=== asac_ is now known as asac === prpplague^2 is now known as prpplague_afk [06:13] persia: you around? [08:36] lilstevie: Am now. [08:39] persia: I am just looking at building kpkgs [08:39] and building one for the transformer [08:43] I usually recommend starting with https://wiki.linaro.org/Resources/HowTo/PackageYourOwnKernel [08:44] jcrigby has a fairly flexible packaging branch that you can merge to get things working. [08:45] ok [08:45] Take care to look through it, and make sure that you've set the flavour correctly in all the places it appears (reading the build log the first couple times should highlight some obvious ones, and running dpkg --info on the result packages will show most other bits that have issues. [08:46] cool [08:46] :) [08:49] jcrigby is usually around on this channel: I'd recommend trying to catch him in your morning though, given timezones. [08:50] Or I'm happy to look at things :) [08:52] :) [08:52] well I am just cloning into a seperate folder from my actively used directory so that it doesn't get messed up :) [08:53] persia: things are coming together nicely though :) [08:59] Excellent! [09:13] persia: http://forum.xda-developers.com/showthread.php?t=1191141 && http://rootzwiki.com/showthread.php?1072-ubuntu-on-the-transformer/ <-- :0 [09:13] same release [09:13] :p [09:16] Firstly, I'll rcommend explicitly listing the license for that code, indicating what folk who get it from you can do with it. [09:17] Ideally, you'll pick something that we can put in Ubuntu: I'd really like to be able to generate an image for that fairly easily, so we can maybe include it in the next release. [09:18] persia: there is no licence available for nvflash [09:18] There is a license, but it doesn't permit redistribution. [09:18] But you want to express a license for *your* code. [09:20] in that pack the only code of mine is 1 line in the kernel which changes the behaviour of a button on the keyboard dock [09:20] Also, you reference a PPA: do you especially want a PPA, or should we just put it in Ubuntu proper? [09:20] Ah, heh, for one line of code, if you don't care, then just file a bug :) [09:21] heh [09:21] most of this took linux nicely just needed tweaks to configs [09:22] well putting in ubuntu proper overall would work, just in the mean time for bt and wifi fixes [09:22] currently bluetooth works but I want it to be a package thus using ppa, and wifi, well that is going to be an active piece of development [09:28] If a package can *exist* (even buggy) and be uploaded by 11th August, we can get it included in oneiric, and keep fixing bugs until sometime in September (depending on the bug and the risk of regressions). [09:28] We can then backport that to the 11.04 release, to make sure that users of the current release are also supported. [09:28] persia: the bt stuff is a bit sketchy though [09:29] That's OK. It's not like we're stiving to have transformer support in the default image set for oneiric. [09:29] cause it also needs the firmware [09:29] no I mean it works :p [09:29] it is firmware [09:29] same as wifi [09:29] I'm not 100% sure on the distribution rights [09:29] What's the license on the firmware? In general, if you can put it in a PPA, you can probably put it in multiverse (unless you're paying for the PPA) [09:29] well I wasn't going to put it on the ppa either [09:29] Double-check. If we can't redistribute, we can'T do a PPA either, because we don't have teh right to upload to the PPA. [09:30] Oh, heh :) [09:30] what I have at the moment is extracting from an android system image and injecting [09:30] post install [09:31] So, expecting the user to already have the firmware available in their andoid install? [09:31] Did you put that in the postinst? [09:31] no I have been telling people to do it :p [09:32] because if you flash pure linux there is no android install on device [09:32] So, there7s a tool for that :) [09:32] Right, so here's a sketch implemetnation [09:32] heh [09:33] Add debconf support to the package, and ask the user if they have a dual install, and if so, where to find android (medium priority) [09:33] Set a default for that question that matches the results of running your installer on an unmodified retail image. [09:34] THen, in the postinst, check to see if the firmware is available already. If not, check if the user told you there was a dual install, and if so, get the firmware from there. [09:34] If there is no firmware already loaded, and there is no