/srv/irclogs.ubuntu.com/2010/05/13/#ubuntu-arm.txt

=== JaMa|Zzz is now known as JaMa
tmztStskeeps: hey :)07:37
tmztcan somebody point me at uds channel for this topic for tomorrow?07:37
plarstmzt: they are actually about to get started, but are scattered in different rooms07:51
plarstmzt: see: http://summit.ubuntu.com/uds-m/track/ubuntu_on_arm/07:52
plarstmzt: typically, the channels take the form of #ubuntu-uds-ROOMNAME, where the room name can be found in that link07:52
plarstmzt: so for example, the ASoC discussion will be in mahogony, so #ubuntu-uds-mahogony.  However since it's live, you would just have 1 way communication on irc to ask questions.  To know what's going on, you will want to listen to the icecast stream07:54
tmztthanks07:57
tmztwhat timezone in the panel clock?07:58
tmztoh, got it07:58
tmztit's later than I thought07:58
DanaGIs there an all-discussions stream?08:01
DanaGOr are there multiple things at the same time?08:01
DanaGAnd do they create transcripts?08:01
tmzteverything is in one room?08:02
tmztthere seem to be two talks (X/dpkg for ex) at the same time08:02
tmztin one track08:02
tmztoh, the room is under the title08:03
DanaGI greatly prefer transcripts.08:03
DanaGIt takes me 5 minutes to read something that could take 45 minutes to listen to.08:03
tmzthttp://www.lczajkowski.com/2010/05/04/uds-m-remote-participation/08:08
DanaGoh, and I want glx_ext_texture_from_pixmap08:09
Stskeepsdon't we all08:09
Stskeeps:P08:09
DanaGThat's 99% of my Linux 3D stuff I use regularly: compiz.08:10
DanaGOr how about reviving that old XGL... on top of GL ES?08:10
hrw-udshi08:10
tmztdoesn't gallium have some kind of es emulation now?08:11
DanaGWhat timezone is uds in, anyway?08:11
DanaGOh, and fglrx claims to have some GL_ES strings.... wonder how you'd use that.08:11
hrw-udsDanaG: european one08:11
hrw-udsDanaG: 09:11 here08:11
DanaGProbably it's as xvba-video was: not public.08:11
DanaGs/was/is/08:12
tmztit's a state tracker08:12
tmztanybody who can comment on why recent ubuntu releases are limited to armv6mmu and if that is changing with the new branching/repo strategy?08:18
hrw-udstmzt: armv6? lucid is armv7a08:18
tmztright, so more generally why is this required?08:19
tmzthow many packages of the full repo actually require neon or vfp08:19
looltmzt: All of main was rebuilt for the v7t2 transition08:19
loolSo all of them  :-)08:19
DanaGhttp://sourceware.org/redboot/08:20
DanaGer08:20
DanaGwrong link08:20
DanaGhttps://wiki.ubuntu.com/Specs/ARMSoftbootLoader08:20
looltmzt: The point is to take advantage of v7 and thumb2 instructions08:20
tmztI'm aware of the requirement, I'm asking for what the justification might have been08:20
DanaGThat reminds me of the Angstrom "kexecboot" loader on Zaurus.08:20
tmztsince only certain packages should benefit from those08:20
hrw-udstmzt: get as much as possible from cpu?08:20
looltmzt: No, everything benefits from it08:20
* DanaG scrolls down and sees that mentioned.08:20
looltmzt: thumb2 is smaller and fits better in cache08:21
tmztthis has made projects such as rhobuntu and mer difficult to support a broad range of hardware with vanilla ubuntu repos08:21
loolthe more code in cache, the faster08:21
DanaG"that" being "kexecboot"08:21
tmztso all packages are now thumb2 code?08:21
looltmzt: Sure, but if we kept v5 only, we'd be irrelevant for v7 projects08:21
looltmzt: thumb2 is the compiled code08:21
loolit's instructions08:21
tmztI can understand that for, say, gstreamer codecs08:22
looltmzt: It really benefits everything08:22
looltmzt: It's not like neon08:22
looltmzt: what you say would typically apply for neon08:22
loolsince most packages dont take advantage of it, and the toolchain doesn't use much of it when it can08:23
looltmzt: But when you tell gcc/gas that they should generate thumb2, it makes a big difference because they will generate 16-bits opcodes instead of 32-bits ones08:23
