**** BEGIN LOGGING AT Fri Jul 25 03:00:01 2014 Jul 25 08:15:52 morning Jul 25 08:16:28 morning Abhishek_ Jul 25 08:17:02 vvu: did you spot a diff between the 2 GB and the 4 GB DT files? Jul 25 08:23:40 Abhishek_: dtb is the diff Jul 25 08:23:45 * vvu thinks so Jul 25 08:23:52 waiting for karki_ to run my code Jul 25 08:43:07 vvu? Jul 25 11:39:17 * vvu thinks that everything is so quiet around here today Jul 25 12:56:38 jkridner: ever tested dfu flashing on the BeagleBone Black ? Jul 25 13:14:51 no Jul 25 13:31:18 Abhishek_ ping Jul 25 13:31:51 karki_: pong Jul 25 13:32:11 how do you reload the pru firmware again :p Jul 25 13:33:16 echo 0:pru0,1:pru1 > /sys/devices/ocp.*/*.prurproc/load Jul 25 13:33:51 thanks :) Jul 25 13:55:51 * karki_ throws a brick at his BBB Jul 25 14:01:55 karki_: what happend? Jul 25 14:02:14 weirdest of errors - 9hrs debugging Jul 25 14:02:27 I'm still clueless as to what is happening Jul 25 14:02:35 :'( Jul 25 14:05:43 karki_: hope it works out Jul 25 14:06:09 yeah, I should have been hacking away on a PWM implementation right now! Jul 25 14:08:05 karki_: in case of emergency reflash :) Jul 25 14:08:34 vvu : problem is with the pru this time :p Jul 25 14:09:00 reflash that too :) Jul 25 14:11:42 karki_: wait until you need to debug timer on a MCU :) Jul 25 14:11:43 that is fun Jul 25 14:12:09 vvu : have you debugged the timer on a pru? Jul 25 14:12:19 that will make your day Jul 25 14:13:41 nee, never touched bbb pru Jul 25 14:15:32 vvu : good decision :) Jul 25 14:15:55 told you, when the kernel boots my job is done on the BBB :) Jul 25 14:37:02 jkridner: i don't think integrating the flasher into buildroot is too hard Jul 25 14:37:12 wouldn't think so Jul 25 14:37:18 now i am trying to do the basic make beaglebone_defconfig with buildroot Jul 25 14:37:21 to see what i get :) Jul 25 14:37:53 how does buildroot give me things, give me an MLO, u-boot, zImage, dtb and filesystem ? Jul 25 14:38:01 or just an .img with all of them inside ? Jul 25 14:41:59 MLO, u-boot, zImage, etc. Jul 25 14:42:15 it doesn't do anything Beagle-specific, such as make a disk image. Jul 25 14:43:06 mhmh, so i need to add a few steps afterwards Jul 25 14:43:25 if i switch my method to DFU then i can use MLO, u-boot and so on Jul 25 14:43:54 but for now i am strained with a disk image Jul 25 14:44:13 vvu: http://buildroot-busybox.2317881.n4.nabble.com/RFC-FIT-image-support-td72927.html Jul 25 14:45:30 still not usable for me, i have already all the framework to put the board into flashing mode Jul 25 14:45:40 * jkridner wants a tiny flasher image that can be boot off of USB or uSD and goes and grabs latest uSD image off http://beagleboard.org/latest-images and flashes it onto eMMC. Requires live Ethernet connection tough. :( Jul 25 14:46:12 jkridner: we kinda have that already :) Jul 25 14:46:24 i just need to edit a script and it grabs latest eMMC images :D Jul 25 14:46:45 http://beagleboard.org/updater is a bonescript-based hack... Jul 25 14:46:58 but it shows you how to parse http://beagleboard.org/latest-images Jul 25 14:47:30 it relies on bonescript running already on the target device. Jul 25 14:47:56 simply executes the 'xzcat' and 'dd' operations to the non-running disk (uSD or eMMC) Jul 25 14:48:12 and 'curl'. Jul 25 14:48:15 jkridner: true that, but i want to make it with the premise that the user has junk on the board Jul 25 14:48:19 and 'md5sum' Jul 25 14:48:21 so bricked boards mainly Jul 25 14:48:56 