**** BEGIN LOGGING AT Fri Dec 07 03:00:02 2018 Dec 07 03:57:40 hello Dec 07 03:58:00 may i get your email address for an enquiry? Dec 07 04:37:51 HELP ECCN Dec 07 04:56:00 FAIL Dec 07 04:57:15 I hope noone just died as a result of not being to get the ECCN in time Dec 07 19:20:45 hi Dec 07 19:28:50 jakke: hi Dec 07 19:42:14 hi Dec 07 19:42:18 I'm new to this Dec 07 19:42:34 but it seems like my BBGW is always running from the onboard memory Dec 07 19:42:52 can I have it to run from a SD card continuously? Dec 07 19:43:27 if a bootable sd card is present, it will boot from that Dec 07 19:44:00 if you want to be even more certain, just wipe eMMC Dec 07 19:44:05 yeah, that's what I thought Dec 07 19:44:20 but I tried to install Home assistant for example Dec 07 19:44:35 and I instantly ran out of space Dec 07 19:44:50 how do I wipe the eMMC? Dec 07 19:44:57 the default image has very little free space because it includes a desktop environment Dec 07 19:45:11 if you have no need for that, reflashing it with the iot image is a good idea Dec 07 19:45:43 ok, sure that is a good Idea, but now I have a 32GB sd card inserted in the BBGW Dec 07 19:45:54 would make even more sense to run the OS from that, right? Dec 07 19:46:04 sure, if space is more important than performance Dec 07 19:46:39 mm ok, i get your point Dec 07 19:47:01 I don't know what "Home assistant" is or how much space it requires Dec 07 19:47:32 the iot image probably leaves somewhere between 1 and 2 GB of free space, haven't checked recently Dec 07 19:47:39 a few GB probably, but now I want to go back to trying out the sensors on my cape Dec 07 19:47:51 a few GB? o.O Dec 07 19:47:56 so how does one wipe the eMMC and install BeagleBone again? Dec 07 19:48:15 I have flashed the SD card with the latest BeagleBone image Dec 07 19:48:21 first make a bootable sd card and boot from that Dec 07 19:49:38 then once you've confirmed you booted from sd card (check that "findmnt /" shows /dev/mmcblk0p1 as SOURCE), you can wipe eMMC using: sudo blkdiscard /dev/mmcblk1 Dec 07 19:50:13 nice, I'm flashing the sd card again just to be sure :) Dec 07 19:50:21 also, there's a good chance that the image hasn't been expanded to span your whole SD card. this isn't hard to fix though Dec 07 19:53:32 https://pastebin.com/ThmLM0pX Dec 07 19:53:51 oh the resize2fs was missing sudo, fixed. Dec 07 19:54:41 and what does this do? Dec 07 19:54:50 span the whole card? what does it mean Dec 07 19:55:28 like, if you just download the 4GB image and flash it onto an sd card, the partition (and the filesystem it contains) will only be 4 GB, regardless of how big the card is Dec 07 19:55:39 ah ok Dec 07 19:55:44 the rest of the card will be free unpartitioned space Dec 07 19:56:04 yeah, makes sense Dec 07 19:56:56 the paste shows how to 1. expand the partition, 2. expand the filesystem Dec 07 19:57:17 both of these can be done even if you've booted from sd card, but in that case you need to reboot in between the two steps Dec 07 19:58:08 how much worse is the performance when running the os from the sd card? Dec 07 19:58:18 if you've not booted from the sd card then you can do the steps right after each other. (if you're doing this on a linux system other than the beaglebone you'll need to adjust /dev/mmcblk0 and /dev/mmcblk0p1 to the appropriate devices of course) Dec 07 19:58:40 I mean it would be pretty sweet to have 32 gb to run apps on, like "Home assistant" for home automatization Dec 07 20:00:12 the eMMC interface is twice as fast as the SD card interface. I think reading data typically saturates the interface, hence you'd see an actual factor 2 in performance difference. write performance is a different matter, and will probably also wildly depend on the card Dec 07 20:00:38 sd cards also tend to have shorter lifetime than eMMC Dec 07 20:01:50 (but this may be offset by the larger size) Dec 07 20:02:06 true, not all blocks will be corrupted Dec 07 20:02:16 looks like i'm on /dev/mmcblk0p1 from the boot at least Dec 07 20:03:03 well that too, but I meant if you're only using 4GB of space on a larger card (which is required to be able to make a fair comparison in