tmzthow much of an improvement is thumb2 over thumb?08:23
loolIn the end, all the little libraries and programs fit better in cache08:23
DanaGUEFI ARM?  That's an odd idea.08:24
looltmzt: It's mostly an improvement in code coverage08:24
looltmzt: it wasn't really possible to use thumb for everything08:24
tmztbut the compiler would fall back to arm when thumb didn't support the needed instruction?08:24
DanaGThu May 13 00:26:10 PDT 201008:26
DanaGtime for me to go to bed. =þ08:26
DanaGAnyway, I suppose the eventual blueprints will be enough of a "transcript" (or rather, will distill down the overall points).08:27
looltmzt: Yes, you'd have to switch back and forth08:28
DanaGOr stream in stereo.08:28
DanaGLeft side in one room, right in the other.08:28
DanaG08:28
DanaGhttp://www.arm.com/about/newsroom/24978.php08:29
DanaGNifty.08:29
DanaGUEFI including ARM.08:29
DanaGNow... we just need Ubuntu to support UEFI!08:30
DanaGActually, I can imagine HP, for example, liking that.... they already use UEFI on their business laptops.08:31
crimsuncf. https://wiki.ubuntu.com/FoundationsTeam/Specs/MaverickUefiSupport08:34
crimsunwhich was discussed the first day ;)08:35
DanaGSweet.08:35
DanaGI've messed with UEFI on my laptop... first-gen Elitebook.  Has somewhat broken memory map.08:36
tmztis there any public discussion of qualcomm kernel-qcm?08:36
=== hrw is now known as hrw-uds
=== markos_ is now known as Guest60377
=== kmargar is now known as markos_
markos_has anyone tried to build mysql-server from the source package (karmic) on arm? version mysql-dfsg-5.1_5.1.37-1ubuntu5.1, here's the pbuilder log: http://pastebin.com/Y7fwHrV710:32
markos_no matter if I build it using pbuilder or without, I get the same result10:33
markos_I'll file a bug report, but I thought I'd ask here first just in case anyone else has seen this before10:33
hrw-udsmarkos_: no idea - I am new member of ubuntu/arm team10:35
markos_hrw-uds: congrats then :)10:36
hrw-udsthx10:36
ograNCommander, around ?11:16
NCommanderogra: yeah, I'm in the basement11:18
NCommanderogra: where are you?11:18
ograNCommander, in the UI heads session11:18
ograNCommander, so i talked to colin11:19
NCommanderogra: (I've been having some thoughts about doing hd-install, and I'm not sure this is the proper solution)11:19
ograNCommander, after a bit of going back and forth we came to the conclusion that actually the best solution is an OEM image11:19
NCommanderogra: ?11:19
ograits trivial to change livecd-rootfs to omit casper and instead add oem-config.-gtk11:20
NCommanderogra: didn't I suggest that earlier?11:20
ograroll a live image thats going to a real FS on antimony11:20
NCommanderogra: its not a bad idea expect then we need a way to resize the DD'ed SD card image11:20
ogranah, leave it to the user to add more partitions, we can put gparted in11:21
NCommanderogra: I thought your objection to doing oem-config in place of ubiquity was a speed issue and such11:21
ograit is11:21
NCommanderogra: ugh, that stinks of HACK11:21
ograbut we need to fix that anyway11:21
NCommanderogra: fair enough. if a live image is acceptable, this works for me.11:21
ograwell, its not a live image11:22
NCommanderogra: since its trivial to drop ubiquity (we just need to change the second package installation pass)11:22
ograits essentially a rootfs on a artitioned image11:22
NCommanderogra: well, from the perspective of image building :-)11:22
ogra*partitioned11:22
NCommanderer11:22
NCommanderhrm11:22
NCommanderactually11:22
NCommandertihs is a new type of image now that I think of it11:22
NCommandersince we'll have no squashfs11:22
ograright11:22
ograwe only use the squashfs as a base11:22
NCommanderogra: I rather just have it dump a tarball and let antimony learn how to deal with it; it will greatly reduce build time11:23
ograbut do the content copying that ubiquity would usualy do during image build11:23
NCommandersquashing an image takes ~1 h11:23
