**** BEGIN LOGGING AT Fri Jan 08 02:59:58 2016 Jan 08 07:19:47 HELLO Jan 08 07:20:32 ARE YOU THERE Jan 08 07:24:27 WE ARE HERE, YOU ARE NOT! Jan 08 10:54:08 tbr +1 :D Jan 08 12:58:47 So, I'm still working on having an image shown during the flashing process. I've done some digging and it appears that having a (flashing) script start as the init process halts further kernel execution, including setting up the display. Jan 08 12:59:19 odd .. well, it might be headless. Can't say I've tried. Jan 08 13:01:42 Apparently /lib/systemd/systemd does something that will not halt kernel execution. Simply having a script loop, echo and sleep will halt execution. What am I not doing right? Is "sleep" the culprit stealing all resources? Or is there a signal that needs to be sent? Jan 08 13:02:28 sleep will wait for its duration Jan 08 13:02:39 but if its in a loop... Jan 08 13:04:29 are you trying to make some form of 'graphical' flasher? I'm not sure I see what you're trying to achieve... Jan 08 13:04:54 you should be able to do something simple with the framebuffer if so Jan 08 13:10:22 I'm using RCNs eMMC flasher, but I would like to show a picture on the display during the flashing process. Jan 08 13:11:49 thats vague Jan 08 13:12:04 I'm just showing a png using "fbi". Jan 08 13:12:06 eliasbakken: static image? Jan 08 13:12:10 Yes. Jan 08 13:12:16 yeah that should be trivial Jan 08 13:12:24 perhaps hook up with rcn in here later on .. Jan 08 13:12:34 he might just push it upstream ;) Jan 08 13:13:04 It should, and it used to work with an older version of the flasher, but now it appears that the console comes up after init is starterd. Jan 08 13:13:24 heres something that may also be trivial. i updated my beaglebone black's kernel and it wouldnt boot. this is after flashing the latest debian jesse image on there. What could be causing this? Do I need to update the bootloader? Jan 08 13:14:21 mistawright .. which method .. roll-kernel or apt-get ? Jan 08 13:15:12 Yes, I'll be happy to push it upstream. Ideally I would like to have a progress indicator, but none of this will work without a working console at the time of the boot :) Jan 08 13:15:22 veremit, apt get. scripts in opt Jan 08 13:15:38 mistawright: do you have a serial debug cable? Jan 08 13:15:41 veremit, needed to install real time kernel to use with bbbmini for arducopter Jan 08 13:15:59 veremit, yes i do and I saw nothing after updating the kernel. when attempting to boot Jan 08 13:16:08 mistawright: oh kewl. probably worth pokin rcn too :D perhaps there's an issue Jan 08 13:16:20 mistawright.. no uboot prompt? Jan 08 13:17:04 should always get a uboot :/ unless you've poked it deliberately :) Jan 08 13:17:11 veremit, no but it may be my board. i have feeling this element 14 board is slightly inferior or im just annoyed with it Jan 08 13:17:35 could be both. the ele14 boards should be at least as-good .. they use the same parts. Jan 08 13:17:48 veremit, in the process of installing this kernel linux-image-4.1.15-bone-rt-r17 Jan 08 13:18:20 mistawright: try a uSD with a cleanimage on .. press the Boot on power-up .. that should eliminate the board .. Jan 08 13:18:42 then you should be able to repair the emmc, if so Jan 08 13:19:03 veremit, yeah that was my next step. i was still poking around with another board that has issues booting which ill tackle in a few Jan 08 13:19:21 too many variables .. heh Jan 08 13:24:49 veremit, yeah like rebooting the board with a microsd card in it that flashes it. facepalm Jan 08 13:26:11 mistawright.. the images on the elinux site Are labelled .. ;) label your cards! :D Jan 08 13:28:03 Ah, I think I just might have gotten a clue! It looks like tilcd is compiled as a module. That might be the reason the display is not coming, up, right? Jan 08 13:28:16 i have 4 64gb microsd cards i use for embedded