the first place) then wear leveling can distribute writes across much more physical blocks Dec 07 20:03:52 yes, seems reasonable Dec 07 20:05:30 it would be nice if SD cards, which tend to be TLC (3 bits/cell), allowed you to reconfigure them to MLC (2 bits/cell) or SLC (1 bit/cell) to sacrifice some space in exchange for lifetime and durability. especially since each step results in about a factor 10 in lifetime Dec 07 20:06:00 true, each card is 10 dollars though Dec 07 20:06:26 I have one more question here, once I get a sweet config set up, can I make a backup of the OS somehow? Dec 07 20:06:37 so that I don't have to configure everything each time? Dec 07 20:06:43 the time you end up spending debugging mysterious problems when the card is starting to fail might be worth more than $10 though Dec 07 20:07:04 just copying the entire contents of the card is the easy solution Dec 07 20:07:52 will be hard on my macintosh Dec 07 20:08:02 why? Dec 07 20:08:15 it's not reading the file format Dec 07 20:08:24 doesn't matter when making an image for the whole card Dec 07 20:08:31 mhm Dec 07 20:09:02 so I make an .img and then flash IT to the sd card whenever I need it in the future? Dec 07 20:09:33 yeah, you can use disk utility to make an image Dec 07 20:09:37 yes Dec 07 20:09:41 ok, thanks :) Dec 07 20:10:08 How do I connect to the graphical interface on the BBGW? Dec 07 20:10:12 only downside is that you'll get a 32GB file regardless of how much space is actually used on it Dec 07 20:10:19 seems like a mystery to me Dec 07 20:10:35 the BBGW has no video output Dec 07 20:11:10 (and the recommended "iot" image has no graphical interface. the "lxqt" image does, but I personally wouldn't want to use it) Dec 07 20:11:30 i flashed the one with graphical interface Dec 07 20:11:45 so it's not possible to connect to it, with VNC for example? Dec 07 20:12:05 it's probably possible to install a VNC server Dec 07 20:12:33 do you really require it? Dec 07 20:13:03 probably not 😃 Dec 07 20:13:13 a desktop environment is a rather heavy thing to run on a beaglebone Dec 07 20:13:48 based on a few seconds of googling, it seems Home Assistant uses a web interface Dec 07 20:14:01 it does Dec 07 20:14:11 I'm used to GUIs I guess Dec 07 20:14:36 so my suggestion would be to use the iot image (or possibly even the console image), install home assistant, and connect to it using the web browser on your own computer Dec 07 20:14:48 yeah, right Dec 07 20:15:12 having a desktop environment on the beaglebone would just be a waste of space and cpu power Dec 07 20:16:45 mm ok Dec 07 20:16:59 I tried to fit the wifi antenna on the board, but it's a bitch to get there Dec 07 20:17:10 is it supposed to be "that hard" to fit? Dec 07 20:17:12 ?? Dec 07 20:17:50 I don't have a BBGW. I do have two BBBWs, and they shipped with antennae attached Dec 07 20:18:15 https://beagleboard.org/static/images/BBGW_vertical.png Dec 07 20:18:30 there's some extension to put on Dec 07 20:18:42 there is also wifi onboard by default Dec 07 20:20:18 actually no, if it's designed for external antenna then the on-board one won't work (and I have no idea why they'd even place it). it looks like the selection is made by placing a 0 ohm resistor or such Dec 07 20:20:42 (switching between two antenna options on the fly is pretty tricky) Dec 07 20:21:25 ehh wth Dec 07 20:21:38 how... but... Dec 07 20:21:43 wtf Dec 07 20:22:23 they just placed the on-board antenna in parallel with the coax connector for the external one? Dec 07 20:22:32 yes Dec 07 20:22:50 there's no way in hell that can give a decent result Dec 07 20:24:25 ok 😃 we'll see, I can't even fit it on the board Dec 07 20:24:32 lol Dec 07 20:24:38 I'm even more curious about the other sensors Dec 07 20:24:43 like the light sensor Dec 07 20:25:07 so far i've only been able to set up the IOT Icebreaker project on GCP Dec 07 20:25:14 well, my (already very low) regard for the green-wireless has just sunk even lower Dec 07 20:25:32 sensors? light sensor? on the green wireless? Dec 07 20:25:40 * zmatt examines schematic more carefully Dec 07 20:26:07 mm, light, buzzer, LED, etc¨ Dec 07 20:26:11 