| 11:02 | jas-hacks_ | rabeeh: hi, if you get time can you test this debian rootf (has GPU/VPU) on cubox-i . http://jas-hacks.blogspot.co.uk/2013/12/imx6-debian-jessie-gpuvpu-3109-100.html |
| 13:15 | MikeSeth | oh yes |
| 13:15 | MikeSeth | my cubox is shipped weeeeee |
| 13:15 | MikeSeth | let's see how many months it's gonna take to ship from north of the country to Tel Aviv |
| 13:19 | xraxor_ | cool |
| 13:19 | xraxor_ | :) |
| 13:19 | xraxor_ | 2 weeks tops |
| 13:26 | MikeSeth | xraxor_: nah, it'll probably be in my mailbox on Sunday |
| 13:26 | MikeSeth | our post office is bad, but not that bad |
| 13:26 | MikeSeth | HELL YEAH MY FIRST ARM BOX |
| 13:29 | xraxor_ | lol |
| 13:29 | xraxor_ | :) which model u get? |
| 13:30 | xraxor_ | i have the i4Pro and loving it, SW could b better, but it will get there |
| 13:30 | sickness | MikeSeth: even before your android mobile? ;) |
| 13:32 | MikeSeth | sickness: it's a brick, not a box |
| 13:32 | MikeSeth | and to get it to do something useful I'd have to kick android off it |
| 13:33 | sickness | eheh |
| 13:42 | MikeSeth | also I lost like 3 of them :P |
| 15:35 | MikeSeth | rabeeh: I got a shipped mail today! No chance I can pick it up already is there |
| 17:27 | jnettlet | dv_, ping |
| 17:50 | hste | jnettlet: how is it going with kernel? |
| 17:51 | jnettlet | hste, okay. Found the bug that was causing the sdhc card to reset and mount read only. I hacked in support to add the Revision to cpuinfo into the kernel source. |
| 17:52 | jnettlet | right now I am trying to figure out the gstreamer plugins. |
| 17:53 | hste | jnettlet: dv_'s work? |
| 17:53 | jnettlet | I was just starting with the gstreamer 0.10 plugins, but will probably test his next. |
| 17:54 | hste | jnettlet: jas-hacks_ compiled both into his jessie image |
| 17:54 | jnettlet | now I need to test something. It seems as though now that I have added the Revision to the cpuinfo readout I am getting slower gpu benchmarks. I need to remove that again to see if they go up. |
| 17:57 | hste | maybe the vpu use up some more resources when firmware gets loaded |
| 17:58 | jnettlet | or there is code in the gpu binaries that turns stuff off when it detects a slower cpu revision |
| 17:58 | hste | you could try just removing the firmware and test |
| 17:59 | jnettlet | I just reverted my change to add the Revision id |
| 18:00 | jnettlet | nah something got changed that is slowing everything down. Xorg is using way too much cpu usage now. |
| 18:01 | hste | yes the cpu have different spped on some revisions. Mine imx6q rev 1.1 only gets 800Mhz if I not hardcode it and not 1Ghz |
| 18:02 | hste | but I hardcode it to 1.2Ghz instead :) |
| 18:06 | jnettlet | yep something is using that to do something. about a 10% difference in the glmark2 score |
| 19:13 | MikeSeth | is there -any- firmware stored in cubox itself? |
| 19:23 | rabeeh | hste: i'm trying the debian image; any idea if the boot console is on the hdmi it hangs lightdm? |
| 19:23 | rabeeh | MikeSeth: the only permanent storage inside the cubox-i is only the chip bootrom (96KB ROM inside the chip itself) |
| 19:24 | rabeeh | everything else is on the micro SD |
| 19:54 | MikeSeth | rabeeh: I am so excited, I got my shipped mail today |
| 20:49 | hste | rabeeh: try dpkg-reconfigure lightdm |
| 20:50 | hste | rabeeh: you can switch to console with alt f1 |
| 21:09 | rabeeh | hste / jas-hacks_: something weird with it |
| 21:09 | hste | rabeeh: whats the problem? |
| 21:10 | rabeeh | putting - 'T0:23:respawn:/sbin/getty -L /dev/ttymxc0 115200 vt100' at /etc/inittab |
| 21:10 | rabeeh | and booting without 'console=ttymxc0,115200' makes it hang in boot |
| 21:11 | rabeeh | debian |
| 21:11 | hste | rabeeh: did u get lightdm up? |
| 21:11 | rabeeh | yes |
