**** BEGIN LOGGING AT Tue Jun 26 03:00:02 2018 Jun 26 04:09:00 e Jun 26 07:29:09 jkridner: hdmi audio works fine for me at all resolutions I tested: 1920x1080@30 1324x768@60 1280x1024@60 1152x864@75 1024x768@60 800x600@60 Jun 26 07:30:22 there does seem to be a weird issue where discontinuing the audio causes my tv to output weird garbage on its headphone output (but not when its internal speakers are used) Jun 26 07:30:41 so maybe the hdmi framer isn't happy about the audio clocks being halted Jun 26 07:31:27 (that would be a bug in the hdmi framer driver, since it should ensure the framer is okay with this when alsa requests it) Jun 26 08:00:16 hotplugging the display (instead of having it connected at boot) works fine too Jun 26 13:06:11 when doing swupdate, there is a single-copy approach and a double-copy approach. if we assume that a) the new image was already tested sufficiently and known to work, and b) any problems (e.g., a power glitch) at update time can still be handled with the single-copy approach by restarting the install, what advantage does double-copy buy you? Jun 26 13:16:35 neither a) or b) are absolutely failsafe Jun 26 13:17:21 do not forget you have a dumb user whom evolution optimized for maximum damage to all kinds of technology Jun 26 13:19:34 as documented by numerous people trying to let their mobile phone swim in the blue goo of a septic tank Jun 26 13:23:51 The hydraulic press channel doesn't count I guess Jun 26 13:27:58 KotH: how would there be any problem with b) ? (short of bad flash, eg.) Jun 26 13:29:00 yates: never underestimate the stupidity of users Jun 26 13:30:18 yates: besides, how do you ensure there is no power glitch? Jun 26 13:30:46 yates: how do you ensure, the user doesnt pull the battery because he didnt want to update just right now and pressed the wrong button by accident? Jun 26 13:30:50 no, no, single copy handles power glitch - it simple reboots and restarts the install Jun 26 13:31:49 KotH: wrong button: hmm, very possible. in that case he can just attach the old usb thumb drive and install the old system. Jun 26 13:32:32 yates: do you have a way to ensure your users are not plain dumb? Jun 26 13:32:48 yates: ie, do you only ship to certified engineers? Jun 26 13:33:02 our use-case is definitely not as general as Jun 26 13:33:09 "the android-using public" Jun 26 13:33:17 does not need to Jun 26 13:33:41 what if the user shoots the devices with a flare gun? Duh, it will break. Jun 26 13:33:50 some level of intelligence is required. Jun 26 13:34:02 otherwise they can bing the device back to the mfr Jun 26 13:34:33 $company i worked for built a gps tracking device. it's only user interface was a push button and a led. yet 10% of the people failed to correctly use it (aka switch it on and off). Jun 26 13:34:43 10% returned after a year with all kinds of stuff broken Jun 26 13:36:02 you could require an IQ test as a condition of purchase.. :) Jun 26 13:36:16 >= 35 Jun 26 13:36:32 heh Jun 26 13:36:35 i wish we could Jun 26 13:37:16 yates: if you're not updating anything necessary to boot the system to the point of being able to restart the installer, single-copy is fine Jun 26 13:37:47 right - we are planning on keeping SPL/Uboot constant Jun 26 13:37:57 hey zmatt! long time! Jun 26 13:37:59 u-boot directly loads the installer? Jun 26 13:38:08 that sounds rather unusual to me Jun 26 13:38:42 yes, i believe in some solutions they use u-boot, others use grub, etc. Jun 26 13:39:25 they pre-set the environment varialbes u-boot uses to decide how to boot Jun 26 13:39:42 ehm, but all of those load a kernel, not an installer :P Jun 26 13:40:21 KotH: lol Jun 26 13:40:45 yes, the single-copy version of swupdate loads a apecial initd into memory, then runs install scripts. Jun 26 13:41:50 http://sbabic.github.io/swupdate/ Jun 26 13:41:59 right, that combined with using double-copy for the kernel is fine (module the usual headaches of MLC flash, which double-copy doesn't protect against either) Jun 26 13:42:16 *modulo Jun 26 13:45:25 zmatt: they call this the single-copy approach. the double-copy uses two entire kernel/rootfs combinations. Jun 26 13:46:21 yeah, to ensure you