/srv/irclogs.ubuntu.com/2013/04/04/#ubuntu-arm.txt

=== calculu5 is now known as calculus
=== calculu5 is now known as calculus
=== ynezz_ is now known as ynezz
kvarleyThe Google Samsung Chromebook has the same ARM chip in as the Nexus 10 - does that mean that I should technically be able to run the ARM versions of Ubuntu on the tablet?08:47
Tm_Tkvarley: yes, doesn't mean it would have all the pheripherals functioning correctly though08:51
Tm_Tpheripherals including but not limited to: display drivers (;08:52
kvarleyTm_T: Hmm, true. Could I boot the 13.04 image with the android kernel though?08:52
kvarleyTm_T: A patched driver has hit the 13.04 repos08:52
Tm_Tkvarley: I suppose it would be worth trying, but I have no knowledge about the said chromebook08:53
kvarleyOr so Phoronix claims http://www.phoronix.com/scan.php?page=news_item&px=MTI5Mzg "The Ubuntu 13.04 repository just received ARM's new universal X.Org graphics driver,  a.k.a. the xf86-video-armsoc DDX. This is a generic ARM SoC DDX driver  for the X.Org Server originally derived from the xf86-video-omap driver."08:53
kvarleyTm_T: I may have to read up on this. I think the stumbling block is having to include the video driver in the rootfs because I'll likely have no wifi access on the device08:54
kvarleyHmm, the SoC is the same but the GPU is different08:55
bmwHello. Can you help me remap arrow keys + Alt to End,PgUp etc... I think xmodmap can't help.08:58
bmwOn Chromebook (:09:02
=== doko_ is now known as doko
hrwkvarley: 13.04 has most of things you need to get ubuntu running on chromebook09:33
kvarleyhrw: Out of the box?09:38
hrwkvarley: define 'out of box'09:38
kvarleyhrw: fresh install09:38
hrwdefine 'fresh install'09:38
kvarleyhrw: lol, as in the images published for the chromebook 13.0409:38
hrwthere will be no images for chromebook09:38
ogra_no, there are no images09:38
ogra_i wouldnt say "will be" ...09:39
ogra_but currently there arent any09:39
hrwno one wanted to join in development of chromebook support and I am too lazy/occupied to work on it09:39
ogra_i use my CB to much, else i would probably invest some time to make images09:39
ogra_but creating the first set simply means a lot of reinstalls09:40
kvarleyhrw: Ok, if I took this image raring-preinstalled-desktop-armhf+ac100.tar.gz would it at least boot on the Nexus 10 if I booted with the Android kernel?09:40
hrwkvarley: nexus 10 != arm chromebook09:40
kvarleyhrw: I know, but they share the same SoC so the configuration would likely be similar09:41
hrwkvarley: ac100 share the same SoC as "LG 4xHD" cellphone - would ubuntu run on a phone?09:41
hrwkvarley: for nexus10 you may ask ubuntu touch guys do they have some image09:42
kvarleyhrw: ok, thanks09:43
ogra_for nexus 4,7,10 and galaxy nexus there are touch images09:43
hrwkvarley: if you use ac100 image and boot it with default chromeos kernel it will start09:43
hrwno x11 probably but will start09:44
ogra_it might even fall back to xfbdev ... who knows :)09:44
hrwanyway that reminds me that 3.4.0-7 is waiting for be built/uploaded09:49
hrwand I finally got microsd->sd adapters which do not stick out from chromebook09:51
hrwogra_: http://dx.com/p/2-in-1-tf-card-to-sd-card-adapter-usb-card-reader-white-12629909:52
hrwogra_: cut usb part and it will not stick out09:53
* ogra_ wonders why the CSS is so messed up 09:54
ogra_its unreadable here ... layers of text over layers of text09:54
hrwgit-buildpackage + chromium-kernel git repo == problems ;d10:08
hrwgood thing is that aptitude works now on armhf10:13
ogra_pfft, aptitude ...10:15
* ogra_ doesnt know anyone using it 10:15
ogra_pretty pointless to have two package DBs installed just for some commandline UI10:16
hrwogra_: flash-kernel still does not have DT support?10:16
ogra_i didnt add any ... and nobody sent patches10:17
ogra_ask ppisati :)10:17
