 (Ubuntu) "Regression: several kernel modules now missing in 12.10 3.1.0-6-ac100 kernel compared to 12.04 3.0.27-1 kernel" [Undecided,Confirmed] https://launchpad.net/bugs/106818216:26
hrwchecking whether to include DRI support... checking for /usr/include/xorg/dri.h... yes16:46
hrwchecking for /usr/include/xorg/dristruct.h... yes16:46
hrwchecking whether to include DRI support...16:46
ogra-cbnice16:49
hrwautoconf suxx16:50
ogra-cbheh, ++16:50
hrwand headers which no longer include required headers suxx too16:51
hrwdri.h requires 1, dristruct.h 416:51
hrwhttp://paste.ubuntu.com/1353543/16:54
hrwthats whole configure.ac patch16:54
kulvehow do I create a sparse file system image suitable for fastboot flash -S?16:59
ogra-cbkulve, you use make_ext4fs17:02
hrwbye17:03
ogra-cbkulve, from android-tools-fsutils (in raring)17:03
kulveah, thanks. Was just about to ask17:03
xnoxkulve: it's also available in a ppa for quantal and precise17:03
xnoxin the install ppa owned by ~ubuntu-nexus717:04
ogra-cbah, right, forgot the ppa17:04
xnox;)17:04
* xnox blogged about it on ubuntu-planet surely everyone saw it =)17:04
* ogra-cb did, but forgot about the ppa nontheless :)17:05
kulveweird. I have debian stable, so I took the package from debian unstable and recompiled. But for some reason that doesn't have -fsutils, even though the ubuntu version with the exact same version number does..17:10
ogra-cbthe ubuntu package has an ubuntu1 version17:12
ogra-cbthat means it carries one change change thats not in debian17:12
ogra-cbguess which one :)17:12
ogra-cb-change17:13
kulveno man page.. Is -l for the whole rootfs size? Like 13GB or something like that?17:14
ogra-cbyeah17:14
ogra-cbypu want -l and -s iirc17:14
xnoxkulve: I did forward the debdiff to debian bug 69285117:16
ubot2Debian bug 692851 in android-tools "android-tools: please package the rest of the fs tools" [Normal,Open] http://bugs.debian.org/69285117:16
xnoxyou can recompile with that patch.....17:16
kulveI already compiled the ubuntu version. I guess  debian wants also man pages always? Maybe that's why make_ext4fs was removed from the packaging?17:18
ogra-cbfastboot doesnt have a manpage either17:18
kulveso it seems17:19
ogra-cbwhile thats a bug i dont think it is a blocker anzmore17:19
xnoxkulve: the point is there is no upstream build system, and makefiles were written for "interesting" tools at the time (fastboot). Now that it's possible to fasboot flash ubuntu/debian, the fsutils became "interesting" so I wrote a few more makefiles to compile them.17:21
xnoxkulve: so just ongoing evolution of packaging bits of android tools git forest.17:21
kulveI didn't even know that there are such things outside android until yesterday17:21
kulvehmm.. fastboot taking 100% and 600MB of memory (and increasing)17:23
ogra-cbare you sure your device is in flash mode and properly connected and detectet ?17:23
kulveat leaste "fastboot erase userdata" worked..17:24
kulvestill more questions. What should the last parameter "directory" actually contain? At least is doesn't seem to like having a /dev in there..17:28
prpplaguehttp://newscenter.ti.com/2012-11-12-New-OMAP-5-processor-based-modules-spur-easier-faster-industrial-designs17:29
kulveI figured out why the fastboot went into a leaking busyloop (or something else weird). I forgot the "M" after "-S 256"..17:33
ogra-cbgreat, now seem that on miniITX boards please so we finally build arm desktop PCs17:33
ogra-cbs/seem/sell/17:33
xnoxCalling hook nilfs17:59
xnoxAdding module /lib/modules/3.5.0-17-generic/kernel/fs/nilfs2/nilfs2.ko17:59