always have a working system Jun 26 13:47:23 in single-copy.. if the installer somehow fails for any reason, you're kinda screwed. it's presumably designed such that it shouldn't ever fail, so *obviously* it'll also never fail in practice right? :) Jun 26 15:01:23 zmatt: cool. Jun 26 15:01:41 rcn-ee did some examination on the S/PDIF. Jun 26 15:08:29 and? Jun 26 15:08:36 I know it works on the McASP side, we use it Jun 26 15:13:47 this is a DT fragment of a working setup we use to create a hacky sort-of-AES3 output (professional S/PDIF over twisted pair) using a beaglebone connected to an rs485 driver: https://pastebin.com/p2kgP27w (but note that we used P9.27 (mcasp 0 data 3) instead of P9.28 (mcasp 0 data 2) because it happened to be more convenient) Jun 26 15:15:33 I think the main difference is op-mode = <1>; on the mcasp0 mode Jun 26 15:16:29 plus whatever is necessary to inform the hdmi framer Jun 26 15:17:32 at least, I *hope* none of the tinkering I did on the mcasp driver was for s/pdif output... it was a long time ago since I don't really remember Jun 26 15:17:42 *so I don't Jun 26 15:20:03 ohhh maybe it was Jun 26 15:21:15 I'll send him a mail Jun 26 15:29:44 zmatt: in a single-copy, the installer is the initrd (and associated scripts). the initrd and those scripts don't change during an update, do they? Jun 26 15:30:35 how would I know? Jun 26 15:30:56 testing slack-to-irc bridge Jun 26 15:31:13 grrr... this little bot doesn't act nice. need a better script. Jun 26 15:31:25 jkridner: please no. bridges like that are just too awful Jun 26 15:31:55 trying to bring rcn-ee in here more often. lots of people already using matrix.org/riot.im as well. Jun 26 15:32:22 awful, yes, but how do we get people that need "app-like" access? Jun 26 15:32:29 well he used to be here *all the time* Jun 26 15:32:31 self-select them out? Jun 26 15:32:39 so it's not like rcn is foreign to irc Jun 26 15:32:56 I assumed he stopped coming because it was too much distraction or something Jun 26 15:32:58 @rcn-ee we are talking about you and wishing you'd start an IRC client. ;) Jun 26 15:33:07 your statement "in single-copy.. if the installer somehow fails for any reason, you're kinda screwed." implies you know something about the installer, where it's located, whether or not the installer itself is updated, etc. Does it not> Jun 26 15:33:25 s/>/?/ Jun 26 15:33:34 yates: no, I'm just making a general statement about the tendency for software to not always work Jun 26 15:33:50 sure. and if the sdcard goes out, even a double-copy won't work. Jun 26 15:34:01 or emmc, or whatever.. Jun 26 15:34:26 plus the fact that a system with a partially installed update that didn't complete is usually not in a great state Jun 26 15:34:54 (plus, if the installer couldn't complete, it might not even attempt to boot back into the system at all anymore) Jun 26 15:35:07 that may not matter. Jun 26 15:35:25 I'm just saying it isn't hard to imagine benefits to double-copy. I'm not saying you should definitely use double-copy (we don't) Jun 26 15:35:28 zmatt: i guess that depends on how the conditional boots are organized. Jun 26 15:36:46 yes, but from your description it sounds like it will boot into the installer until it completes. as you said, if a power failure occurs, it is desirable to just restart the installer next time Jun 26 15:37:00 yes, exactly. Jun 26 15:37:09 probably a failed install would act much the same way, unless it has some clever fallback recovery mechanism. Jun 26 15:37:30 but in that case, whatever caused the install to fail probably simply happens again on the next attempt Jun 26 15:37:43 i.e. device is effectively dead Jun 26 15:37:49 multiple power failures?!? Jun 26 15:37:54 no Jun 26 15:38:27 I'm not talking about power failures, but something going wrong with the install itself Jun 26 15:39:01 i can see one scenario: a wrongly-formatted usb drive. then yeah, restarting won't help things. i agree and see that point. Jun 26 15:40:07 ohh, it's expecting the update to be available on an usb stick throughout the install? Jun 26 15:40:27 that probably adds a lot more failure possibilities, ok Jun 26 15:43:10 anyway, I'm not in a