well, having no uSD seems like the main requirement to me.... and avoiding the time it takes to program then boot a uSD. Jul 25 14:49:12 we have that Jul 25 14:49:20 if you have no uSD, then you need to boot from somewhere besides the eMMC, otherwise you could corrupt it and be bricked. Jul 25 14:49:33 but no GUI/download. Jul 25 14:50:02 at this point i have support for flashing eMMC just hook up the board while pressing S2 and run the script Jul 25 14:50:05 it takes care of everything Jul 25 14:50:10 the buildroot benefit is to make a tiny ramdisk image for you to run to do the flashing with 'curl', 'xzcat', 'md5sum' and 'dd' and not a lot else. Jul 25 14:50:14 and it flashes the eMMC Jul 25 14:50:43 script runs on a Linux host to which it is connected? Jul 25 14:50:50 yes Jul 25 14:51:19 look in the BBBlfs repo Jul 25 14:51:38 i do the same as you guys do with the basic images, expose a partition to the linux host Jul 25 14:51:43 and then xzcat and dd on it Jul 25 14:51:53 g_mass_storage from the BBBs kernel Jul 25 14:52:50 not just a partition, you'd need to share the whole disk such that the partitions can be created. Jul 25 14:53:04 yes, i expose whole /dev/mmcblk0 Jul 25 14:53:08 which is eMMC Jul 25 14:54:35 That is a fairly cool method as it turns the BeagleBone into a flash drive and avoids the need for the Beagle to have web access. Just need a good cross-platform solution, with GUI and grabbing the latest images. Jul 25 14:54:50 It is difficult for me to push any solutions to the community that aren't cross-platform. Jul 25 14:55:00 yes, now Linux fairly works Jul 25 14:55:05 still waiting for people to test Jul 25 14:55:09 windows I will look into next Jul 25 14:55:19 and osx in a month or so...don't have a machine Jul 25 14:55:34 if you give me some base content, I can do a post on Adafruit's blog and credit you, but I'm leary of the time to make the content. Jul 25 14:56:03 over the weekend I can make some sample text Jul 25 14:56:23 GUI w/ installer. people are dumb you know. dumb as in they understand what they understand and don't want to understand what you understand. Jul 25 14:56:37 true that... Jul 25 14:56:40 thanks. we'll publicize this a little bit. Jul 25 14:56:53 jkridner: for Linux GUIs what should I use ? Jul 25 14:57:04 i am 90% console and 10% Chrome :) Jul 25 14:57:05 something cross-platform, like Qt. Jul 25 14:57:12 hmmm... Jul 25 14:57:19 a Chrome plug-in would be really nice! Jul 25 14:57:28 that sounds neat Jul 25 14:57:29 that would give interesting cross-platform support. Jul 25 14:57:39 yes but still need the backend Jul 25 14:57:46 with a plug-in, you can get lower-level network access, but I'm not sure about libusb. Jul 25 14:57:49 libusb code is the thing that is driving me crazy Jul 25 14:57:58 u-boot peeps told me to switch to DFU stuff Jul 25 14:58:10 but need to get in contact with the dfu subsystem maintainer Jul 25 14:58:11 http://stackoverflow.com/questions/11736587/chrome-experimental-usb-api-does-not-find-any-device-connected Jul 25 14:58:19 hmm... seems Chrome has a USB api Jul 25 14:58:31 hehe nice :) Jul 25 14:59:02 but still the issue is that AM335x ROM is seens as a network interface and all the kernels take it under their hands, and userspace cannot talk easily to it Jul 25 15:06:03 chromium guys must be interested in supporting DFU from Chrome: http://www.chromium.org/chromium-os/u-boot-porting-guide/2-concepts#TOC-Device-Firmware-Update-DFU- Jul 25 15:06:44 