hrwogra_: show me other useful APT UI for console and I may switch10:17
ogra_aptitude is *not* an apt UI ... it has nothihng to do with apt at all10:17
hrwI use it as such ok?10:18
ogra_it is its own package tool ... simply duplicating what apt does10:18
ogra_just wastes extra disk space with an additional DB10:18
* ogra_ stays with apt :)10:19
hrwI went dselect -> console-apt -> aptitude10:20
ogra_well, except the gui part aptitude today has no features that apt doesnt provide also10:20
ogra_(it used to in the past)10:21
hrwsure, but I use that UI part10:21
ogra_yeah10:21
ogra_i have better usage for teh several MB that wastes on my disk :)10:22
hrw;D10:22
hrwhm. hacking flash-kernel does not have sense10:29
hrwit can guess where rootfs is from /proc/cmdline but can not where kernel is stored10:29
ogra_in /boot indeed :)10:30
ogra_along with the dtb10:30
hrwogra_: will you package u-boot for chainloading?10:30
ogra_i pull whatever linaro packages ... as usual10:30
hrwLinaro does not do any chromebook stuff10:31
ogra_if i even have to pull ... i think jcrigby can just upload nowadays10:31
ogra_linatro has the mainline branch as base for their packages ... so you could ask for a binary that goes into ubuntu i guess10:31
hrwI have no idea which ver of u-boot works for chainloading10:32
* ogra_ neither10:33
RjsI use aptitude all the time, largely because of the UI - it is much easier to select packages from multiple choices or look at suggests and recommends manually using an interactive UI than trying to add things to the apt-get command line iteratively10:33
Rjsthough mainly on debian, and especially on minimal single-use embedded systems (instead of a desktop), so I guess the situation may look different from an Ubuntu desktop perspective...10:33
hrwRjs: I doubt that ubuntu desktop comes with something similarly usable10:34
ogra_well, especially on embedded i wouldnt use aptitude ... unless you have a lot of diskspace10:36
hrwyep10:37
hrwthere are other tools to get rid of useless stuff10:37
hrwdebfoster for example10:37
Rjsogra_: hmm, on embedded I've found that aptitude is the best way (for me) to keep the number of installed packages minimal and still have all features I need (I use it with auto-install recommends off and select manually perhaps around 70% of the recommends)10:37
hrwRjs: what is your 'embedded' hardware?10:38
ogra_i'm not talking about comfort :)10:38
ogra_aptitude adds its own package database10:38
ogra_in parallel to the apt one10:38
Rjshrw: mostly alix's embedded geode pc's (i.e. x86) and a few openmoko gta02's10:38
hrwRjs: alix was nice hw. I used 1c for few years10:39
ogra_ogra@chromebook:~$ du -hcs /var/lib/apt/lists/10:39
ogra_95M/var/lib/apt/lists/10:39
Rjsif plain apt had an interactive UI with the features of aptitude, I'd probably switch to it, but I haven't found anything comparable...10:39
hrwhrw@krolik:/var/lib$ du -hs aptitude/ apt10:39
hrw7,8M    aptitude/10:39
hrw84M     apt10:39
ogra_it wont actually double that up but will definitely add a lot on top10:39
hrw-rw-r--r--  1 root root 4046282 kwi  4 12:14 pkgstates10:40
hrw-rw-r--r--  1 root root 4046326 kwi  4 12:13 pkgstates.old10:40
hrwthat's all?10:40
ogra_dunno, as i said, never using aptitude10:40
RjsI've found on my system that keeping auto-install recommends on (which I guess would be the only practical choice with a non-interactive UI) would add much more than that amount of diskspace, plus lots of services and background processes that I have no use for10:41
ogra_so switch it off :)10:41
Rjsbut it's probably heavily dependent on the user preferences and what the systems are for (e.g., I have several desktop systems without gvfs or udisks or network-manager or all that complexity, but still a somewhat minimal and very usable X11 system for my purposes)10:42
ogra_pfft, X11 .... Mir is the future10:42