ograyeah, true, we can add such a function11:23
NCommanderogra: /etc/fstab, /etc/passwd, /etc/group11:23
NCommanderf-k-i will have to be run11:23
ograthe only thing that concerns me a bit is the actual image size11:23
ograyou dont need fstab beyond proc11:24
NCommanderogra: we might want to look at using livehelper instead, since this looks like a massive pain to hack onto livecd-rootfs11:24
NCommanderoh11:24
NCommanderfair enough11:24
ograpasswd and group are set up by oem-config11:24
NCommanderogra: but then there won't be an fstab entry for /11:24
ograyou dont need one11:24
NCommanderogra: I thought oem-config needs an account added to run at boot time11:24
ograit doesnt11:24
NCommandercasper makes one for the livecd environment, and the installer adds an ubuntu:ubuntu user for first boot11:24
ograit adds one actually :)11:24
ograthats the live session11:25
ograoem-config doesnt need any user11:25
NCommanderogra: then what user does GDM log in as?11:25
ograso i'd like to build two kinds of images ...11:25
ograthere is no gdm running, oem-config is the DM11:25
NCommanderDM?11:25
ogradisplay manager11:26
NCommanderogra: the dm is started up by GDM, which is set to autologin on first boot with oem-config11:26
ograan oem-config image boots into a stripped down ubiquity11:26
ograad the DM11:26
ograinstead of GDM11:26
NCommanderI've got to be missing something11:26
NCommanderBut that's fine11:26
ograyes :)11:26
NCommanderogra: er, ubiquity launchs GDM though in all cases though ...11:27
ogratry a rootstock build and install oem-config-gtk in it11:27
NCommanderogra: I'll take your word for it :-)11:27
ograubiquity doesnt touch gdm11:27
ograso11:27
ograi'd like us to build two images11:27
NCommanderlivecd-rootfs needs to grow a new function mode11:27
ograstarting out with  minimal one11:27
ogra+a11:27
NCommanderogra: minimal? like just a basic command line system?11:28
ograwhich is only a basic debootstrap+oem-config11:28
ograright11:28
ograwe'll need a seed for that11:28
NCommanderogra: we can't use tasks in PPAs11:28
NCommander(limitation of Launchpad)11:28
ograand the other one should be a stripped down netbook11:28
ograthats why i want to use a seed :)11:28
ograso we can have a metapackage11:29
NCommanderogra: right, but we'll have to make sure we keep the metapackages up to date11:29
NCommander:-)11:29
NCommanderah good, we're on the same page, we're just cross-typing each other11:29
ograthey live in the PPA, shouldnt be a prob11:29
NCommanderright, we did this for Moblin11:29
* NCommander shivers11:29
ograso what i'd like us to come up with is some idea for the size issue11:29
NCommanderogra: hrm, we can't put stuff on an ext4 image though11:29
ograi dont want to end up with 3G images11:30
NCommanderAt least, not if we use antimony11:30
ograwe can use ext311:30
NCommanderSlow, but doable11:30
ograno pob with that11:30
NCommandere2tools is VERY slow :-/'11:30
ogranope, pretty fast11:30
NCommanderIt is? Not the last time I used it11:30
ograright, we will use gene2fs11:30
NCommanderOH11:30
NCommanderThat works11:30
ograindeed11:30
ograand doesnt need root11:30
NCommanderRight, so d-cd needs to grow a new ext2/3 backend11:30
ograneeds to tinker with the journal afterwards though11:31
NCommanderI already have code on a local branch for that11:31
ograso it gets converted to ext311:31
NCommanderogra: how do we do that though, we can't loopback mount as non-root11:31
ograwhy would you loopback mount anything ?11:31
NCommanderogra: to twiddle the journal with tune2fs11:31
ograno need for that11:32
NCommanderunless you another way to get journals enabled11:32
ogratune2fs can work on .img files11:32
ogradirectly11:32
NCommanderdidn't know that11:32
NCommanderNifty11:32
NCommanderogra: the lack of ext4 is a bit disappointing, but not a show stopper11:32
ograext3 is fine11:32
NCommanderogra: so the images will have two partitions. vfat /boot, and / ext3, right?11:33