position to do a risk analysis for you. it does sound like update failures should be recoverable as long as parts critical for launching the updater (e.g. u-boot, kernel, initrd) use a double-copy update mechanism Jun 26 15:43:46 but it does mean that if something goes wrong, the device may be borked until a successful install has been done Jun 26 15:43:57 unlike double-copy which always leaves the device functional Jun 26 15:44:07 (and lets you roll back if the new version turns out to have problems) Jun 26 15:45:11 you just have to weigh those benefits against the reduction in max image size Jun 26 15:51:05 * tbr had a "fun" failure mode to deal with over the long weekend… failed eMMC, but still reporting itself to the SoC → "brick" Jun 26 15:52:40 nothing you cannot fix with a hot air station and lots of flux Jun 26 15:52:41 after some prolonged prodding and testing, I hot aired the eMMC off the PCB; managed to write the special sauce boot stuff to the existing eMMC; it booted on the first try, surprisingly Jun 26 15:52:50 exactly :) Jun 26 15:53:14 sir is it possible to give internet connection using usb modem to beagleboneblack Jun 26 15:53:29 yes Jun 26 15:53:31 "yes" Jun 26 15:53:44 yes* Jun 26 15:53:50 tbr: you could just have forced boot from sd Jun 26 15:53:56 *) some assembly required Jun 26 15:54:22 tbr: which hot air station do you use? own or makerspace? Jun 26 15:54:30 zmatt: didn't find enough documentation for this S905x based STB and also didn't manage to put it in reset Jun 26 15:54:35 ohhhh Jun 26 15:54:45 sorry I assumed you were talking about a beaglebone Jun 26 15:54:52 yes Jun 26 15:54:55 thinkfat: something made out of chinesium. let me check Jun 26 15:55:04 lol Jun 26 15:55:11 tbr: banggood link appreciated Jun 26 15:55:17 zmatt: no worries, I forgot to mention it wasn't bbb Jun 26 15:56:00 sir is it possible to give internet connection using usb modem to beagleboneblack . if it is yes than how can i make it Jun 26 15:57:07 same as any other linux system Jun 26 15:57:36 thinkfat: Yihua 853D. It's neat as an average all in one.https://www.aliexpress.com/item/Yihua-853D-USB-1A-Mobile-Phone-Repair-Soldering-Station-with-LED-Light/32663668836.html Jun 26 15:58:00 ouf Jun 26 15:58:02 usb Jun 26 15:58:02 thinkfat: if you just need the hot air, then there are ones for like 40$ or such Jun 26 15:58:04 for realz? Jun 26 15:58:17 it's just a fixed 5V supply for that ridiculous light Jun 26 15:58:37 it's china, *everything* has USB Jun 26 15:58:43 sir can u fix my problem Jun 26 15:58:56 nobody is giving to replay to me Jun 26 15:59:23 tbr: I'll have one in 220v color Jun 26 15:59:29 but it's not cheap, eh? Jun 26 15:59:41 thinkfat: Yihua 858/858D is the stand alone Jun 26 15:59:55 40$ ish Jun 26 16:00:14 ah, I see Jun 26 16:00:20 this one has a power supply, too Jun 26 16:00:33 which makes sense for phone repair Jun 26 16:01:43 50ish, including shipping https://www.aliexpress.com/item/YIHUA-858D-220V-Hot-Air-SMD-Rework-Station-Heat-G-un-SMT-Solder-soldering-iron-Welding/32847096616.html Jun 26 16:02:26 while you're at it, get one of these https://www.aliexpress.com/item/Hot-Air-Gun-Station-Heat-Resistant-Insulation-Magnetic-Silicone-Pad-Desk-Mat-Fix-iPhone-BGA-Soldering/32814258841.html ;-) Jun 26 16:04:34 tbr: ah yes Jun 26 16:04:41 tbr: I think I worked with one of these last weekend Jun 26 16:04:53 tbr: it was a disaster, but likely not due to the hotair station Jun 26 16:05:02 hehe Jun 26 16:05:20 tbr: I roasted the pcb pretty badly Jun 26 16:05:49 tbr: tried to reflow some stuff but I likely used the wrong solder paste Jun 26 16:06:00 I don't have much hot air experience (I mostly use it to shrink tubing), but it seems to work somewhat. If I get into SMD rework and expect the parts to survive, I'll have to practice Jun 26 16:06:18 tbr: so, the flux always evaporated before the tin could properly melt Jun 26 16:06:26 yuck Jun 26 16:06:56 from what I saw: gratuitous extra flux seems to be always in order Jun 26 16:07:13 tbr: yes, and just tin the pads, don't use paste Jun 26 16:07:17 I got a hot station from work as a discard. It looks older than dirt, but it should suffice to determine if I will