If you could use a Chrome extension for the GUI for now and have another app sitting there for the hardware connection, that is a start. Jul 25 15:07:03 Anyway, I'd say Qt or Chrome for the GUI. Jul 25 15:36:39 jkridner: i will go for qt Jul 25 15:50:08 jkridner: enjoyed embedded.fm this week Jul 25 15:50:16 cool. Jul 25 16:51:55 * Abhishek_ is "Bootstrap"ping the web interface Jul 25 16:52:08 (pun intended) Jul 25 16:52:10 jkridner: even if i try to use chrome usb api we will have same issue with the kernel locking the device Jul 25 16:52:25 and the request permision function from google API is only chromeos specific :( Jul 25 16:52:34 k Jul 25 16:52:48 Abhishek_ : I think the quotes explain the pun ;) Jul 25 16:52:48 usb coding is not so nice after all Jul 25 16:53:07 i will go with qt because i need to do something in qt for ti also :) Jul 25 17:16:25 alexanderhiam: ping Jul 25 17:18:46 rseethamraju: hey Jul 25 17:19:17 found the snapshot pipeline that works!! Jul 25 17:19:23 awesome! Jul 25 17:19:38 so can you do two tees and use the same source for all 3? Jul 25 17:19:41 gst-launch-0.10 v4l2src ! ffmpegcolorspace ! pngenc ! filesink location=test.png Jul 25 17:19:47 yes Jul 25 17:19:54 great! Jul 25 17:20:04 though there’s a warning Jul 25 17:20:05 libv4l2: warning v4l2 mmap buffers still mapped on close() Jul 25 17:21:14 I’m thinkind there must be a similar uncomplicated and small pipeline for recording too s Jul 25 17:21:42 so the output is in .avi Jul 25 17:23:58 the snapshot works for me too, same warning but let's just ignore that :P Jul 25 17:25:18 ok Jul 25 17:36:47 alexanderhiam: they say that Debian has disabled ffmpeg encoders, so install Marillat's package Jul 25 17:47:54 huh Jul 25 18:17:34 you can do an updater purely from the eMMC safely with the caveat of.... Jul 25 18:17:48 ...a failed update could require a uSD to recover Jul 25 18:19:51 i wanna do it with no assumption about uSD, eMMC or whatsoever Jul 25 18:20:00 just blank eMMC and blank uSD Jul 25 18:20:37 that can be done but require some careful planning Jul 25 18:20:46 this is what i am doing now Jul 25 18:20:47 very tiny ramdisk Jul 25 18:21:06 i am shipping to the board a small kernel+small ramdisk Jul 25 18:21:14 and expose the eMMC to the linux host using g_mass_storage Jul 25 18:21:41 and just dd things on the drive Jul 25 18:22:08 this expensive USB stick approach is fine to get something going Jul 25 18:22:18 it is already working :) Jul 25 18:22:27 and i can flash in like ~5mins the full 2gb image Jul 25 18:22:29 are you following up the dd with anything else? Jul 25 18:22:37 following up as in ? Jul 25 18:22:43 checkingf if the data is right ? Jul 25 18:22:51 problem with dd is you assume a fixed size Jul 25 18:23:17 so your 2gb image could waste 2gb on a board w/a 4G flash or even larger Jul 25 18:23:30 i am not checking this.. Jul 25 18:23:34 need to put some support in here Jul 25 18:23:41 if you are doing a ext2/ext3, resize2fs could help Jul 25 18:24:32 to enlarge the partriton right ? Jul 25 18:24:35 but you would need to fix up the partition table Jul 25 18:24:49 no, resize2fs enlarges the fs to the size of the partition Jul 25 18:25:03 but dd will screw with the partition table so you need to fix that up first Jul 25 18:25:13 then get the kernel to reread it....last thing is resize2fs Jul 25 18:25:34 i should do some communication so the board can report its revision Jul 25 18:25:42 dd works today cuz no one is looking