https://github.com/GoogleCloudPlatform/IoT-Icebreaker Dec 07 20:26:20 works pretty shitty :) Dec 07 20:26:33 oh external sensors Dec 07 20:27:00 right, there is a cape to it, where you can fit the sensors Dec 07 20:29:22 any suggestions on fun project with the BB? Dec 07 20:30:09 ehh, not sure how to answer that. you probably do things you consider fun, not things I consider fun ;) Dec 07 20:30:13 *you should Dec 07 20:30:56 suggestions on projects then? :) Dec 07 20:31:07 I have no clue on what to do Dec 07 20:31:31 have any interest in network services? Dec 07 20:31:34 I guess I can kinda understand how these grove-based modules could be nice for initial experiments. it's a bit unfortunate however that once you outgrow them you're left with the worst variant of the beaglebone family Dec 07 20:31:45 jakke: you didn't have any specific goal when you acquired this in the first place? Dec 07 20:34:54 I just got certified at GCP, so I wanted to try out IoT a bit, which is why I got the BBGW Dec 07 20:36:05 so, the basic recipe for IoT is: you take a Thing, and then connect it to the Internet ;) Dec 07 20:36:22 (optional step 3: your device gets owned and used as part of a botnet) Dec 07 20:36:57 right, I got some light bulbs from IKEA as well, the wireless ones Dec 07 20:37:19 maybe I'll connect them, and then turn it into a botnet during 2019 Dec 07 20:39:42 on a random note, if you've never read about "Internet Census 2012", it's a fun read Dec 07 20:40:57 (a huge dataset gathered by an unknown individual using a benevolent botnet consisting entirely of devices that had an open internet-facing telnet (!) port with trivial credentials, e.g. root/root or admin/admin) Dec 07 20:41:51 benevolent is maybe the wrong word... but, carefully designed to be non-harmful (and actually try to keep out a harmful botnet that was doing the same thing) Dec 07 20:43:39 yeah, ive heard about that Dec 07 20:44:07 also that there is a viruz which destroys IoT devices with insecure setups Dec 07 20:44:12 fun stuff Dec 07 20:44:14 I liked the quote: As a rule of thumb, if you believe that "nobody would connect that to the Internet, really nobody", there are at least 1000 people who did. Dec 07 20:44:26 I think it's fun to put a honeypot out there and see how long it takes to get compromised. Dec 07 20:44:43 after all, you can instantly re-image. Dec 07 20:44:54 what is the benefit? Dec 07 20:45:19 becoming cognizant of network surroundings in terms of tools and techniques. Dec 07 20:45:25 what goes on? Dec 07 20:45:31 water we talkin about Dec 07 20:46:00 IoT (Idiots on TCP, or Internet of Terror) Dec 07 20:46:13 oh fun Dec 07 20:46:13 lol :) Yep...fun to play. Dec 07 20:46:56 uh, why? why would we ever discuss IoT in any of its acronym-definitions? Dec 07 20:47:00 its poison Dec 07 20:47:05 I put a pi with a weak password in my DMZ and allow free access to it. Dec 07 20:47:18 My test with FreeBSD@R-Pi was boring, not even Ken Thompson showed up Dec 07 20:47:21 ...why... Dec 07 20:47:25 usually in less than 30 minutes, if that, it's a goner. Dec 07 20:47:38 oh, as a honeypot/test Dec 07 20:47:40 ok Dec 07 20:48:24 yeah, thats about what you'd expect, especially considering that port 22 on basically every ip on the internet is swept once every minute. Dec 07 20:48:27 put a text file on it called "creditcard_numbers.txt" Dec 07 20:48:45 when I had a beaglebone on a network where the DHCP server actually gave you a _public_ IP (unfirewalled), I made sure as hell to only allow ssh login using public-key authentication :P Dec 07 20:49:55 zmatt ugh, I went one step farther. My home network's link is gigabit, and my IP is (mostly) static, so i figured i'd want to set up a remote link capability Dec 07 20:49:57 (also the rare occasion where I enabled a firewall on my laptop... probably on the beaglebone too) Dec 07 20:49:59 mention even one good thing to do with the beaglebone Dec 07 20:50:35 but, i went very very very paranoid, mainly need hardware-backed 2fa to authenticate to my server via ssh Dec 07 20:50:53 i keep it down most of the time t Dec 07 20:50:54 jakke: we integrate them into our products Dec 07 20:51:17 *though, because im still terrified of the possibility that i missed a setting while hardening Dec 07 20:51:20 like what, your fridge? Dec 07 20:51:21 (plugged into a custom base board) Dec 07 20:51:38 no, we make speakers (https://dutchdutch.com/) Dec 07 20:51:56 jakke are you here to try and like, claim its bad? because no its not a great computer but as for an SBC dev board its pretty great Dec 07 20:52:23 i use it for tinkering/hacking projects before i move to a lower-level MCU or FPGA based solution Dec 07 20:52:32 it programs and controls the DSP, monitors various stuff, controls LEDs and fan, has a web-interface, etc Dec 07 20:52:54 its nice to have something simple and easy to work with for doing quick poking at all of your interfacing devices Dec 07 20:53:25 much better than the RPi, imo. Dec 07 20:54:16 jakke: fridge-iot? good idea... it sends you emails when you're running out of beer Dec 07 20:54:17 ok, point taken, I'm not saying it's bad, Dec 07 20:54:28 I want to have ideas on what to do Dec 07 20:54:42 with it? so you have the board and nothing to do with it? Dec 07 20:54:56 yeah, or when the wife opens the door, i get a text Dec 07 20:55:09 hedgeberg: yeah :) Dec 07 20:55:55 huh, i mean i try to not get systems unless I have a project that needs them at this point because i have 2.5 million dev boards and haven't done anyting with at least a quarter of them Dec 07 20:56:35 it's for tinkering with, I guess, and learn stuff Dec 07 20:56:44 I mean, you could play with developing in baremetal, which imo is a great learning experience Dec 07 20:56:51 largely you can use the beaglebone for similar purposes as other embedded systems (like the rpi). some of the things that makes the beaglebone stand out are: 1. it has a lot of GPIO and a large variety of interfaces 2. its PRU subsystem is awesome 3. the hardware is open and the components are available for purchase in low quantity, so you can create your own custom version or derivative if you want Dec 07 20:57:15 its not the best platform for that since its got a big SoC in the middle, but its a lot better than the RPi Dec 07 20:57:18 4. the SoC has excellent public documentation (~5000-page technical reference manual) Dec 07 20:57:21 ^ Dec 07 20:57:51 doing dev for an SoC in general can be really complicated, so having a good set of documentation is really, really amazing Dec 07 20:58:32 but uh, if you just want simple tinkering projects? i hear wirepython has good gpio support Dec 07 20:58:51 in contrast for example, the rpi has no full schematic available (afaik?), its SoC is not available for purchase (it doesn't even have a product page) and is very poorly documented Dec 07 20:59:02 zmatt not just that Dec 07 20:59:14 the blobs needed to achieve boot are closed source proprietary Dec 07 20:59:27 so baremetal on the rpi2 and rpi3 is really really complicated Dec 07 20:59:32 also it runs /hot/ Dec 07 21:00:18 jakke got a 3d printer? if so you can run octoprint on the beaglebone as a way to make it remotely controlled, which is pretty fun imo Dec 07 21:00:20 hedgeberg: the blob running on the videocore boot cpu you mean? I don't really consider that a big deal. I can see it as just an extension of bootrom Dec 07 21:00:56 ok, sounds like a good idea, with the printer Dec 07 21:00:58 i mean i guess, but id rather bootroms be OSS too, to be fair Dec 07 21:01:24 hedgeberg: yeah well, it isn't on the am335x either Dec 07 21:01:29 I need to get into the linux flavour first, it's kind of a steep curve in the beginning Dec 07 21:01:37 jakke yeah, octoprint is great. gives you a nice web UI and everything Dec 07 21:01:44 do you not do any linux? Dec 07 21:01:49 because its just debian really Dec 07 21:02:25 it's debian, but still, as soon as there are errors, it's just a hassle to figure it out Dec 07 21:03:01 I mean, look at this Dec 07 21:03:02 https://pastebin.com/RXxcEvfS Dec 07 21:03:20 and zmatt yeah ik but i wish it was. i really like truly OSS projects, makes debugging way easier. also, bootroms are a huge security nightmare and being open source at least makes it more transparent Dec 07 21:03:45 aw, pastebin is blocked here??? Dec 07 