ever figure out hot air. (grin) Jun 26 16:07:38 :) Jun 26 16:08:04 Ragnorok: and it's sure good enough to do heatshrinking ;-D Jun 26 16:08:28 tbr: I wonder if this is already a knock-off of a knock-off Jun 26 16:08:34 tbr: second generation knock-off Jun 26 16:08:39 possible Jun 26 16:09:05 going to check louis rossmans channel ;) Jun 26 16:09:45 btw: Yihua is "bigclivedotcom" 'approved' (aka he also uses one, to do heatshrinking) Jun 26 16:09:53 hah Jun 26 16:10:10 I have little use for heatshrinking equipment Jun 26 16:10:48 I don't do it that often, but when I do then something better than spending 20min looking for matches is appreciated Jun 26 16:13:05 and now it finally proved some of its money worth by making the eMMC pop off Jun 26 17:50:13 Does anyone have experience with OpenCL? It is compatible with the BBB? I'm just looking for a simple program to manipulate matrices, ONLY for learning purposes Jun 26 17:51:40 CoffeeBreakfast: why do that on a BBB ? and why not use something like numpy ? Jun 26 17:52:48 I know that.. It's just theoretical. Nothing for production. I'm a curious person Jun 26 17:53:37 (Also I enjoy learning) Jun 26 17:54:54 google resulted in some mentions of OpenCL for ARM but I didn't see anything concrete and the links I tried to follow were dead Jun 26 17:54:59 the PVR/SGX core on that thing is quite dated. would be surprised if there are (public and free) openCL drivers for it Jun 26 17:55:13 oh you definitely can't use the SGX Jun 26 17:55:50 the only possibility would be to use this "OpenCL™ for ARM NEON™" mentioned in some press release Jun 26 17:56:06 might be in MESA? Jun 26 17:56:21 why would it be there? Jun 26 17:56:32 that sounds nonsensical to me Jun 26 17:57:34 some stackoverflow comment pointed to https://github.com/pocl/pocl Jun 26 17:57:47 https://github.com/ARM-software/ComputeLibrary Jun 26 17:59:02 a "Computer Vision and Machine Learning library" doesn't sound like an openCL implementation Jun 26 17:59:13 sounds more like something that *needs* an openCL implementation Jun 26 18:00:01 ah, optionally uses one Jun 26 18:00:10 regardless, not what CoffeeBreakfast was asking for Jun 26 18:00:41 CoffeeBreakfast: I think if you insist on wanting OpenCL, this pocl thing might be a start Jun 26 18:02:25 zmatt: are there other alternatives for the BBB, apart from numpy? Jun 26 18:02:29 ok, I didn't look that deep. So it's an abstraction above opencl/neon/etc, I guess Jun 26 18:02:40 CoffeeBreakfast: uhh, anything that runs on linux? Jun 26 18:02:52 zmatt: yes Jun 26 18:03:04 that was a rhetorical question, sorry Jun 26 18:03:07 CoffeeBreakfast: uhh, anything that runs on linux. Jun 26 18:04:17 "a simple program to manipulate matrices" is extremely vague. I gave one suggestion. asking "are there other alternatives" sounds like you expect me to enumerate every piece of linux software in existence capable of manipulating matrices. I'll pass on that, thanks Jun 26 18:04:51 zmatt: With the GPU Jun 26 18:05:31 the GPU supports openGL ES 1 and ES 2 (if you put in the effort to get the drivers working) Jun 26 18:06:16 Yeah, sorry if I was vague Jun 26 18:11:54 anything you want to do with the GPU will have to be done through one of those APIs, so using it for general purpose computation is likely to be awkward and not particularly fast Jun 26 18:14:44 I see Jun 26 18:16:43 I don't exclude the possibility that it *might* be possible to use the gpu for *certain* general-purpose computational tasks with useful performance if enough effort is put into it, but I do not get the impression that's what you were after. especially not since you were asking for "simple" :P Jun 26 18:21:15 Are you talking about digging in the AM335x Tech Ref manual? Sure I'll pass if that's the case Jun 26 18:21:55 the TRM is quite readable, but it contains absolutely nothing of use for what you're asking Jun 26 18:22:20 it's OpenGL ES 2 you'll need to dig into Jun 26 18:23:43 Well, sounds reasonable Jun 26 18:24:11 as for getting performance, that might require magic, luck, trial, error, and a lot of persistence Jun 26 18:24:14 :P Jun 26 18:24:32 Persistence always wins in the end. Jun 26 18:24:38 and there