ogra_*g*10:42
ogra_at least on arm10:43
Rjsbut not the present :)10:43
ogra_well, i heard it works on nexus 4 and 7 already10:43
ogra_the advantage is that it doesnt need any special drivers ... you can just use the android blobs10:44
hrwogra_: http://paste.ubuntu.com/5676209/ - everything needed for device tree support?10:44
ogra_hrw, dunno, as i said, talk to ppisati10:45
hrwppisati: ping  ^^10:45
* ogra_ has no clue at all how dtb stuff is implemented in our kernels10:46
hrwogra_: s/dtb/DT/10:46
ogra_and given that nearly all future ubuntu arm development will be android kernel based ....10:47
ogra_(except for  server i guess)10:47
hrwogra_: I heard rumours that s/eglibc/bionic/ is also planned10:48
ogra_hrw, nope10:49
ogra_that would have been an april fool :)10:49
hrwogra_: there was such? :D10:50
ogra_we do make heavy use of libhybris to bridge between the two10:50
hrwI was mostly travelling on 1st April10:50
hrwSaarbr├╝cken -> FRA -> TXL -> home10:50
ogra_yeah10:51
ogra_i read g+ :)10:51
ppisatihrw: dtb support for flash-kernel i suppose?10:53
hrwppisati: yes10:53
hrwppisati: just other way to identify machine10:53
ppisatihrw: iirc we had another patch for dtb support from... marvin?10:54
ppisatimarvin24: ^10:54
ogra_it was similar iirc10:55
hrw19:24 < marvin24> hi, I'm about to add http://paste.ubuntu.com/5602702/ to our version of flash-kernel10:55
* ogra_ remebers it at least using the same path in /proc10:55
hrwthat's not that10:55
hrwhttps://launchpadlibrarian.net/133791292/flash-kernel_3.0~rc.4ubuntu29_3.0~rc.4ubuntu29-tegra.diff.gz10:56
hrwalso not that10:56
hrwhe just added support for BT based Tegra board using FDT entry from /proc/cpuinfo10:56
hrw~hail /lastlog marvin10:57
hrwech, 3.4.0-6 has broken keyboard11:38
marvin24hrw: what is BT based?12:13
hrwmarvin24: DT? Google Snow (aka Samsung ARM Chromebook)12:13
marvin24ppisati, hrw: my goal is to find a way to tell uboot which DT to load until we have full DT support in uboot12:13
marvin24hrw: no, this is for ac100, but all tegra2/3/4 boards should be supported12:14
marvin24btw, does the flash-kernel db support wildcards?12:14
marvin24e.g. Machine: nVidia Tegra* (Flattened Device Tree)12:14
hrwmarvin24: your change used name from /proc/cpuinfo while mine adds reading of model from DT12:15
marvin24hrw: no, I use /proc/device-tree/model as you12:15
marvin24you posted the link12:15
marvin24https://launchpadlibrarian.net/133791292/flash-kernel_3.0~rc.4ubuntu29_3.0~rc.4ubuntu29-tegra.diff.gz12:16
marvin24this is used to copy the device tree to /boot where uboot can find it12:16
hrwmarvin24: you use /proc/device-tree/model to copy DTB but not to identify device in flash-kernel12:16
marvin24hrw: sorry, but why is this necessary?12:16
marvin24the kernel supports multi-soc12:17
hrwmarvin24: Chromebook returns "SAMSUNG EXYNOS5 (Flattened Device Tree)" which covers also Andale which can not boot Chromebook kernel12:17
hrwmarvin24: I have only 3.4 kernel for Chromebook12:17
marvin24looks more like missing Andale kernel support for me - right?12:18
hrwmarvin24: so my goal is to use DT model when cpuinfo reports DT device. this way same SoC devices may run other flavours12:18
hrwmarvin24: I want to support *one* device as I do that in my free time12:18
marvin24I also check for "grep -q 'Flattened Device Tree' /proc/cpuinfo " fiest12:18
hrwwhen Linaro kernel teams will get Andale into working condition then they will have 3.9-rc etc12:18
marvin24*first12:18
hrwmarvin24: moment...12:19
marvin24hrw: are you also planing to copy the device tree to /boot ?12:19
marvin24or just use the device tree from uboot12:19
marvin24I don't know how far exynos support has grown12:20