21:03:47 come on Dec 07 21:03:58 jakke: lol, very informative Dec 07 21:04:06 i dont want to deal with getting around the proxy/firewall/domain filter Dec 07 21:04:23 jakke: is there not any more context than this? Dec 07 21:04:26 it's a pre-made script, which runs, and then screws me over on the last part Dec 07 21:04:41 curl -O https://raw.githubusercontent.com/home-assistant/fabric-home-assistant/master/hass_rpi_installer.sh && sudo chown pi:pi hass_rpi_installer.sh && bash hass_rpi_installer.sh Dec 07 21:04:43 what does the script do? what is it from? Dec 07 21:04:47 (also, shame on them for still using python2 instead of python3) Dec 07 21:04:57 right??? Dec 07 21:05:11 zmatt gnuradio still uses python2 and it drives me up a freaking wall Dec 07 21:05:21 i had to fight to get people here to roll up to python3 Dec 07 21:05:57 I also really don't understand why debian doesn't label python2 packages explicitly as "python2-", with "python-" as deprecated compat aliases Dec 07 21:06:31 hedgeberg: didn't hass have a generic installer? using an rpi-specific installer sounds like a bad idea Dec 07 21:06:56 https://www.home-assistant.io/hassio/installation/#alternative-install-on-generic-linux-server Dec 07 21:07:00 yeah, idk. i like that arch maintains all python 2 packages as python2- everything, makes finding packages in the aur way simpler Dec 07 21:07:23 ?? i wasnt the one installing hass, idek what it is Dec 07 21:07:32 sorry I meant jakke: Dec 07 21:07:52 oh no, humble bundle breach Dec 07 21:07:56 uuuuuuuuugh Dec 07 21:08:51 jakke: or better question is why it even started attempting to install anything... the "chown pi:pi hass_rpi_installer.sh" command should have failed since that user doesn't exist Dec 07 21:09:05 yeah, I changed that to the debian user Dec 07 21:09:11 then it ran fine Dec 07 21:09:19 until the end, where that error occured Dec 07 21:09:20 except it didn't Dec 07 21:09:42 it made it for a good 5-6 mins before closing down Dec 07 21:10:02 so now there's a half installed application, which I have no idea on how to remove :) Dec 07 21:10:29 most likely it's just missing some dependency Dec 07 21:10:38 the script actually looks pretty benign Dec 07 21:11:23 i'm using the other link now, for advanced users ;) Dec 07 21:12:40 what was the context of the error message, i.e. the stuff printed before it? Dec 07 21:13:15 NameError: name 'sys_platform' is not defined ---------------------------------------- Command "python setup.py egg_info" failed with error code 1 in /tmp/pip-install-OJnvYW/cffi/ Dec 07 21:14:01 please don't paste stuff like that in irc, use a paste service like pastebin.com Dec 07 21:14:29 weird error Dec 07 21:15:09 right? Dec 07 21:15:20 that's my thought too Dec 07 21:16:22 I'm finding some reports that upgrading pip and then setuptools is causing the problem, while upgrading setuptools and then pip works fine Dec 07 21:16:37 (the script does pip and then setuptools) Dec 07 21:16:54 try sudo pip install --upgrade setuptools Dec 07 21:17:42 hass.io needs docket? network-manager? wtf is this thing Dec 07 21:17:44 *docker Dec 07 21:18:14 you run Google Assistant on it, and then you hook it up to your chromecast Dec 07 21:18:24 Hue lights, etc. Dec 07 21:19:26 I'm also note really clear on the connection between hass.io and home-assistant Dec 07 21:19:43 it's an abbre Dec 07 21:19:58 abbreviation, i assume Dec 07 21:19:59 no I don't think so Dec 07 21:20:51 home assistant is part of hass.io, but the latter seems to be something much bigger Dec 07 21:21:05 ok.. Dec 07 21:21:21 this paragraph makes that very clear: https://www.home-assistant.io/hassio/#upgrading Dec 07 21:22:31 https://www.home-assistant.io/faq/#home-assistant-vs-hassio Dec 07 21:22:42 also huh, that FAQ is referencing python 3 Dec 07 21:23:06 yet that install script you were using was python 2 Dec 07 21:23:14 is there another SBC with PRU-type capability? Dec 07 21:23:23 sicelo: beagleboard-x15 Dec 07 21:23:35 non-beagleboard, that is :) Dec 07 21:23:52 PRU is a TI thing Dec 07 21:24:46 i haven't used it yet, but it sure seems to be amazing Dec 07 21:24:47 there might be other SBCs with a