**** BEGIN LOGGING AT Fri Apr 29 02:59:58 2011 Apr 29 03:11:15 dcordes Apr 29 07:14:54 ogra_: hi Apr 29 08:48:08 hey ndec Apr 29 08:48:33 ogra_: just wanted to confirm that I will be at UDS, from Mon to Wed. Apr 29 08:48:38 Wed all day Apr 29 08:48:57 awesome !!!!! Apr 29 08:49:52 dont forget to register on LP Apr 29 08:49:59 ;) Apr 29 08:50:18 ogra_: that's done .. Apr 29 09:47:44 ndec: welcome, man Apr 29 09:50:20 ndec: cool Apr 29 13:37:42 guys, did you ever use jtag on the mvl-dove board? and if yes, which hw/sw did you use? Apr 29 13:37:49 have you ever tried with openocd? Apr 29 13:39:46 i dont think anyone has touched dove in over a year Apr 29 13:39:59 yep Apr 29 13:39:59 (and i dont think anyone in the team is eager to do so) Apr 29 13:40:04 :) Apr 29 13:40:08 no no Apr 29 13:40:19 it's just that i'm debugging an issue on ti-omap3/4 Apr 29 13:40:30 and to me, it looks like a stack corruption Apr 29 13:40:32 we had a special dongle for jtag iirc Apr 29 13:40:43 so i would like to see how it works on another arm platform Apr 29 13:41:08 i have a fly swatter and i'm using openocd Apr 29 13:41:30 i just need to know which profile i should use Apr 29 13:41:33 BTW Apr 29 13:41:46 ppisati, asl pprplague Apr 29 13:41:49 *ask Apr 29 13:41:51 as far as you know, is our dove equal to a sheevapolug? Apr 29 13:42:02 no, not at all Apr 29 13:42:06 uh ok Apr 29 13:42:19 its similar but not equal to the armada chip Apr 29 13:42:25 ok Apr 29 13:42:30 its a completely experimental platform Apr 29 13:42:48 i see Apr 29 13:43:20 talk to NCommander for more details, he maintained the port Apr 29 13:43:26 ooooook Apr 29 13:43:45 btwm who is pprplague? Apr 29 13:44:10 works at tincantools and is a contractor at TI afaik Apr 29 13:44:32 prpplague is good person to ask about JTAG stuff Apr 29 13:44:40 oh cool Apr 29 13:44:47 my jtag is from tincantools :) Apr 29 13:45:15 thats why i pointed you to him ;) Apr 29 13:45:36 ok, i'll chase him Apr 29 13:46:16 my jtag equipment is from openmoko Apr 29 13:48:36 anyway, i senti an email to saeed Apr 29 13:48:50 let's see if he knows the correct profile/cpu for our dove board Apr 29 14:41:57 ndec: https://launchpad.net/~rsalveti/+archive/unity-3d-gles/ Apr 29 14:44:35 ndec: activate the ppa, install compiz-core and compiz-gnome then once you have your unity-2d session up and running call "compiz --replace composite opengl move resize decor" Apr 29 14:52:35 rsalveti: compiz gles will land in oneiric? Apr 29 14:53:01 hrw: that's the idea Apr 29 14:53:25 I believe that will be the natural path once Amaranth merge his branch with upstream Apr 29 14:53:36 but still need more love to make it work properly Apr 29 14:54:15 one step closer to desktops Apr 29 14:54:47 hrw: you meant to modern desktop... as it's already a desktop Apr 29 14:55:07 ndec: my desktop is not modern then - no compiz Apr 29 14:55:11 ;D Apr 29 14:55:39 hrw: could it be that you are not modern ;-) Apr 29 14:55:45 pfft, unity-2d is so much better anyway Apr 29 14:56:20 ndec: have to admit that I don't use unity3d, but standard desktop, but i still use the effects... Apr 29 14:56:23 ndec: if modern==use-unity then I will never be modern Apr 29 14:56:47 [ 1422.107879] Buffer I/O error on device mmcblk0p2, logical block 326038 Apr 29 14:56:47 [ 1426.970520] mmcblk0: error -110 transferring data, sector 6924832, nr 8, card status 0xc00 Apr 29 14:56:52 I hate SD cards Apr 29 14:57:02 hrw: not what i meant... i am not using it either. Apr 29 14:57:06 [ 1411.778564] JBD: I/O error detected when updating journal superblock for mmcblk0p2. Apr 29 14:57:09 [ 1411.786682] journal commit I/O error Apr 29 14:57:14 hrw: panda or beagle? Apr 29 14:57:19 panda Apr 29 14:57:29 EA1/ES2.0 Apr 29 14:58:00 rsalveti: I hope that at luds-o there will be some good non-TI boards available for linaro people Apr 29 