too hard at the free space on a 4G board Jul 25 18:25:56 that'd would require you to keep a database Jul 25 18:25:59 it is easier then this Jul 25 18:26:14 at boot up in your ramdisk - ask the device what its real size is. save it Jul 25 18:26:26 hook USB disconnect and undo g_Massstorage Jul 25 18:26:39 rework partition, force a re-read (it is an ioctl) Jul 25 18:26:46 resize2fs Jul 25 18:26:50 flash LED saying you are done. Jul 25 18:27:38 ok, need to read now :) Jul 25 18:27:42 but will ping you Jul 25 18:27:50 :) Jul 25 18:28:01 yeyy have work for the weekend now :) Jul 25 18:28:15 android does the hooking of usb disconnect so you can peek there for hints if needed Jul 25 18:28:41 the ioctl() can be invoked by one of the fdisk programs... something like sfdisk -R or like Jul 25 18:29:01 and reworking partition can be scripted to feed into fdisk Jul 25 18:29:28 so as i understand is get real size at boot, report it to linux host Jul 25 18:29:33 dd the image Jul 25 18:29:45 call all the things to make partition bigger Jul 25 18:29:56 or i am half wrong ? Jul 25 18:31:52 pretty much right Jul 25 18:32:27 goodie, now i need to find out how to get the real size of stuff. in my ramdisk df -h shows me nothing Jul 25 18:32:37 you can cheat on getting the real size - zero out the partition table with dd, then run fdisk. fdisk will do the work to figure out the real size Jul 25 18:32:38 there is something in /dev/ for sure that reports the real size Jul 25 18:32:56 there is an ioctl for it Jul 25 18:33:15 if you cheap, save the old table Jul 25 18:33:32 I suppose you can go further and run it through sed and have fdisk reload that for you at the end Jul 25 18:34:18 so I would do all of this while g_mass_storage exposes the eMMC to the linux host Jul 25 18:34:38 i want to keep to a minimum the "command" communication on g_serial Jul 25 18:34:49 or whatever method I use to tell the BBB to execute something Jul 25 18:35:22 hmmmm that could work..not sure how clean is g_mass_storage's exposure. worth a try Jul 25 18:35:32 until now works like a charm Jul 25 18:35:50 speed is like 6.7MB/s or something like that Jul 25 18:35:57 why shouldn't it be clean ? Jul 25 18:36:14 it may take short cuts in the size stuff Jul 25 18:37:12 until it fails i will keep this method to see what happens Jul 25 18:43:06 ds2: the problem is that I do not have a 4gb board to test things Jul 25 18:43:21 in this case if i make a smaller .img that has 1gb partiton should be the same trick right? Jul 25 18:44:27 how do you unsigned int in PRU C? Jul 25 18:44:35 alexanderhiam ? Jul 25 18:45:25 I mean does it have support Jul 25 18:57:08 so weird. I have a "unsigned int x" Jul 25 18:57:21 x reads 0xF0000000 Jul 25 18:57:52 then int y = (x & 0xFF) Jul 25 18:58:08 yet y holds 0xfffffff0 Jul 25 18:58:17 and not 0xf0 Jul 25 18:58:54 karki_: How does it look like in the disassembly? Jul 25 19:47:41 * karki_ needs classes in unit testing...... Jul 25 19:58:12 karki_: don't wish that Jul 25 21:11:28 Abhishek_: how is nodejs world going? Jul 25 21:23:20 mranostay: this wad yesterday's update: https://www.youtube.com/watch?v=VGhf3b3TneA Jul 25 21:23:25 *was Jul 25 21:24:09 * mranostay bookmarks for later Jul 25 21:24:35 Today did some work on the HTML source and refactoring the js Jul 25 21:24:50 Should go into github tomorrow Jul 25 21:25:13 It's a 1min video btw Jul 25 21:27:31 Gonna do some more Bootstrap tomorrow **** ENDING LOGGING AT Sat Jul 26 03:00:00 2014