hrwwhen user runs 'flash-kernel' script it checks /proc/cpuinfo for Hardware line and then do whatever it has to. For your Tegra work it is enough as you have one kernel for all devices + set of DTB files. So flash-kernel do what it has to and then kernel postinst runs your script to put proper DTB in /boot/ according to /proc/device-tree/model file. In my case on Exynos5 we do not have that luxury and different devices (so far) use different kernels so I need to hav12:21
hrwand Chrombook is using signed kernels...12:22
hrwmarvin24: but I see how my patch may make your work harder12:24
marvin24hrw: thanks for the explaination12:27
hrwhttp://paste.ubuntu.com/5676446/ may be better as allows to cover your usecase (one database entry for whole DT platform) and mine (many database entries per DT platform)12:27
hrwwill require some comments in code to describe it12:28
marvin24hrw: I don't see why your original patch should cause problems12:29
hrwmarvin24: your entry from https://launchpadlibrarian.net/133791292/flash-kernel_3.0~rc.4ubuntu29_3.0~rc.4ubuntu29-tegra.diff.gz will be skipped as DT model name would be used12:30
hrwmarvin24: http://paste.ubuntu.com/5676446/ checks /proc/cpuinfo, gets "NVidia Tegra DT" identifier. Checks database - o, we support it already - go.12:30
hrwmarvin24: when there is no "This platform DT" in database it will read DT model name and search for it.12:31
marvin24ah, machine is something like Toshiba AC100 again instead of common "Tegra20 ..."12:31
hrwyes12:31
marvin24hrw: ok, so your latter patch should match the common device tree early12:32
hrwyes, if it is in flash-kernel database12:32
marvin24now I got it - thanks!12:33
hrwmarvin24, ppisati: http://paste.ubuntu.com/5676460/ - is this comment readable enough?12:34
marvin24hrw: looks good to me12:36
hrwcool12:36
marvin24now how can I get the db entry into offical flash-kernel?12:37
hrwmarvin24: report a bug12:37
hrwwith debdiff12:37
marvin24ok, thanks again12:38
hrwmarvin24: so on AC100 what is in /proc/device-tree/model and what is a name of dtb?12:38
* marvin24 has to boot up first12:39
hrwcause on Chromebook it is 'Google Snow' + exynos5250-snow.dtb so your hook for copying DTB may not work12:39
hrwworks12:40
marvin24hrw: model is "Toshiba AC100 / Dynabook AZ"12:46
hrwthx12:46
marvin24so this would match the old fastboot entry in the db12:46
marvin24while /proc/cpuinfo gives "nVidia Tegra20 (F..D..T)"12:46
* marvin24 needs to go to the car garage ...12:47
hrwBug #116448412:48
ubot2`Launchpad bug 1164484 in flash-kernel (Ubuntu) "Add support for checking Device Tree model name" [Undecided,New] https://launchpad.net/bugs/116448412:48
ogra_hrw, i dont think that patch will fly, you completely ignore the device DB13:29
ogra_hrw, you rather want an entry for each supported DT device in the DB and tell it to use DT ...13:31
ogra_instead of ignoring *all* checks by just runing "if ! check_supported "$machine"; then"13:32
hrwogra_: I do not ignore DB13:34
ogra_check_supported checks the DB13:34
hrwyes13:34
ogra_you *want* that and a proper DB entry13:35
hrwogra_: so for all Tegra devices you want DB entry?13:35
ogra_imagine machines with DT and NAND ... or special bootloader setups13:35
ogra_for now you only need one for the ac100 ... andf it needs to switch between the old and new ways13:36
hrwok, I start to see13:36
hrwhangout now - be back13:36
ogra_since the distro doesnt use marvins DT kernel yet13:36
hrwre13:40
hrwogra_: if distro supports DT platform with one kernel - add DT entry. if with several ones - remove DT entry, add DTnames entries13:41
hrwor we can change check to: grab cpuinfo, check for DT, check for DTName = use if present, use DT if not13:42
hrwplease add comment to the bug13:42
ogra_hrw, yeah, second option sounds good13:43
marvin24can we also add wildcards to the machine detection? or is this too dangerous?13:59
marvin24currently, it seems we only detect exact matches: if [ $value = $machine ];  ...13:59
ogra_i dont think the code can handle that13:59