14:58:28 I am tired of omap[34] based boards Apr 29 14:58:56 well, make nvidia sign up :) Apr 29 14:59:00 get an ac100 Apr 29 14:59:19 hrw: don't know when we'll be getting snowballs, and also not sure about the current state of the sw for it Apr 29 14:59:38 but that will be an option soon Apr 29 15:00:06 ogra_: as linaro engineer I should use linaro hw so even ac100 will not free me from using other boards ;S Apr 29 15:00:07 nvidia is harder, they don't want to play the opensource game correctly Apr 29 15:00:18 rsalveti: We get those every year in winter. Oh, you were referring to something else. Apr 29 15:00:34 GrueMaster: hehe :-) Apr 29 15:00:39 GrueMaster: snowball is ST-Ericsson A9 board Apr 29 15:00:49 Ah, my bad. :P Apr 29 15:00:58 GrueMaster, well, keep the software after you melted one ;) Apr 29 15:01:09 in a bucket :) Apr 29 15:01:56 oh hey I have a comment on the Oneric goals thing for ARM.. I think supporting a raw partition for kernels etc. is kind of a weird idea. Any attempt to override some kind of filesystem management of kernels is going to just make everyone hyper-reliant on flash-kernel to update their systems where it could just be n ext2 partition. There's absolutely nothing wrong with U-Boot ext2 support (except one weird error code which we have a patch for :) Apr 29 15:02:14 I forget who's little braindump that was listed under, maybe rsalveti? Apr 29 15:02:24 no, it was mine Apr 29 15:02:28 ndec: at this ppa you can also grab the unity-2d package, I enabled the qt gles backend by default on it Apr 29 15:02:41 is it specifically to support weirdo systems like AC100 that don't actually run U-Boot? Apr 29 15:02:42 with the redesigned flash-kernel it shouldnt be a prob to rely on it ;) Apr 29 15:02:54 ndec: but for that you'll need the latest driver at omap-trunk ppa, I exposed EGL_NV_post_convert_rounding so Qt can work fine with the PVR driver Apr 29 15:03:04 no, it is mainly for the crappy vfat stuff we have to use to boot from Apr 29 15:03:12 ndec: would be good to fully implement this extension later, will follow that on an email Apr 29 15:03:29 a real redesign or the "loic jiggling"? Apr 29 15:03:42 ogra_: try uboot-as-linux-kernel way? Apr 29 15:03:54 the debian redesign thats going on since a year or so Apr 29 15:04:03 yes, loic works on that too Apr 29 15:04:09 ogra_: nokia n900 devs got that way Apr 29 15:04:11 and i will too Apr 29 15:04:22 hrw, for what ? Apr 29 15:04:33 "echo b >/proc/sysrq-trigger" - best command on panda Apr 29 15:04:35 making bootloaders data driven is an expensive configuration nightmare.. every time it needs new data for each board, there's a new little bit of code that ends up being board specific just to make it boot. I don't think you can encapsulate bootloaders in a generic way and still capture all their useful features. Apr 29 15:04:38 ogra_: for ac100? Apr 29 15:04:44 hrw, why ? Apr 29 15:04:56 i have a pertty good bootloader atm Apr 29 15:05:02 no need to change it Apr 29 15:05:05 ok then Apr 29 15:05:25 my primary concern though was the fact that flash-kernel was not designed to run off kernel hooks, I am glad you brought that up Apr 29 15:05:36 it needs kicking 10 other packages in other arches though Apr 29 15:05:42 well, initramfs-tools upstream did Apr 29 15:06:05 initramfs-tools calling flash-kernel if it exists causes a bunch of weeeeeeeird hacks to be required Apr 29 15:06:11 i wouldnt have changed the way flash-kernel is used atm if upstream would have agreed on my bug suggestion Apr 29 15:06:17 ?? Apr 29 15:06:29 * ogra_ hasnt seen any weird hacks that would require Apr 29 15:07:50 they're in ubiquity.. some sort of "move flash kernel out of the way, then install the right kernel, then put flash kernel back, then install uboot-tools and so on which triggers update-initramfs" hacks to make flash-kernel