| 21:11 | rabeeh | without the 'L0:23...' it boots fine and i get lightdm without issues |
| 21:12 | rabeeh | btw - notice that there is no need for bootcmd.scr script at all |
| 21:12 | rabeeh | all what is needed is a uEnv.txt file that has - |
| 21:12 | rabeeh | bootfile=/boot/uImage |
| 21:12 | rabeeh | mmcargs=setenv bootargs root=/dev/mmcblk0p1 rootwait video=mxcfb0:dev=hdmi consoleblank=0 |
| 21:15 | jas-hacks_ | rabeeh: the hang will occur because it's done in inittab and getty must be blocking |
| 21:15 | rabeeh | agetty then? |
| 21:15 | jas-hacks_ | try it |
| 21:18 | jas-hacks_ | there is another way which I used in the xubuntu build which was to detect if the serial port is present however that effects the shutdown. The console output is not always shown on shutdown. |
| 21:20 | jas-hacks_ | I had some problems trying to run the VPU stuff on the C1 doesn't seem to be enough memory with a X desktop running |
| 21:20 | rabee | 21:20 * rabeeh reflashes |
| 21:21 | rabeeh | jas-hacks_: do you have gpumem=64M on C1? |
| 21:21 | rabeeh | s/C1/HB |
| 21:21 | jas-hacks_ | 64M seem to affect X |
| 21:22 | hste | I used gpumem=32M on mine and that run ok, but got the ipu locking when running glmark2 |
| 21:23 | rabeeh | jas-hacks_: is it reproducible? how can i check? |
| 21:26 | jas-hacks_ | rabeeh: Have to retest on the C1 when I get a chance |
| 21:27 | rabeeh | one other item; any suggestion for Windows users how to resize ext4 partition? :) |
| 21:27 | rabeeh | currently the images i'm spitting out for either 2GB or 4GB micro SD cards; that for Linux it's easy to resize |
| 21:28 | rabeeh | although lots of boards uses raw image flashing utilities; but none has a good recipe for resizing the partitions after flashing the raw image |
| 21:29 | jas-hacks_ | not sure if it is possible in cygwin |
| 21:36 | rabeeh | jas-hacks_: agetty doesn't work too |
| 21:36 | jas-hacks_ | rabeeh: will look at a solution |
| 21:39 | rabeeh | jas-hacks_: using 'xz -9' the image shrunk fro ~695MB to ~407MB |
| 21:40 | rabeeh | i mean your tarball to my 2GByte raw image |
| 21:40 | jas-hacks_ | just need some script to check if serial port is in /dev before launching getty |
| 21:40 | rabeeh | but what will the script do? |
| 21:41 | jas-hacks_ | only launch getty if port is present in /dev/tty... |
| 21:41 | rabeeh | launching getty to ttymxc0 from /etc/inittab should work |
| 21:43 | rabeeh | jas-hacks_: another item; i was willing to write a script on my linux x86 machine to prepare an ubuntu /debian tarball from scratch |
| 21:43 | hste | jas-hacks_: I did put a launcher for owl-video /usr/share/applications/owl.desktop after putting the video-player.sh in /usr/local/bin http://stende.no-ip.info/files/owl.desktop |
| 21:44 | rabeeh | is there anything special besides debootstrapping, installing xubuntu-desktop and then hacking the gpu to force it to install? |
| 21:46 | hste | there is some problems when installing som debs with debootstrap. |
| 21:47 | jas-hacks_ | rabeeh: you would have to cross compile some of the VPU stuff and patch the gstreamer code |
| 21:47 | rabeeh | why cross compiling when you can native compile? |
| 21:48 | jas-hacks_ | I thought you were building everything on x86 |
| 21:48 | hste | so its easier to make a minimal image on pc and have a script you start on first boot that do the rest of installation thereon the device |
| 21:49 | rabeeh | you can build with qemu-arm-static |
| 21:50 | rabeeh | so the idea is getting a base image; debootstrap via qemu-arm-static |
| 21:50 | wbx | the rpi people use raspi-config to expand the rootfs after boot. |
| 21:50 | rabeeh | then 'apt-get install' again using qemu-arm-static; and then native build |
| 21:51 | jas-hacks_ | I start with a minimal console rootfs and then build on the target which is similar to what your doing |