=== hggdh_ is now known as hggdh
marvin24something like [[ $value =~ $machine ]]14:01
marvin24(don't know if dash handles this)14:01
=== fabo_ is now known as fabo
marvin24arr, no regex in dash14:19
hrwdash lacks everything and more14:24
RjsI have no idea what you're talking about, but just in case this helps: all Bourne shells should support case which supports matching by glob patterns (not regexps)14:25
Rjse.g: case "$value" in foo-*) echo match ;; *) echo no match ;; esac14:25
hrwmarvin24: grep -e?14:25
marvin24hrw: maybe one should just replace dash by grep which has more features ;-)14:28
hrwogra, marvin24, ppisati: http://paste.ubuntu.com/5676804/ then14:32
Rjs"case" is I think the standard thing to use for wildcard matching in shell scripts, unless you specifically need regexps instead of glob patterns; grep is quite complicated to use in comparison (something like  if echo "$value" | grep regexp >/dev/null 2>&1; then  except if value begins with -, and I may have forgotten some other detail too)14:32
hrwRjs: if [ -n "`echo something|grep`" ]; then ;D14:33
hrwI have it in one of my scripts...14:34
ogra_lool, any opinion about hrw's patch ? ^^^14:34
Rjsthe case equivalent would be something like: case "$machine" in *device tree*) machine="$(get_devicetree_model)" ;; esac14:35
Rjs(though $(...) is also non-portable, should use `...` instead)14:36
Rjsthough if the script is complex enough to have user-defined functions, you could possibly consider just specifying /bin/bash and forgetting about portability :) (unless it's used somewhere where it's possible that only dash or e.g. busybox sh is available)14:38
marvin24hrw, ogra_: still fine I think14:39
marvin24I'll do some more tests in the next days14:40
marvin24maybe I can also solve the regex problem ;-)14:40
hrwBug #1164484 - v3 added14:44
ubot2`Launchpad bug 1164484 in flash-kernel (Ubuntu) "Add support for checking Device Tree model name" [Undecided,New] https://launchpad.net/bugs/116448414:44
loolhrw, ogra_: Might be best to call this device_tree_model instead of machine14:54
hrwlool: and add zillion of 'if/else/fi' in code everywhere?14:56
hrwlool: machine is also wrong named then - should be proc_cpuinfo_hardware14:57
loolhrw: machine is the historical name of it in the code and of the command-line override14:58
hrwlool: machine is also variable which tells which entry from deviceDB to use - right?14:59
loolYes; currently these are real Hardware names though14:59
hrwlool: it was Yes/No question14:59
hrwlool: so what my patch does is just expanding ways of checking for names14:59
hrwlool: otherwise we will go back into flash-kernel < 3 version where we had huge selects instead of database15:01
=== NekoXP is now known as Neko
loolhrw: It's just that I'd rather grep the db for model entries15:04
loolhrw: e.g. search for DeviceTree-Model: xyz in the db15:04
hrwgood point as well15:05
ogra_lool, yeah i think thats what i said above too15:06
=== W1N9Zr5355 is now known as W1N9Zr0
marvin24mmh, Xorg on raring crashes19:15
marvin24paste.ubuntu.com/567765319:16
marvin24I've seen something like this before19:16
marvin24when pci is missing19:16
marvin24indeed, pci probe19:22
ogra_marvin24, hmm20:02
ogra_segfaulting isnt nice though ... and it works  on todays ac100 image20:03
ogra_oh, indeed ... 3.8 ...20:05
tassadar_ogra_: do you have nexus7 with ubuntu desktop laying around? Would you please show me it's cmdline?20:39
ogra_i dont ...20:40
tassadar_okay, no problem20:41
tassadar_it's just that the prebuilt boot.img has console=tty0 in cmdline, and the installer then changes it to tty1, and I'm not sure which one is correct20:41
ogra_it is the cdmline the bootloader sets plus "root=/dev/mmcblk0p7 ro quiet splash"20:41
ogra_doesnt matter20:42
ogra_as long as its a tty20:42
ogra_the prob is that the bootloader puts console= in20:42