boards. all sandisk ultra plus cards. hard to differentiate so i will need to label soon Jan 08 13:28:31 eliasbakken: that would surprise me, but may be possible on the flasher images... Jan 08 13:28:53 mistawright: I've put label-printer labels on the back 'contact'-side :) Jan 08 13:29:26 eliasbakken: ah, on reflection, the lcd isn't initialised by default in case those IO pins are inuse elsewhere Jan 08 13:29:51 veremit: I'll try to compile it in to the kernel and see if it has an effect. Jan 08 13:30:12 you should be able to add it to /etc/modules.conf or whatever its called Jan 08 13:34:31 veremit: you are right. Since it is an embedded platform, it makes sense to have it as a module. This is a BBW I'm working on now, so even less lightly than with BBB. Jan 08 13:35:28 eliasbakken: ah not the dilemma with the Black .. :) Jan 08 13:36:31 veremit: not really, but I'm seeing the same symptoms with the black.. Jan 08 13:36:39 ah okies Jan 08 13:40:08 veremit: I double checked now, tilcdc is also a module on the black. This is almost certainly the problem. But now for the million dollar question, how do I load the modules? Can I simply modprobe from the flasher script? Jan 08 13:46:56 eliasbakken: just use the standard module-loading mechanism .. I'm sure you can do something in /etc/modules Jan 08 14:08:07 Hi, i'm actually trying to improve the BBB response time for real time purposes. I was wondering what's the "best" value for HZ timer frequency, I just read that ARM arch needed to be set to HZ=100 but i've changed it to 250HZ and after some test the system is better than when it was set to 100... but what's the limit ? how to test it ? any references ? thanks :) Jan 08 15:08:14 I'm gonna work with the bbb and ros. Now I'm searching for info if I should better use ubuntu or armstrong (on my sd-card). Is there something you could recommend? Jan 08 15:31:53 Eva__: the official distro is debian. Historically angstrom (or more recently 'yocto') was the choice for embedded systems, but the Pi/Debian collaboration has kinda opened that up Jan 08 15:34:06 Thanks. So, I should put ubuntu (deb) on my sd-card Jan 08 15:43:17 "Should I use A or B, what do you recommend?" "Actually, C is the preferred choice nowadays" "Ok, so I should use B?" Jan 08 15:49:38 Well, I tought that with debian he referred to ubuntu. Jan 08 15:53:42 debian != ubuntu Jan 08 15:53:48 they are very different distros Jan 08 15:54:01 ubuntu is very desktop-oriented Jan 08 15:54:37 lol Jan 08 15:55:11 yeah, that is why ubuntu is the #1 OS used in cloud setups :P Jan 08 15:55:16 Well I was confussed, because the ROS page only refers to ubuntu arm and armstrong. Because ubuntu is based on debian I though you meant that. Jan 08 15:55:44 ogra_ .. I fail to understand why ubuntu is used for web servers .. its known to be full of security holes Jan 08 15:55:52 huh ? Jan 08 15:55:54 although it does have a LTS policy Jan 08 15:56:08 you heaven't heard about that then .. :) Jan 08 15:56:26 no, it isnt fuller of scurity holes than other distros Jan 08 15:56:47 and ubuntu hanst focused on desktopp development in about 3 ears now Jan 08 15:56:51 *years Jan 08 15:56:58 *hasnt Jan 08 15:57:15 ogra_ I saw a chart recently of vunerabilities, and ubuntu was in the top 10 .. along with windows and ios and android iirc. Jan 08 15:57:17 vfocus is on embedded, could and phones since several years Jan 08 15:57:21 *cloud Jan 08 15:57:52 * ogra_ saw a chart saying it would rainn today ... Jan 08 15:57:58 was all sunny :P Jan 08 15:58:00 me too :/ Jan 08 15:58:06 but its gettin dark now :D Jan 08 15:58:11 yeah Jan 08 15:59:07 Eva__, if you are interested in ROS on ubuntu, talk to kyrofa in #snappy Jan 08 16:29:34 there are ros ubuntu packages.. Jan 08 16:43:04 veremit: i guess ubuntu is popular in "cloud" services because modern devops are not real