dual-install from which to get the firmware, notify the user (again, via debconf) that you were unable to find the firmware, and point them at instructions for extracting it themselves. [09:35] heh [09:35] that would do it [09:35] http://pastie.org/2288888 [09:35] ^^ wifi firmware licence [09:36] 2.2 meets the requirements for multiverse. [09:37] Basically, that states that we'll only redistribute it unmodified and under the same license terms, and that the redistributor promises to take responsibility if they break this rule. [09:37] ok [09:37] good to know [09:38] I believe the bt firmware would be the same, although I haven't been able to find the licence agreement for that, it is on the same silicon [09:39] Mind you, I'm not an archive admin, so my opinion isn't authoritative, but that it's a "restriction on distribution" clause, rather than a "no distribution" clause, and seems to say something like "If you redistribute this, be sure the recipient can'T tell the difference between what they get from Broadcom and what they get from you", it seems like it ought be OK. [09:39] ok [09:39] Unfortunately, we need to get a license, because the terms of that license require the distribution of the license for any redistribution. [09:39] yeah [09:45] check-config is failing [09:45] http://pastie.org/2288934 [09:49] I'm not that familiar with check-config: you may need to change some things to be modular, or make sure you hit minimum limits. [09:49] ah ok [09:49] Also, check the n900 or ac100 kernels: maybe there's a workaround in one of those (or some clue that mayhelp) [10:22] fixed it up :p [10:22] Cool! [10:22] reading the instructions helped :p next instructions delt with it [10:22] What was wrong? [10:22] Heh :) [10:22] well it is expected [10:22] thats all :p [10:23] made the changes and the only 2 things that had to be commented out were apparmor and ram_size [10:25] Not much we can do about ram size, but why doesn't apparmor work? [10:25] Did you merge in the ubuntu sauce branch? [10:26] heh not that it doesnt work I just haven't fixed up that part of the config yet [10:27] fixed that up now [10:27] Ah, OK. [10:28] building now :) [10:50] Hi! What kind of toolchain is used to compile ubuntu-arm? [11:11] ppisati, the new 3.0 panda kernel should have good wireless support? [11:11] the one in oneiric [11:23] janimo: uhm, i guess it [11:23] janimo: but never really tried it since i don't have antennas for it [11:24] ppisati, hmm I thought it should work w/o antennas too. I never mounted any antenna on the panda. [11:24] anyway will test. I dislike running a long ethernet cable through the room for the panda only [11:25] it worked at one point in natty but was flakey. Need to make sure it works without a need for software from PPAs [11:25] janimo: k, let me know then [11:30] janimo: 3.0 linaro works on panda with wireless [11:31] without antenna I got 4Mbps from AP under panda === zyga_ is now known as zyga-x11 [12:07] hrw, stock packages from the archives only right? [12:28] janimo: let me check [12:29] heh. panda rootfs died [12:29] linux-image-3.0.0-1402-linaro-lt-omap_3.0.0-1402.3~ppa~natty_armel.deb sounds like non ubuntu package [12:43] hi! I want to compile ubuntu for Pandaboard, because i need some driver. Panda boots form SD, ubunzu needs 4GB, but I only have 2 [12:44] should I follow this instructions? [12:44] http://www.omappedia.org/wiki/Ubuntu_kernel_for_OMAP4 [13:45] persia: hit another error I cant figure this one out though :/ [14:18] pandaboard is omap4 right? [14:45] zul: correct. [14:53] lilstevie: What error? [14:53] http://pastie.org/2290069 [14:57] persia: I am trying a start again approach [14:58] seems to be what to do [15:00] lilstevie: Yeah, that's probably how I'd handle the same issue. No idea why it suddenly stopped working. [15:00] persia: well seems to be a common fix to the issue [15:02] :) [15:17] issue fixed :) [15:17] \o/ [15:17] now I just need to modify flash_kernel for my hardware :p [15:18] linux + flash-kernel + d-i is usually the enablement set [15:19] well, if you can use the vendor bootloader. If you can't, then you need that as well. [15:19] d-i? [15:19] I am using the vendor bootloader [15:20] d-i is debian-installer. Every installer we write is eventually refaactored to be based on d-i, because all the problems are already solved there. [15:20] ah [15:21] (or well, for the problems not solved there, solving them there ends up solving them for all the different