| 21:52 | jas-hacks_ | Although I found some of the packages are broken for arm builds which is where the problems start |
| 21:52 | rabeeh | jas-hacks_: but your way is all done manually and not reproducible |
| 21:52 | hste | rabeeh: I had a script doing most of it http://stende.no-ip.info/files/buildgk802.sh |
| 21:53 | rabeeh | hste: this runs natively on gk802 i assume? |
| 21:53 | hste | this is for installing on a pc |
| 21:54 | hste | a minimal system |
| 21:54 | hste | have to change kernel and uboot |
| 21:54 | jas-hacks_ | rabeeh: The problem is that you need to go through the whole steps at least N number of times as nothing works immediately |
| 21:54 | rabeeh | jas-hacks_: yeah; i'v been there :) |
| 21:55 | hste | :) |
| 21:55 | rabeeh | i know the grief of debootstapping again and again |
| 21:55 | rabeeh | then intalling 1GByte of binaries again and again :) |
| 21:55 | rabeeh | but at the end of the day; it's a perfect example for lots of other distros |
| 21:55 | jas-hacks_ | then something breaks upstreams |
| 21:57 | jas-hacks_ | maybe the way forward would be to override our deb packages with the upstream versions |
| 22:00 | rabeeh | hste: where have you been hiding this script :) |
| 22:00 | hste | its not easy to make a script as the closed stuff change and userspace have to match kernel |
| 22:00 | hste | rabeeh: have had it there for long time |
| 22:00 | rabeeh | if you rerun what are the things the typically break? |
| 22:01 | rabeeh | gpu mesa crap? |
| 22:02 | jas-hacks_ | the gpu stuff breaks with any changes to upstream mesa changes + xorg |
| 22:02 | _rmk_ | I did keep saying that "/dev/sd" should not be used to mean "sd card" |
| 22:03 | _rmk_ | I guess you've pulled the trigger on the gun you keep loading |
| 22:03 | hste | yes I have made some better scripts on that part http://stende.no-ip.info/files/mkubuntu.sh. havn't modified the buildscript |
| 22:03 | jas-hacks_ | for mesa we can script it to move the mesa libs and recreate the symlinks |
| 22:06 | jas-hacks_ | rabeeh: the main problem happens when you do 'apt-get upgrade' |
| 22:07 | hste | and I think it could be ok to keep the imx6 libs in one place like /usr/lib/imx6 and not in /usr/lib |
| 22:08 | hste | could be solved by holding back packages from upgrade |
| 22:09 | jas-hacks_ | hste: thats ok, but if you want to compile any package on the target device then you might need the mesa headers back |
| 22:10 | hste | yes but safer for regular users |
| 22:27 | rabeeh | jas-hacks_ / hste: is there a way to rebuild mesa packages to natively include the vivante headers? |
| 22:27 | rabeeh | i mean this sounds the most reasonable way to do it |
| 22:28 | rabeeh | what do you think? |
| 22:29 | jas-hacks_ | may be possible, however the vivante headers seem to be older than the mesa ones |
| 23:36 | obinou_ | xraxor_: hi ! Thanks for the geexbox demo . I'm going to get my cubox really soon now, and I want to experiment XBMC on it, with some required features |
| 23:37 | obinou_ | (like recording, and DVB decoding + DVB record+play feature) |
| 23:40 | xraxor | cool |
| 23:41 | xraxor | its really fast HW you will love it :D |
| 23:52 | rabeeh | hste - http://imx.solid-run.com/forums/viewtopic.php?f=2&t=326 |
| 23:55 | obinou_ | xraxor_: I'm looking for timeshifting , DVB and recording capabilities, basically. |
| 23:56 | obinou_ | I already own multiple imx6-based gadgets (gk802, sabre-lite, and udoo |
| 23:56 | obinou_ | so I know what to expect |
| 23:57 | obinou_ | I'm kind of worry about the thermal aspects: I already fried a gk802 , and I have thermal issue on other projects with imx6 as well |
| 23:57 | rabeeh | obinou_: like what? |
| 23:58 | rabeeh | what are other imx6 that has thermal issue? |