sysadmins, and don't want to cope with debian :P Jan 08 16:44:07 samael .. I wouldn't run debian either .. but I s'pose debian is deemed 'stable' by bein ancient lol Jan 08 16:44:46 I've seen fedora, centos and bsd being popular though Jan 08 16:45:04 samael.. but I agree with your first point ;P Jan 08 16:46:28 all of them seems valid alternatives to me, given you don't like debian. but i personally fail to understand how you would choose ubuntu over debian for a server Jan 08 16:47:19 ask the companies that do, it obviously pays off well enough to fund ubuntu phone and snappy development Jan 08 16:47:42 anyway, de gustibus :) Jan 08 16:48:16 ogra_: someone must be payin those ubuntu devs .. so I guess it works .. :) Jan 08 16:48:17 ogra_: phones are completely different target. and debian is not really an alternative in this field Jan 08 16:49:02 samael, phones are the next desktop in ubuntu ;) (and snappy is the next plumbing layer) so it isnt *that* far off Jan 08 16:49:27 who's signed up to installing ubuntu instead of Google-Android or iOS or .. *laughs* windoze .. ? Jan 08 16:49:46 my Dad has a windoze phone and hates it .. Jan 08 16:49:53 china mobile ... Jan 08 16:49:55 can't blame him Jan 08 16:50:01 lol ah right. Jan 08 16:50:11 not like .. motorola or anyone ? Jan 08 16:50:46 oh, well, along with chinese manufacturers indeed Jan 08 16:50:59 ogra_: oh i see. in my servers i don't need/use such /things/. of course, ymmv :) Jan 08 16:51:24 everything is made in china/taiwan/hong kong isn't it? Jan 08 16:52:07 well, wikipedia, netflix, uber, most of the bigger movie rendering studios all run ubuntu clusters Jan 08 16:52:24 so i wouldnt say that ubuntu is actually rare in the server market :) Jan 08 16:52:26 netflix run ubuntu?! really? Jan 08 16:52:33 yep, along with bsd Jan 08 16:52:49 ah so their main platform is bsd I should imagine Jan 08 16:53:09 it certainly won't be windows .. not with that level of users Jan 08 16:53:13 i think the frontend is bsd and the streaming backend is ubuntu Jan 08 16:53:16 or some such Jan 08 16:56:13 https://openconnect.netflix.com/software/ Jan 08 16:56:54 interesting they use Amazon though Jan 08 16:57:59 https://insights.ubuntu.com/2015/08/11/how-netflix-tunes-ubuntu-on-ec2/ Jan 08 17:00:37 I saw they used xen .. but I don't see how the ubuntu bit fits in .. unless that's their xen host?! or runs in the 'xen's perhaps Jan 08 18:23:29 hi, if anyone present knows about writing linux drivers? Jan 08 18:25:31 The 'net is full of examples on this, based on a week of searching. There appears to be little BBB specific about it. Jan 08 18:32:47 i can write driver but i have problem with interput timer Jan 08 18:34:15 Ah. Got me. I'm only tipping my toes in that - haven't really done a lot of looking even. Jan 08 19:07:49 Hi everyone. I'm using the Cpru compiler for a bitbanging application. I am having trouble, however, in measuring time. Actually, I need some sort of timer or at least some sort of access to instruction counter. Can anyone share any thoughts on this? Jan 08 19:10:18 sherring: Research cpu cycle counters in the am335x, if you need that level of data. Otherwise the ARM has a number of timers available with example code, which I have only found and not attempted to employ. Jan 08 19:12:10 Ragnorok: I wish there was something in the PRUSS itself, since accessing wide memory is slower. Jan 08 19:12:30 There is. In each of the pair. Jan 08 19:14:40 Ragnorok: Ok, just found it! looked for it yesterday and found none :@ Thanks! Jan 08 19:14:50 One is happy to be of service. Good luck! Jan 08 19:46:37 Hm... Not trivial this whole splash-image-during-flashing business. Even though fbdev is enabled in the kernel, the display is very slow to get enabled. It looks like systemd has a part in setting up the console, is tat right? Jan 08 19:47:23 I highly