are performance analysis tools for the GPU, but I've never made an attempt to get it to work Jun 26 18:24:59 iirc it requires capturing data on the target and then stuffing the output into some analysis tool that only runs on Windows Jun 26 18:26:14 though if it's just for "learning purposes" as you said, performance presumably doesn't matter I guess Jun 26 18:26:29 Yep Jun 26 18:28:16 performance is really tricky because it requires figuring out which things do or don't trigger a "slow path", i.e. it requires getting well versed in the deep lore of the specific driver you're using :P Jun 26 18:28:51 one obstacle for general-purpose computation would be getting data back *out of* the GPU (e.g. glReadPixels() is known to be sloooooooooow) Jun 26 18:32:10 anyway, if you want to get the GPU working: 1. flash a recent GUIless image to the beaglebone (iot or console) to ensure no mesa packages are installed Jun 26 18:32:25 2. sudo apt-get install ti-sgx-ti335x-modules-$(uname -r) Jun 26 18:33:44 3. install the sgx userspace libs. you can try installing the quick and dirty debian packages I built last year and hope they still work: https://liktaanjeneus.nl/sgx/ Jun 26 18:34:12 Maybe I should read about the sgx, or not? Jun 26 18:35:28 if you want to. I'm not sure how much about it is relevant to you Jun 26 18:37:50 the docs that you can find aren't very detailed, and most of it won't matter if you're using it for general computation rather than graphics Jun 26 18:41:04 zmatt you know a lot. Are you embedded systems engineer or something like that? Jun 26 18:41:22 yup Jun 26 18:42:08 humm electrical engineering, with that branch? (I'm guessing) Jun 26 18:43:10 no. although I do know a fair bit about hardware (and e.g. offer advice to our hw engineer when needed or double-check schematics) I'm mostly a software and math guy Jun 26 18:43:41 especially low-level software (although I'll write some javascript code if I really really really have to) Jun 26 18:44:20 low-level software is fun to me Jun 26 18:45:37 you could do it in GLES Jun 26 18:45:49 that will use the SGX Jun 26 18:46:04 that's what I said. ES or ES 2 Jun 26 18:46:09 I think ES2 would make more sense? Jun 26 18:46:55 I mean GPGPU with GLES instead of CL Jun 26 18:47:11 CL is more generalized for computation but.... Jun 26 18:47:26 sure it's possible Jun 26 19:00:48 zmatt: on a device wihtout network connection, how else would the update be loaded?!? Jun 26 19:01:51 yes, the user could yank the usb out. sure. or the usb thumb could be bad. or .. or .. or.. sure. the user could smash the device with a brick, too. Jun 26 19:05:36 it's becoming obvious you're biased towards double-copy solution. Jun 26 19:13:20 woe me, I couldn't read your mind to know the details of your device's situation (which you didn't mention) Jun 26 19:13:37 and I have no such bias. You asked what benefits double-copy might have, so I tried to answer that Jun 26 19:13:43 17:35 < zmatt> I'm just saying it isn't hard to imagine benefits to double-copy. I'm not saying you should definitely use double-copy (we don't) Jun 26 19:14:33 I'm also kinda done spending any further braintime on it Jun 26 19:14:43 yeah, me too. Jun 26 19:15:49 The term "biased" sounds political to me haha Jun 26 22:33:56 hi!! Jun 26 22:38:28 hi to all Jun 26 22:45:01 hi Jun 26 23:03:10 (hi... hi.... hi.... hi... hi...) Jun 26 23:18:31 hi :) Jun 26 23:18:57 i realy need help! Jun 26 23:25:52 then perhaps you might want to consider asking an actual question? merely saying you need help is not enough information for anyone to be able to provide any help, hence in practice you typically will not get any reply at all Jun 26 23:37:24 ok im noob and i read a lot the last days Jun 26 23:45:43 i cant charge the image to the bb Jun 26 23:46:00 i follow all steps Jun 26 23:46:18 maybe somthing wrong whit my bbb? Jun 27 02:34:15 Hello again. I have debian stretch on my desktop PC. I need to compile for the BBB, but Emdebian toolchains are no longer supported, apparently Jun 27 02:47:50 nvm, I got it Jun 27 02:59:36 That book is old. Jun 27 02:59:54 sorry... Jun 27 02:59:59 forget it. **** ENDING LOGGING AT Wed Jun 27 03:00:01 2018