recent TI SoC (am335x/am4/am5), not sure. at the very least some SOMs have a base/experiment board Dec 07 21:25:07 probably all of them a lot more expensive than the beaglebone Dec 07 21:25:24 PRU is pretty awesome yes Dec 07 21:25:49 jakke: this is reaaaaly confusingg Dec 07 21:26:03 jakke: the Home Assistant FAQ says "There will be no support for Python 2.x. Python 2 will be EOL in 2020 and it doesn’t make any sense to support a release which was planned to be retired in 2015. Dec 07 21:26:50 I mean, If I knew how to edit the script to have it run on pip3 I would do it Dec 07 21:27:15 jakke: my question would be why was it using pip rather than pip3 in the first place Dec 07 21:27:29 poor scripting? Dec 07 21:27:30 does raspbian use python3 as default? Dec 07 21:29:13 hmmnope, it seems "pip" is python2 on raspbian too Dec 07 21:29:19 very strange Dec 07 21:30:22 anyway, the dependencies of hass.io are going to make it a big headache to install Dec 07 21:30:24 indeed, I can't even curl the http link on the hass.io page there, so I'm not the right guy to answer Dec 07 21:31:25 maybe see if this home assistant thing has a support irc channel? Dec 07 21:31:39 yeah, I'll stop bugging you guys Dec 07 21:32:10 at least I have more space on my BBGW thanks to you Dec 07 21:32:18 :) Dec 07 21:45:55 Hello, I was wondering if anyone knew how to enable cloud9 to run as root? I know how to use sudo, but I want to run the ide as root. Any thoughts? Dec 07 21:46:11 I am running the latest version of stretch. Dec 07 21:47:01 sudo su will give you a root shell Dec 07 21:47:17 please don't use "sudo su", that creates two nested login sessions Dec 07 21:47:23 yeah, its not ideal. Dec 07 21:47:24 use "sudo -s" or "sudo -i" Dec 07 21:47:44 (the latter gives a clean environment like you've just logged in) Dec 07 21:47:45 Thank you. I know, but I am actually looking to allow the apps that I am editing to run as root. Dec 07 21:48:04 zmatt, full confession, i find myself using sudo su far more often than i should... Dec 07 21:48:19 :) Dec 07 21:48:20 MrR^2 ?? Dec 07 21:48:31 MrR^2: so, there are many dubious things here... almost never should you need to run an application as root, and even if you do there's no reason to run an ide as root Dec 07 21:48:38 yeah... Dec 07 21:48:55 can you break down the situation as to why you want that? Dec 07 21:49:06 All of the robotics elements need root to run. Dec 07 21:49:07 the "debian" user already has a huge amount of privileges Dec 07 21:49:15 what? why? Dec 07 21:49:27 pwm Dec 07 21:49:34 pwm doesn't require root privileges Dec 07 21:49:35 motocontrol Dec 07 21:49:37 pwm doesn't need root Dec 07 21:49:43 it does when using rcpy Dec 07 21:50:01 Oh, I may have not installed the libraries correctly. Dec 07 21:50:01 still shouldn't, if your user is in the right groups Dec 07 21:50:05 I will try that Dec 07 21:50:34 Should the default user be in the correct groups by default? Dec 07 21:50:35 and, if what you're using does need that, you really shouldn't be using it unless you know the linux kernel very, very well Dec 07 21:50:47 i mean, is this software that ships default in stretch? Dec 07 21:50:58 Sort of Dec 07 21:51:01 otherwise, no, you probably need to add some groups to your user Dec 07 21:51:25 I have a beaglebone blue and using python libraries to run the motor control Dec 07 21:51:47 The examples are built in, but do not work by default Dec 07 21:51:53 the /etc/udev/rules.d/81-pwm-noroot.rules rule ensures pwm is accessible by any user in the "pwm" group Dec 07 21:52:00 the debian user is in the pwm group by default Dec 07 21:52:26 if you have the latest version of the bb-customizations package you'll also get cool symlinks in /dev/pwm/ Dec 07 21:53:04 Do you know how to ensure the default user is in the correct group for an application I install? Dec 07 21:53:18 I don't know what you mean by that Dec 07 21:53:58 So, if I do apt-get install, do I need to run that as root? Dec 07 21:54:11 apt-get install needs root privileges yes Dec 07 21:54:18 use sudo for that Dec 07 21:54:31 If so, the app and library installed is not accessible to the current user correct? Dec 07 