doubt it Jan 08 19:47:36 console is a kernel thing Jan 08 19:48:25 hence you can read errors/crashes on the console if the kernel is active still Jan 08 19:48:37 usually you can get kernell dumps on it even Jan 08 19:49:26 console != framebuffer Jan 08 19:49:42 although there is such a thing as a framebuffer console of course, but I'd disable that in your case Jan 08 19:49:44 zmatt, +1 like debian != ubuntu lol Jan 08 19:50:33 zmatt, framebuffer should come up immediately at boot surely? Jan 08 19:51:37 not if the tilcdc driver is compiled as module Jan 08 19:51:42 stt_michael, really? Jan 08 19:52:00 eliasbakken: check lsmod and/or kernel config Jan 08 19:52:35 Well, it's the framebuffer that is needed, yes. Jan 08 19:53:31 we established its a module earlier... Jan 08 19:53:42 But it does not appear until some 8 seconds into the boot, while systemd starts at around 5 seconds Jan 08 19:53:46 but I can't remember how you load them... Jan 08 19:53:47 Yeah.. Jan 08 19:54:07 you can do a uboot splash too, eliasbakken Jan 08 19:54:07 also, the framebuffer needs to be initialized Jan 08 19:54:23 Ok, is that not still super hcaky? Jan 08 19:54:25 zmatt, go on ... ;P Jan 08 19:54:30 true, or bypass fbdev and use dri like plymouth does Jan 08 19:54:50 fbdev seems to be headed towards "deprecated" actually Jan 08 19:55:10 though there's another issue Jan 08 19:55:58 since I'm currently trying to get a system to start up with a framebuffer app, and without fbcon it won't init by itself :P Jan 08 19:56:14 eliasbakken, google "u boot splash screen beagleboard" I get about 5 good results at the top Jan 08 19:56:18 head -n 1 /sys/class/graphics/fb0/modes > /sys/class/graphics/fb0/mode Jan 08 19:56:25 zmatt, that's pants Jan 08 19:56:38 that enables the thing, but I'm pretty sure that's not how you're supposed to do it Jan 08 19:56:48 I can't get fbset to cooperate though Jan 08 19:57:20 the mode it reports also doesn't change as a result of aforementioned action Jan 08 19:57:38 not sure what do make of that, other than fbset apparently being useless Jan 08 20:03:30 A u-boot splash screen would definitely solve my problem, even better than the fbi splash solution. I'll look into it some more, from what I read the last time I googled it, it was too hackish, but perhaps that post was old. Jan 08 20:06:07 eliasbakken, quite likely lol Jan 08 20:11:18 the lcd controller is very easy to init and displaying a static image requires no irq handling or any other maintenance Jan 08 20:12:09 I did it in a little baremetal test, though not as a very useful size since I haven't finished my EMIF-init code yet, and 64 KB isn't much space for a framebuffer ;) Jan 08 20:16:18 that'd depend on your resolution Jan 08 20:17:00 I am squeezing a FB in 8K or less Jan 08 20:31:43 Ragnorok: a few hours before, you told me about the clock register for PRU. I tried accessing it over its offset (0xC) and using the clock() c command, but got no luck. Actually, it seems that clock() is reseting the pru(?). Any other thoughts? Jan 08 20:32:31 Hurm. Sadly no. I just know I've read it's there, and seem to vaguely recall finding an example of how to access it. Sorry. Jan 08 20:34:49 ds2: fair enough Jan 08 20:35:56 Well, thanks any way. I cant find the base address for PRU_ICSS_PRU_CTRL Jan 08 22:01:12 rcn-ee, ping? Jan 08 22:01:30 hi stt_michael how's ti going Jan 08 22:01:41 goes good mostly .. tis friday :D Jan 08 22:01:43 hows you? Jan 08 22:02:14 good day to get lots done. ;) (quiet at work) Jan 08 22:02:26 ah perfect .. yeah my boss was away at a meeting today :p Jan 08 22:02:38 but I did get a bit way-laid earlier.. lol Jan 08 22:04:32 rcn-ee, wondered whether you could shed any light on gettnig a uboot splashscreen on the flasher image, or a static image on the framebuffer(Or similar) our friend eliasbakken here has been