installers) [15:23] heh [15:23] well flash-kernel is somewhat of a tough nut [15:24] I'm not happy with the hardcoded nature [15:24] Yeah. lool keeps talking about redesigning it to be sane, but it's a hard problem. [15:25] Ideally, we'd like a generic solution for all fastboot devices, but feel free to hardcode it for now. [15:25] I thikn ogra landed hist ac100 stanza in oneiric: if not, you can pull it from his PPA. [15:26] the problem is once we solve the issue of uploading to a 0x808 type partition with nvflash without having to set a size tag for it there will only be 3 partitions in 1 configuration vs the 10 in an android dualboot config [15:26] Right. [15:27] 0x808 tells nvflash to fill to the end of the partition minus $SIZE fwiw [15:27] My general strategy is to apply a hack that works today (assuming I'm not breaking someone else's stuff), and then fix it once the necessary blockers are out of the way. [15:27] so setting size large enough means having the img size unallocated at the end of the emmc [15:28] heh well today hacking it in hardcoded will work [15:28] Blocking on the right solution tends to make it hard to get the first draft working. [15:28] Yep :) [15:29] flash-kernel has ended up bein used in many more kinds of devices not originally planned [15:30] or well, it was never planned. it was just a script to get kernels and initramfs to nas device nand flashes [15:30] suihkulokki: Indeed. IT has (sadly) become a generic means of placing kernels, rather than just writing to flash for certain devices. [15:31] heh [15:31] What do you think about requesting u-boot-installer and fastboot-installer (to parallel lilo-installer, grub-installer, etc.), at least for those platforms that these support? [15:32] well I still need something like it, cause of making the bootimg [15:32] lilstevie: RIght, but for non-ARM, we tend to have bootloader-specific scripts, each of which is fairly generic [15:32] ha [15:34] ah* [15:59] persia: http://pastie.org/2290481 <-- kernel packed [16:02] Are we going to have the Xubuntu armel51 images for A3 ? [16:02] * charlie-tca won't be disappointed to not have them yet [16:04] lilstevie: Wonderful!!! Post the .dsc (and all the files referenced in the .dsc) somewhere we-baccessible, and I'll pull, review, and upload as soon as I get a chance. [16:05] charlie-tca: THere's some issues related to the mx5 images. At least the patches I wrote that were supposed to generate them ended up with the screen not working (I've been working on graphics drivers this week). [16:05] persia: I am repacking with a things changed [16:05] charlie-tca: I have heard of others working on mx5 images, but haven'T seen anyone be successful yet. [16:05] persia: do you also want a source package? [16:05] Thanks for not forgetting us, persia [16:05] So, I'd say, very unlikely for A3 [16:06] lilstevie: .dsc+references files == source package (yes) [16:06] No problem. Just have to check once in while [16:06] ok :) [16:06] charlie-tca: Never :) Although I'm not sure I'll run Xubuntu on my mx5 hardware, the problem in generating the images is the same for all flavours. [16:07] persia: launchpad prefered? [16:07] or anywhere [16:10] lilstevie: I don't really care: I just want them all in the same place, so I can get them in a single virtual directory over http or https [16:11] so not in a tarball then? :p [16:12] If you especially want: a tarball is harder for me to download. [16:12] * persia uses dget to get packages, which auto-unpacks stuff, whereas a tarball has to be downloaded, unpacked, then unpacked agin, etc. [16:13] its cool, [16:13] I enabled directory browsing over http [16:13] Didn't need that, as long as they were all in teh same directory. [16:18] heh [16:35] persia: uploading now [16:36] Cool. [16:37] Now that you're less distracted, I should probably ask you about the tab kernel. [16:38] heh I was wondering if you were going to ask about that [16:38] now that I have the process for generating this down I will get that one set too, [16:40] we have the 10.1 working now [16:40] like same bug with the touchscreen [16:40] but :) [16:42] That's OK. Software always has bugs. Let's get it in the repo before the deadline, and then, if ywe're lucky, we'll have users, and we can get more input to fix the bugs. [16:44] :) [16:51] hmm.. i'm trying to use gst-launch to play some video, but i'm only able to get non-accelered stuff to show (via fbdevsink)... using v4l2sink shows only black screen [16:51] what am i missing? [16:56] lilstevie: I'm getting told me move, but please post the URL, and I'll pull it as soon as I'm sitting somewhere again. [16:58] http://lilstevie.geek.nz/kernelsrc [16:58] upload will finish in about 15 minutes [17:15] persia: ok it is up, [17:51] live ubuntu arm image I can save to the sdcard of my atrix4g phone? [17:51] I just wish to run apps... it already has an x-server and linux-arm boot scripts [17:51] and linux kernel etc [17:56] ubuntu-arm livecd or qemu image? [18:07] mynameisdeleted: You might have some luck with our new Ubuntu-Core image as a chroot environment. From it, you can apt-get install all of the ubuntu-Arm packages. http://cdimage.ubuntu.com/ubuntu-core/ [18:08] It is Oneiric based, so there may be some instabilities, but this is one of the use-cases that we designed this image for. [18:08] (chroot environment that is). [18:10] and I use mont --bind to share /tmp and X11 sockets and such with hostos as well as /proc etc [18:10] and then the chrooted apps dont know the xserver pre-installed is on a different root [18:12] 802.11n built into phone is cool [18:12] as is dual-core 1ghz arm processor and 1GB ram [18:12] I dont think most arm netbooks have more than 1 core at 1ghx [18:12] nore do most atom netbooks [18:13] Not sure how to resolve dependency issues. But if you have a big enough SD (2-4G minimum), having duplicate binaries won't be an issue. It won't try to relaunch an X session. [18:13] so for ubuntu.... intel atom vs arm risc processor? [18:13] which is better chocie? [18:14] depends on what you want to do [18:14] I guess arm is slower but uses less power? [18:14] not sure about 1ghz dual-core arm though [18:14] an a8 is ~20% slower than an atom [18:14] vs single-core atom [18:14] Again, it really depends. Video playback is very comparable. [18:14] a8 at 800mhz [18:14] single-core atom [18:15] running at 1.6 [18:15] I have olympus made 2 core arm processor at 1 ghz precisely and 2000 bogomips [18:15] hwo does 1ghz armv7 core comapre to 1.6 ghz atom core? [18:16] bogomips = bogus mips [18:17] mynameisdeleted: There are a lot of benchmarks that show pluses & minuses for each arch. If you plan on running games, you may be out of luck. But for video playback (with accelerated drivers), they are very comparable. Also web usage and OpenOffice are fairly close. [18:18] It is really hard to compare without a specific task and as close to equivalent hardware configuration. [18:19] For example, Atom with Dual SATA raid will clobber arm with eMMC. [18:19] openoffice + web browsing I think [18:20] will class10 32GB sdcard with 20MB/s read sustained and 10MB/s write sustained be comperable to sata? [18:20] Then you should see similar performance, albeit with longer battery life on arm. [18:20] how does class10 flash memory comapre in performance to a traditional drive? [18:21] if it sustains 20MB/s read and 10MB/s write [18:21] It depends on the SD implementation on your system. I have seen dismal performance on some high speed SD cards. [18:21] so it could be good or could be bad [18:21] And I don't know of any SD that compares to SATA. [18:21] it will be an experiment [18:21] Most are designed for high speed sequential reads/writes. Linux is more random. [18:21] many solid-state drives blow sata out of water [18:21] ahhh [18:22] mynameisdeleted, SD card performance for the vast majority of systems will absolutely suck compared to any SATA drive [18:22] but there are flash-based filesystems that are more optimized for using sequences when possible [18:22] a better choice of filesystem might make sd performance more acceptable [18:23] The flash filesystems are still in heavy development, but are greatly improved over traditional (ext3) filesystems. [18:23] is internal memory likely faster than microsd? [18:24] on flash a read can be random .. just not a write right? [18:24] I have experimented with btrfs & nilfs. For other reasons, we are not using them in our default SD images, but we are keeping a close watch on their progress. [18:24] usually ext4 and btrfs are a lot faster on SD card [18:24] or boht read and write work better sequentially? [18:24] it's mostly about mballoc and copy on write features, it reduces the load [18:24] ext3 on an SD card is abysmal [18:24] so ext4 is the most stable sd choice that will work well [18:25] If you want a simple test, use bonnie++ to do filesystem benchmarking. [18:25] that's what we ship on the Efika [18:25] can