21:54:43 what? no? Dec 07 21:54:51 packages are system-wide Dec 07 21:55:20 the privileges apt-get needs to install packages has nothing to with privileges needed by or for an application or library Dec 07 21:55:36 that is what I thought, but when I run an app I made (or the one pre installed), it cannot find the library unless I run as root. Dec 07 21:55:51 that sounds very unlikely. what is the exact error message? Dec 07 21:56:22 Traceback (most recent call last): File "/var/lib/cloud9/examples/robot/python/rcpy/motorandencodertest.py", line 1, in from rcpy.motor import motor1, motor2 ImportError: No module named rcpy.motor Dec 07 21:57:00 But, if I run as root, it is fine Dec 07 21:57:19 sudo python3 motor.py Dec 07 21:57:43 why is the name of the script suddenly different? Dec 07 21:58:16 they are two scripts that do the same thing. Dec 07 21:58:17 also, are you sure the relevant python library was installed by apt? it could be installed in a user-specific python directory (e.g. if you used pip) ... although that would be a bit weird for root Dec 07 21:58:50 I did install it with pip, but also installed it with apt Dec 07 21:59:09 so if you just do "python3 motor.py" you get the same importerror? Dec 07 21:59:10 So, the python libraries are user independent? Dec 07 21:59:24 whatever you install with apt is global for all users Dec 07 21:59:57 the same is not true for pip (although I thought it was when using pip as root, but I'm not sure.. I've never used pip as root) Dec 07 22:00:05 Okay, that is beginning to make sense. Dec 07 22:00:24 I will try using pip as non-root, and then see what happens. Thank you. Dec 07 22:00:27 regardless, not being able to find packages has nothing to do with privileges Dec 07 22:00:39 it has to do with not being able to find packages ;) Dec 07 22:00:50 lemme check how to ask python where a library was found Dec 07 22:01:04 So, then why does it work when I run as root? Dec 07 22:01:23 different module search path Dec 07 22:01:27 I remember going through this with a raspberry pi a bit ago as well. Dec 07 22:01:59 often, people do stuff as root when they shouldn't have, and then end up having to do more stuff as root because of the earlier stuff they did as root Dec 07 22:03:17 I totally think that is the situation I am in. I just looked at the install instructions for rcpy. It looks like I am installing the libraries locally. Dec 07 22:03:29 try: echo 'import rcpy.motor; print(rcpy.motor.__file__)' | sudo python3 Dec 07 22:03:37 it should show the file path of the module Dec 07 22:07:03 So, I do think it is local. Dec 07 22:07:04 /usr/local/lib/python3.5/dist-packages/rcpy-0.5.0-py3.5-linux-armv7l.egg/rcpy/__init__.py:116: UserWarning: > Robotics cape initialized warnings.warn('> Robotics cape initialized') /usr/local/lib/python3.5/dist-packages/rcpy-0.5.0-py3.5-linux-armv7l.egg/rcpy/__init__.py:127: UserWarning: > Installing signal handlers warnings.warn('> Installing signal handlers') /usr/local/lib/python3.5/dist-packages/rcpy-0.5.0-py3.5-linux-armv7l.eg Dec 07 22:07:37 that's still system-wide though Dec 07 22:07:52 Oh. Odd Dec 07 22:07:57 but, what is all this output? Dec 07 22:08:09 this is all output of the my command? Dec 07 22:08:09 The last warnings are what you get when it runs. Dec 07 22:08:18 oooooookay Dec 07 22:08:23 and if you run it without sudo ? Dec 07 22:08:46 Rather than just printing "Hey this is running now" they put it in a warning. Dec 07 22:09:53 it's pretty rude they're doing a lot of stuff when the module is loaded rather then when an initialize function is called or a class instantiated Dec 07 22:10:08 like, normally you want to be able to import a module to e.g. query its version without too much side-effects Dec 07 22:10:21 This is odd. It worked. Dec 07 22:11:34 btw, I notice rcpy's documentation asks for "roboticscape" to be installed... that's an old and presumably deprecated ancestor of librobotcontrol Dec 07 22:11:45 it seems rcpy hasn't been updated in a long time Dec 07 22:12:49 It is. Dec 07 22:12:59 also WTF... "sudo mv /usr/local/lib/python3.5/site-packages/* /usr/local/lib/python3.5/dist-packages/" that's bad and