trying .. Jan 08 22:04:55 rcn-ee, you're too knowledgeable not to ask :D Jan 08 22:05:23 there is a u-boot video driver.. so you could use that. ;) Jan 08 22:05:55 part of the problem, to support all the lcd cape overlay's, they have to be modules, so video comes up late.. Jan 08 22:06:18 rcn-ee, eliasbakken - this was the problem Jan 08 22:06:51 if we get lcd overlays working when it's built-in, it should show up sooner... i'll look again at 4.4/4.5.. Jan 08 22:08:15 i do like how the sunxi guys do it... fb in u-boot, then use kernel's drm 'simplefb', nice smooth u-boot/kernel transition.. Jan 08 22:13:18 rcn-ee: got some lcdc things I'd love your opinion about too... but first train to catch :/ afk Jan 08 22:13:38 safe travels zmatt .. :p Jan 08 22:13:54 rcn-ee, making progress with sunxi .. or just observations ;) Jan 08 22:14:55 stt_michael, just observations... bananapi_m3 works nice now. ;) Jan 08 22:15:05 oooo tell me more .. :p Jan 08 22:15:14 h3? Jan 08 22:15:15 :p Jan 08 22:15:25 a83t core. ;) Jan 08 22:15:36 well one cpu core works.. no mmc... Jan 08 22:15:39 but u-boot works.. Jan 08 22:15:48 oh wrong fruit .. you're a Pi behind me :p Jan 08 22:16:02 I want a lemon Pi next .. or a blueberry pi .. or .. Jan 08 22:16:07 yeah, that orange... Jan 08 22:16:29 I want the whole fruit bowl .. no Apple pi yet either .. :( Jan 08 22:16:59 so have we got a sunxi branch on eewiki now? :p Jan 08 22:17:16 hum, though i pushed the orange pi out... guess not.. Jan 08 22:17:38 stt_michael, right now, they are just in the "netinstall" https://github.com/RobertCNelson/netinstall/tree/master/hwpack Jan 08 22:18:06 but they aren't much different then the olimex a20 guide.. just different config names.. Jan 08 22:19:03 yup .. so that's an easy clone-copy-paste wiki page lol :D Jan 08 22:26:52 rcn-ee, must let me know when you get your orange-pi. The NAND and wifi are a definite plus ;p Jan 08 22:28:44 it's in my hand... Jan 08 22:29:41 doh, and it has the smaller dc power jack that everyone in china uses... i don't have that one at my desk.. Jan 08 22:46:34 yeah which one IS that?! Jan 08 22:46:41 banana pi? Jan 08 22:46:43 I had to find a crazy adapter-bodger for it Jan 08 22:46:47 ORANGE-pi Jan 08 22:46:53 :D Jan 08 22:46:55 ffs.. keep up lol :P jk GenTooMan Jan 08 22:47:42 right .. time to hit the road again .. bbiaf Jan 08 22:47:51 the banana pi is a quad core I considered getting it but I'll work with some BBB's first. Jan 08 22:47:54 rcn-ee, yeah I think if you got a meaty usb supply .. its fine Jan 08 22:48:21 B-Pi is yes, so is the O-Pi, but the O-Pi plus comes with SATA *and* Wifi *and* onboard NAND Jan 08 22:48:30 it -just- trumps the imx6 wand Jan 08 22:48:49 for the price. Jan 08 22:48:59 stt_michael, but the wand can do 3d. ;) Jan 08 22:49:30 rcn-ee, eh ok .. Jan 08 22:49:31 stt_michael, https://www.youtube.com/watch?v=1OJ1sVt0_Jc Jan 08 22:49:39 but for the price .. the O-pi is damn good Jan 08 22:49:58 yeah it is.. Jan 08 22:50:00 its 1/4-1.3 price of the wand .. wtf you want .. make dinner and clean the house?! Jan 08 22:50:06 lol Jan 08 22:50:43 anyhow .. must shut up and get outta here .. I may have started late .. but 11pm is too late to be at office on a friday! Jan 08 22:50:47 bbiab!! :D Jan 08 23:30:09 Hey guys, I have a question, is Beaglebone suitable to control a quadrocopter? Jan 08 23:30:26 I want to write everything from scratch. Jan 09 00:11:23 It's possible to control a quadrocopter with it yse. Jan 09 00:14:28 as for doing everything from scratch. I suggest you do some preliminary experiments you would have a lot to think about. Jan 09 00:14:56 Ok done talking to myself (sigh) Jan 09 00:18:16 tea. Jan 09 00:19:21 tea, yes, good plan .. Jan 09 00:26:40 and I really need to figure out where the hell that weird glitch on the framebuffer is coming from Jan 09 00:27:35 especially