run only once on install Apr 29 15:07:51 adding a huge amount of complexity that isnt needed imho is not really what i was after when filing my bug Apr 29 15:08:14 that bit will not change at all Apr 29 15:08:35 the flash-kernel-installer wont be touched ... it is for doing the initial setup, nothing else Apr 29 15:09:15 the kernel postinst integration wont change that Apr 29 15:09:46 nor will the redesign of flash-kernel change the need to have an initial setup of the bootloader from the installer Apr 29 15:09:53 if flash-kernel just worked exactly like update-initramfs (taking an updated ramfs and flashing it or prepping it, taking an updated kernel and flashing it or prepping it, update boot script, and a final "commit it all if not done already") so it can be run in stages and only flash individual parts.. and more importantly happen before update-notifier is called so the little red "you need to reboot" icon isn't shown Apr 29 15:10:11 what might change is that it will get as horridly complex as grubs setup is already Apr 29 15:10:14 which relies on that, and the idea that kernel hooks are numbered like udev rules, sysctl rules, and everything else Apr 29 15:10:28 alphabetical, filesystem sorted order is a terrible idea Apr 29 15:10:31 since we will mostlxy focus on moving flash-kernel to be treated similar Apr 29 15:11:23 as long as that doesn't mean it's called zz-flash-kernel... :0 Apr 29 15:11:23 and grub is a complex and ugly beast Apr 29 15:11:43 it will be called whatever is required Apr 29 15:12:00 flogging a dead horse here aren't I Apr 29 15:12:03 we wont change initramfs-tools or the kernel postinst mechanism Apr 29 15:12:33 we will adjust flash-kernel to work with the existing implementations instead Apr 29 15:12:38 no need to change the mechanism just the script names so you can standardize on what order things run without it being reliant on locale sort order and filesystem case sensitivity and all kinds of other possible freakish flags that apply to a directory listing of files Apr 29 15:13:11 the same decision that was made for udev, sysctl, every other rules/hooks/triggers system on Debian and Ubuntu (and every other distro too..) Apr 29 15:13:31 well, then we will follow exactly that sheme Apr 29 15:14:09 lilstevie: pong Apr 29 15:14:11 Hi ! Apr 29 15:15:00 all it takes is like 50-update-initramfs, 90-flash-kernel-commit, 99-update-notifier (90-update-grub for x86...) instead of it having no explicit numerical ordering Apr 29 15:15:16 dcordes: fixed usb :p Apr 29 15:15:24 lilstevie: what was the problem ? Apr 29 15:15:34 and if you need to do it in steps and handle anything between certain standardized kernel hook numbering scheme, you can do it without trying to predict what letter other packages start the names of their hooks with Apr 29 15:15:54 lilstevie: I have been thinking about it a lot, your strange binary init etc Apr 29 15:15:54 Neko, initrd hooks have a builtin dependency system, no need for file ordering Apr 29 15:15:57 sysctl.d has a quite well defined numbering for system things, user things, board specific weirdnesses... Apr 29 15:16:02 dcordes: a param Apr 29 15:16:06 lilstevie: and how it would disabale usb Apr 29 15:16:15 lilstevie: CMDLINE ? Apr 29 15:16:16 ogra, talking about kernel postinst hooks Apr 29 15:16:20 dcordes: bootloader pram sets it Apr 29 15:16:48 and how does it pass it to the kernel? is it kernel commandline ? Apr 29 15:17:12 I thought one of the things that got thought but dropped for Natty was "we want to get rid of initramfs because it's annoying, slows down boot, and AC100 doesn't have enough space for it anyway" Apr 29 15:17:31 dcordes: we have this weird partition that has the params, when the kernel module gets loaded, it loads in usb state Apr 29 15:17:32 ac100 isnt officially supported Apr 29 15:17:41 and