ograi was planning to implement the same setup in rootstock to overcome the usage of root11:33
ograso i can just copy/paste the PoC from there into antimony11:33
ograthree11:33
NCommanderogra: we can expand the partition map on first boot now that I think about it. OEM had to do that on the Dell mini's once11:34
ograerr11:34
ogratwo11:34
NCommanderWe just have to look at how they did it, and then make the ext filesystem grow11:34
ograbut no vfat /boot11:34
NCommanderogra: ?11:34
ogradpkg doesnt cope with vfat /boot11:34
NCommanderugh11:34
ograwe will do something similar to imx5111:34
NCommanderogra: so flash-kernel voodoo?11:35
NCommanderyeah11:35
NCommanderOk, so I see how this will come together11:35
ograa raw hidden partition that pretends to be a flash11:35
NCommanderd-cd will grow ext2/3 image support, that image is passed into {post-}boot scripts, and then we handle it like we do on imx51 and dove11:35
ograflash-kernel will then mount that and cp uImage/uInitrd to it11:35
ograthat code from oem ... does that automatically grow the partition ?11:36
NCommanderogra: it had to, but I don't know the full story behind it11:36
NCommanderogra: but its a solved problem, we just need to figure out how they solved it11:37
ogracan you try to get your hands on that code and test it ?11:37
NCommanderogra: let me see if I can figure out who solved it originally, this was almost a year ago11:37
ograi'll look into livecd-rootfs and the build scripts11:37
ogra(or feel free to do the d-cd changes and give them to me)11:37
ograso we have something to get started from11:38
NCommanderogra: I can handle the build scripts/d-cd if you want, I have local infrastructure11:38
ografine then11:38
ograi'll put an action point list together and write up an explanation of what we'll do to david11:39
NCommanderogra: right, so I'll make d-cd grow ext support, you'll make livecd handle making a pre-installed image11:39
NCommanderor do you want me to handle the entirity of the image stack?11:40
ograright, and you try to get your hands on the grow partition code11:40
NCommander(this is work I enjoy doing so I don't mind taking it)11:40
ogralet me make up the bulletpoint list and we look who does what then (by mail), if you want to start already, start with d-cd11:41
NCommanderogra: execellent, I'm just fixing up my local build infrastructure here now11:42
=== Meizirkki_ is now known as Meizirkki
=== NCommander is now known as Guest79402
=== JaMa is now known as JaMa|off
aaron_what 's the dbfifo  use for ?14:55
aaron_armel-rootfs-201005131535.tgz ?14:59
aaron_i cannot find the img file14:59
ogradid you tell rootstock to actually build an img ?14:59
ograby default it only builds a tgz14:59
ograyou need to use --keepimage15:00
aaron_but how to convert armel-rootfs-201005131535.tgz  to an img file ?15:00
ograits described on the RootfsFromScratch wikipage (see channel topic)15:01
aaron_dd if=/dev/zero of=ubuntu-arm.img bs=1MB count=0 seek=102415:16
aaron_what size i should to set ,i don't want to waste the space15:17
hrw-udsdd if=/dev/zero of=myimage.img bs=1M count=102415:35
hrw-udsmkfs.ext3 myimage.img15:35
hrw-udsmount myimage.img /mnt -o loop15:35
hrw-udssudo tar xf tarball.tgz into /mnt15:36
=== jmcgee is now known as jmcgee|gone
=== jldugger is now known as pwnguin
=== JaMa|off is now known as JaMa
waleedhello21:15
waleedam trying to run ubuntu in my iphone21:15
waleedis there any one tried to do something like that?21:15
tmztwaleed: rhobuntu22:05
tmztwhat phone?22:05
waleedIphone22:06
waleediphone 2g22:06
tmztwith the new android kernel support it should be possible22:06
tmztI can try to help you in #htc-linux but I'm not in #iphonelinux or whatever22:06
=== ndec is now known as foo_
=== foo_ is now known as ndec
=== ndec is now known as ndec_
=== ndec_ is now known as ndec_home
=== ndec_home is now known as ndec
=== JaMa is now known as JaMa|zz

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