ogra_without any value20:42
ogra_that confuses a few things20:43
marvin24ogra_: I think we can fix it - just a moment20:45
* marvin24 compiles new xserver on his ac10020:45
ogra_marvin24, give it to tjaalton in #ubuntu-x20:51
ogra_(or mlankhorst if it takes until tomorrow)20:51
marvin24ogra_: ok, thanks - will try20:51
marvin24ogra_: no reply - I guess they will just point me to the bug tracker21:20
marvin24basicly, a simple NULL pointer check is missing21:21
ogra_yeah, i see it in the channel21:21
ogra_i guess he went to bed already21:21
marvin24yes, also time for me21:21
ogra_ha, no, he didnt21:21
marvin24dentist is waiting for me late in the night21:22
marvin24oh21:22
wookey_what's the gpu in a nexus 7?21:30
wookey_and an n900?21:31
wookey_I need to build a gles app and am wondering which way lies least pain21:31
wookey_bloody gpu people are such a PITA, making everything painful21:32
ogra_nexus7 is tegra21:32
ogra_321:32
ogra_if yu use the ubuntu image you even have a proper driver preinstalled21:32
ogra_n900 should be PVR ... no idea what version, no idea about drivers21:32
wookey_I thought the ubuntu image used !X of some sort?21:33
gildeandid you see this about tegra-drivers: http://www.phoronix.com/scan.php?page=article&item=nvidia_tegra_3d&num=121:33
ogra_the desktop image uses X ...21:33
ogra_we have two for the nexus721:33
ogra_wookey_, https://wiki.ubuntu.com/Nexus7/Installation21:33
ogra_thats for the desktop image21:33
ogra_(i would do the manual install ... but make your choice)21:34
wookey_right I read tha tpage yesterday - I didn't notice anything about two diff images21:34
ogra_gildean, yeah21:34
gildeanogra_: didn't take them long, only like three years21:35
wookey_OK and 'desktop install' means 'normalish ubuntu'21:35
ogra_wookey_, https://wiki.ubuntu.com/Touch/Install ... thats the android based Ubuntu Touch21:35
ogra_right, desktop means literally desktop :)21:35
wookey_OK, just 'desktop on a small screen'21:35
ogra_yep21:35
ogra_sadly not scaling very well :)21:35
wookey_so you'll still need a keyboard and mouse to do various things21:36
ogra_yup21:36
ogra_there is an onscreen kbd21:36
wookey_OK. I was hoping that I could use the touch UI and make it work reasonably for the small set of about 3 programs I need to run21:36
ogra_but for programming you want a real one or at least ssh21:36
ogra_sure, use the touch image for that ... but that doesnt use X21:37
wookey_but as you say with onscreen keyboard and unity and some tweaking maybe the desktop version can be OK21:37
wookey_problem is that  the main app uses wx and wx needs X (I think)21:37
ogra_you have the choice :)21:37
ogra_ah, yeah21:37
ogra_that keeps you stuck on the desktop image21:38
wookey_OK. that clears some things up anyway.21:38
wookey_It does seem like the least-painful option for now21:38
wookey_hopefully it'll be a bit less crufty than debian on an openmoko :-)21:39
wookey_which was pretty damn crufty21:39
ogra_heh21:41
ogra_it is snappy and pretty usable actually21:41
ogra_i used it as my living room tablet and as ebook reader ... worked fine for that21:42
tassadar_the stuck-mouse-button bug is still there unfortunatelly :/21:42
ogra_yeah, there is a fix upstream though21:43
ogra_sadly apparently hard to backport21:43
tassadar_yeah, saw the discussion on launchpad21:43
tassadar_but from what I can understand from that bug entry, it is set as must-fix before raring release, right?21:44
ogra_well, it was ... now touch came around ...21:44
ogra_we'll see21:44
tassadar_hm, yeah, and mir21:44
ogra_Mir wont enter the archive before release21:45
tassadar_no, but I suppose the nexus7 image doesn't have very high priority...or does this happen on all touchscreen devices, even PC?21:46
=== rsalveti_ is now known as rsalveti

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