we wont get rid of initramfs Apr 29 15:18:08 x86 and ARM already rely on the concept that the bootloader can understand a Linux filesystem and it would be rather weird to have a Linux kernel that could not read its root filesystem Apr 29 15:18:28 ogra_: what does it take to make it officially supported ? Apr 29 15:18:47 wouldn't Toshiba have to ship an official Unlock Android tool so you can root it? :) Apr 29 15:18:56 and then install any OS you like? Apr 29 15:18:56 dcordes, a contract with nvidia with either canonical or linaro i would guess Apr 29 15:19:24 dcordes, i plan to produce an unsupported community image in 11.10 though Apr 29 15:19:31 dcordes: I also found out that the bootloader can boot the baseband, havent figured out how to make it do it though Apr 29 15:19:35 lilstevie: hm I don't know oyur SoC. in qsd8250 we have some cpu parameters we set in the bootloader. that is on a very low level, now kernel involved. I don't understand what role your usb bootloader parameters have Apr 29 15:19:47 lilstevie: what is your SoC anyway ? Apr 29 15:19:55 hummingbird Apr 29 15:19:56 now that we have a proper kernel and many ac100 bits are mainlined Apr 29 15:20:06 ogra_: heh that sounds nice Apr 29 15:20:25 dcordes: the driver takes the param Apr 29 15:20:58 dcordes: but i had to load the module that reads the params Apr 29 15:21:32 lilstevie: ok that's interesting. do you have a place to document all this ? Apr 29 15:21:53 lilstevie: I bet many could benefit from that Apr 29 15:21:58 dcordes: kinda, I am trying to get a new place sorted out Apr 29 15:22:15 I just started a twitter for the project :p Apr 29 15:23:40 yerg, 200kb/s for a torrent download of natty? my life is over :( Apr 29 15:23:56 excuse me while I go and kick the internet in the pants Apr 29 15:24:07 lilstevie: did you try unity-2d on galaxytab yet ? Apr 29 15:24:14 dcordes: yes Apr 29 15:25:36 lilstevie: does it work well? any flickering ? Apr 29 15:26:38 only problem I have with it is when I use gestures too many times my touch driver stops working correctly Apr 29 15:37:18 lilstevie: sounds good Apr 29 15:37:47 unity-2d flickering on htcleo (hd2) seems to be a device specific thing Apr 29 15:37:58 our framebuffer driver has room for improvement Apr 29 15:38:11 my touch driver has room for improvement :p Apr 29 15:39:28 well congrats you have it working at all ! Apr 29 15:40:07 I would be happy to further improvement hd2 multitouch support - with ginn I had some first success in maverick. but now I can't really go on due to the flicker problem Apr 29 15:40:17 I was looking to make use of the working mt in unity.. Apr 29 15:43:36 heh it is why I pushed the update button for natty Apr 29 15:46:29 the problem is there is nothing much I can do about it. personally I lack the capability to improve the driver Apr 29 15:46:44 ah Apr 29 15:46:46 somebody was working on proper driver (xf86-video-msm) for our device Apr 29 15:47:30 nice Apr 29 15:50:07 I even have a kernel binary with it Apr 29 15:51:00 :o Apr 29 23:44:56 do i have to put a kernel image somewhere http accessible for rootstock? or can i use something like file:///? Apr 29 23:48:47 steev_: it'll try to use wget to grab the kernel, so to put it somewhere Apr 29 23:49:05 feel free to hack it and add proper file:// support Apr 30 00:03:09 I had put it somewhere online for my rootstocks Apr 30 00:03:15 I'll be damned if I can remember WHERE Apr 30 00:04:31 steev, http://packages.efikamx.info/pool/main/l/linux-source-2.6.31.14.20-efikamx/ this'll do if you don't mind not being headless until you can drop a new kernel in there Apr 30 00:04:57 I believe the image I gave to wschaub was built with that particular one Apr 30 00:06:02 yeah, if i can get it to compile on natty, i'll just throw it up on my devspace **** ENDING LOGGING AT Sat Apr 30 02:59:58 2011