xnoxare the raring panda images working?17:59
xnoxI guess I need to add partman-nilfs to complete the circle.17:59
xnoxthen the next question will be "will it fit" in the panda/ac100/nexus initramfs.18:00
xnoxand whether it is built....18:00
ogra-cbxnox, panda has no restrictions18:20
ogra-cbonly fastboot based bootloaders do18:20
ogra-cband iirc it did fit into ac100 last time i tried18:21
ogra-cbnexus shouldnt differ much18:21
xnoxogra-cb: good.18:21
xnoxare the raring panda images working? Will try to do partman-nilfs module for the installer.18:21
ogra-cbi havent tried one yet18:22
xnoxok =) I will tomorrow.18:22
ogra-cbi usually started that with A1 preparation18:22
ogra-cbi guess i have to make up some personal schedule now :)18:22
xnoxlol. well x86 desktop are borked because of secure-boot, but colin fixed it, so tomorrow should be better =)18:22
xnox"opt-in milestones"18:22
ogra-cbheh, yeah18:23
xnoxtechnically this is second week since UDS, such that we should have a fully working set by thursday.....18:23
ogra-cbwell, i doubt i can do it wrt nexus until thu18:23
ogra-cbstill no kernel18:23
xnoxwhich is interesting since kernels don't need much review to go through.18:26
xnoxand we just had ppc kernel whizz through....18:27
kulveAny hints why my nexus fails to mount the fs that create with "sudo make_ext4fs -s -l 4G rootfs.img build/" and flash with sudo fastboot flash -S 256M userdata rootfs.img"?18:28
ogra-cbxnox, well, the nexus kernel sits in the queue longer, i watched the ppc one go bz whiule reloading the queue page over the weekend18:29
kulvekernel just tells that it failed to mount it as ext2/3/4/etc, shows the list of the partitions and reboots after 10sec18:29
ogra-cb*by18:29
ogra-cbkulve, sounds like your initrd isnt used18:30
* ogra-cb needs to try something ... brb18:30
kulveogra-cb: yeah, I'm trying to manage without it. It did manage to create "normal" ext4 and mount that, but it had the space limitation18:31
ogra-cbhrw, wow, that touchpad snippet helps a lot !!18:33
hrwogra-cb: xorg.conf one?18:34
hrwogra-cb: thanks should go to ojn not me18:34
ogra-cbit makes it so much more sensitive18:35
ogra-cb[    24.541] (EE) AIGLX error: dlopen of /usr/lib/arm-linux-gnueabihf/dri/armsoc_dri.so failed (/usr/lib/arm-linux-gnueabihf/dri/armsoc_dri.so: cannot open shared object file: No such file or directory)18:41
ogra-cb[    24.541] (EE) AIGLX: reverting to software rendering18:41
ogra-cbhrw, from your Xorg log18:42
ogra-cbso i guess ripping out dri1 completely is actually needed18:42
janimoogra_, can you give me the exact config diff you use to talk to cbc_multi? When I switched the android out and also made things modular the kernel did not build - the same cable_somthing error you saw too19:04
hrwogra_: fun is that it looks that dri1 is not even used in code - only in configure19:40
ogra-cbogra@chromebook:~$ /usr/lib/nux/unity_support_test -p|grep supported19:52
ogra-cbUnity supported:          yes19:52
ogra-cbogra@chromebook:~$ grep rendering .xsession-errors19:52
ogra-cbcompiz (core) - Info: Unity is not supported by your hardware. Enabling software rendering instead (slow).19:52
ogra-cbi dont get that19:52
ogra-cbit works fine on the nexus19:52
rsalvetiogra-cb: what is the return core from /usr/lib/nux/unity_support_test ?20:30
rsalvetihm, but by the message the limitation is inside compiz20:30
rsalvetilack of extension?20:30
rsalvetiremember that this is a different driver :-)20:30
rsalvetiso different kind of fun20:30
opus_hey guys, any support for the AMLOGIC 8726M?22:54

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!