since I have trouble imagining it being due to lcdc itself, it's too dumb for that Jan 09 00:29:51 and it's weird that X11 doesn't show it Jan 09 00:40:07 hmm boiling the kettle really would expedite tea .. brb :) **** BEGIN LOGGING AT Sat Jan 09 00:45:38 2016 Jan 09 00:47:05 I do think boiling a kettle sounds like a good idea... I suspect they have a rather high boiling point Jan 09 00:47:10 *do not Jan 09 00:52:37 so you have have display "issues" with the framebuffer driver? Jan 09 00:55:23 yeah, though I now have a different one than before it seems, although they appear related ("before" was using the fbconsole on stock 4.1-ti, I'm now using a custom config and use the framebuffer device directly for graphics, fbcon disabled) Jan 09 00:56:45 issue 1 was that it seemed to scan out from a pointer slightly before the actual framebuffer, i.e. some garbage pixels at the start of the first line and the right side of the display ended up on the left (and one line down) Jan 09 00:57:08 it would jump between this and normal seemingly at the slightest provocation Jan 09 00:57:14 offset always the same Jan 09 00:58:09 different kernels, different monitors and resolutions via hdmi framer, also manifests without hdmi framer (lvds transmitter connected directly) Jan 09 00:59:42 the new phenomenon is that the last column contains wrong pixel data, namely a copy of a column some distance to the left Jan 09 01:00:07 and I think the distance is the same as the offset of the previous glitch Jan 09 01:02:49 the first one should be easily reproducible just by using the framebuffer console... Jan 09 01:06:08 the new glitch really puzzles me though, I don't understand what could possibly affect only a single pixel of each line Jan 09 01:07:13 unless the kernel's compositor (or whatever it is that's e.g. overlaying the cursor) is really screwing up Jan 09 01:07:27 but that would show up on more platforms Jan 09 01:35:43 One pixel offset? Does the frame buffer have an internal offset register to scroll the display? Jan 09 01:35:53 not one pixel Jan 09 01:36:16 a significant offset, few dozen pixels maybe, didn't count them Jan 09 01:36:53 Something might be diddling the display offset register, depends on how the frame buffer is configured. Jan 09 01:37:16 the frame buffer is just a piece of memory, and I'm not even sure whether the kernel overwrites the same memory or supplies a new frame Jan 09 01:38:58 (lcdc reads a contiguous piece of memory and passes it via an async fifo to the transmitter) Jan 09 01:47:33 Is it possible the frame buffer controller is not being configured correctly? I remember something about a base address for the display a display pitch and an offset for scrolling the display. It's been a few years since I looked at it (LOL). I wish I had total recall but as you get older... heck you forget. Jan 09 01:49:37 you just write pointers to start/end of framebuffer to the peripheral Jan 09 01:51:49 the dma controller doesn't operate on pixels or lines but just on words of data or preferably power-of-two bursts thereof Jan 09 01:55:50 that's why the new issue (single column copied to last column) puzzles me: neither of lcdc's components seems plausibly capable of creating the effect; the dma controller doesn't know or care about pixels and the transmitter splits the data stream into pixels but can't plausibly remember anything to make it show up again many pixels later Jan 09 01:57:47 my next plan (after getting sleep) is to check the pointer as stored in the peripheral, if only one or a few buffers in memory are being used, then I can also check them for bogus content Jan 09 02:00:54 Hmmm ... I assume their is a vertical refresh flag or something and a horizontal refresh flag as well. If ya ignore the horizontal one things can get messy. **** ENDING LOGGING AT Sat Jan 09 02:59:58 2016