that be an ext4 disk image saved ontop a fat file system? [18:25] or is ti better to partition an sd-card? [18:25] you want to partition it [18:25] either use the full card so the filesystem starts at 0 [18:25] or if you partition it, make the partition starts aligned to 4MB [18:25] that seems to be the best case for performance [18:26] so when my 32GB class10 card copmes in I'll format/install it from my mac [18:26] (otherwise, MBR takes up 512 bytes and the filesystem is kind of 512 to 4 kilobytes out of alignment) [18:26] and set it up as an ubuntu arm drive [18:26] you end up doing twice as much work to write a block than you would otherwise if it was aligned [18:27] I can make the mbr take exactly 4MB though right? [18:27] and then the ext4 starts in right place [18:28] do something like parted /dev/mmcblk0 -anone and then mkpart primary 4M 100% [18:29] how hard is it to make x-sever use built in display in multitouch mode rather than use external hdmi? [18:29] I guess thats called find an x-server app for andoird [18:31] also there problaably is no ubuntu version for mobile phone... only for netbooka nd maybe tablet [18:31] as long as it uses a keyboard and a real display ubuntu is fine [18:31] I can run a terminal emulator on the phone for commandline stuff outside that [18:44] lilstevie: Pulling now. Thanks. [19:00] mynameisdeleted: you might want to look into utouch [19:01] I might do that [19:43] rsalveti, do you know if the TI firmware package is still used with the omap4 wlan driver that is in the kernel? AIUI the driver from the PPA is no longer needed, but firmware should still be there :) [20:45] janimo: firmware should come now from the linux-firmware package [20:45] it's all upstream afaik [20:47] rsalveti, ah you're right. I found the ti-connectivity binaries in a git repo with all firmware, but forgot there's one package for all firmwares now. thanks [21:02] janimo: The wlan firmware has been part of the linux-firmware since Natty. It is in the desktop images by default, but not in the headless/server images. [21:07] GrueMaster, that may explain my wifi woes on my headless panda. But I hope I'll put an end to them now that I know what to look for [21:08] Yea, I haven't gotten it to work fully on headless in a while, but forgot what I had to install to get it working. Not really a high priority image test. [21:26] of course if instead of firmware not found errors, the panda locks up when ifup wlan0 is called is progess, just not the kind I hoped for [21:51] janimo: which kernel, 3? [21:51] rsalveti, yes. But after a reboot it did not hang anymore [21:51] hm [21:51] so hung twice but now seems ok [21:53] I'm running the daily desktop on one system with 3.0 kernel and wifi is fine there. I'm in the process of reimaging my cluster of 4 server systems, will bang on one as soon as it is up to see what may be missing. [21:53] cannot ping it again. Will debug in the weekend [21:59] Ok, that was easy. sudo apt-get install wireless-tools;sudo iwconfig wlan0 essid grue-net2 ; sudo dhclient wlan0 [21:59] Currently ping flooding it now. [22:02] janimo: What image were you running? [22:02] GrueMaster, an old natty dist-upgraded and full of cruft [22:02] on a USB external disk [22:03] my ARM build machine [22:04] Ah. This is oneiric. I could try from my SRU tester. I put Maverick, Natty, and Oneiric on separate partitions of one usb drive for easy SRU testing. [22:04] GrueMaster, I wonder why that does not work for me [22:04] I am on oneiric as well [22:04] simple iwconfig essid never worked for me [22:04] plbkac? [22:04] could be my router/AP being more demanding [22:05] no idea, but there is probably some dose of pebkac involved [22:05] heh [22:05] not literally of course, I stand, do not use a chair [22:06] but no problem, imx arrives next week, and instead of two half functional ARM/Oneiric devices (panda, ac100) I'l have 3! joy [22:06] Personally, I can't stand all of this testing I have to do...so I sit. :P [22:06] I haven't even booted mine. ENOTIME. [22:06] And lack of an image has made it very low-p. [22:08] could be hw issue with my panda(doubt it), more iffy AP (openwrt) or indeed something I overlook. But I never had good wifi on my panda [22:10] Yea, I kind of have a bit of overkill here. Two 802.11N AP's at either end of the house (one 3 feet from my panda stacks), and the entire house wired for gigabit.. [22:11] Although I have just filled my 16 port 1Gb switch in my office. sigh.