wrong in every way Dec 07 22:13:22 So, when I run the file from a non root terminal prompt, this is what I get. /usr/local/lib/python3.5/dist-packages/rcpy-0.5.0-py3.5-linux-armv7l.egg/rcpy/__init__.py:116: UserWarning: > Robotics cape initialized warnings.warn('> Robotics cape initialized') /usr/local/lib/python3.5/dist-packages/rcpy-0.5.0-py3.5-linux-armv7l.egg/rcpy/__init__.py:127: UserWarning: > Installing signal handlers warnings.warn('> Installing signal handl Dec 07 22:13:31 I think anyway Dec 07 22:13:35 oh, huh Dec 07 22:14:26 I think it got cut off. I will try again. Sorry for the double post. Dec 07 22:14:29 that's such a weird step to include... if that directory is not part of python's path? what installed something there and why? Dec 07 22:14:31 ERROR: in rc_pru_shared_mem_ptr could not open /dev/mem: Permission denied Need to be root to access PRU shared memory ERROR in rc_encoder_pru_init, failed to map shared memory pointer ERROR: failed to run rc_encoder_pru_init Dec 07 22:14:32 please use pastebin Dec 07 22:14:49 oh yeah the morons use /dev/mem Dec 07 22:14:57 yuck Dec 07 22:15:15 there's absolutely no need for that -.- Dec 07 22:15:24 are you even using the pru encoder stuff? Dec 07 22:15:41 But, without rewriting the wrapper, I cannot avoid it correct? Dec 07 22:16:10 * zmatt sighs Dec 07 22:16:15 Or, throwing my students into learning C. Dec 07 22:16:26 Oh, me too;) Dec 07 22:16:38 which functionality of rcpy are you actually using? Dec 07 22:17:06 We are using it to build a robot before jumping into the roborio. Dec 07 22:17:22 that doesn't answer my question Dec 07 22:17:37 since for example pwm control is trivial in pure python Dec 07 22:17:44 so, we need it to control the motor, servo, encoder, di, ai, and imu of the robot Dec 07 22:17:58 ok yeah, a bit more Dec 07 22:19:23 anyway, so it does sound like the module was found when you ran it as non-root Dec 07 22:19:35 so the error from cloud9 is still unrelated to privileges Dec 07 22:20:40 Oh, that is a good point. So, I guess I still need to teach the students to launch their apps in terminal Dec 07 22:20:56 I don't know anything about cloud9... can you run this in it: import sys; print(sys.path) Dec 07 22:20:58 so, then I can teach them to use sudo if necessary. Dec 07 22:21:26 that should show where it looks for python modules Dec 07 22:22:36 well, doesn't hurt to learn to launch apps in terminal using sudo :) Dec 07 22:23:14 true Dec 07 22:25:43 It says it is looking in a few places separated by a colon. I tried moving ../python/3.5/... to the front, but it did not seem to change anything. Dec 07 22:27:02 Oh well, thank you for your help and time. Dec 07 22:28:01 can you be more specific and just quote what it says? Dec 07 22:28:13 (using pastebin preferably) Dec 07 23:02:16 Hi! Does anyone have any ideas of BBB enhanced reliability? I have seen postings of BBB not booting when power is applied and I am concerned these errors carry over to the enhanced/enhanced industrial board. Dec 07 23:02:50 are you referring to filesystem integrity or something else? Dec 07 23:02:59 put it on a good PSU, see if you have a problem. Then worry about it. Dec 07 23:03:42 my bbbs stay up for months. Dec 07 23:03:47 we reconfigure the eMMC into SLC mode with reliable writes enabled, to increase its lifetime and hopefully its resilience to power interruptions Dec 07 23:04:29 our devices never get the opporunity to shut down cleanly (they're simply turned off by cutting power) and so far this hasn't been a problem for us Dec 07 23:05:37 For some context, my company is planning to use BBB for a space application and we are interested in minimizing the likelihood of errors during boot along with memory, processing, or any other critical aspects of the boards Dec 07 23:05:43 of course if you can prevent unclean shutdowns, that would be even better Dec 07 23:05:58 wait, what? Dec 07 23:06:06 you cannot use BBB in space Dec 07 23:06:18 its a great sbc but it is /not/ right for that application Dec 07 23:06:36 the SoC is not rad tolerant Dec 07 23:06:39 It actually has been used a few times with some minimal issues, like 1/14 had a critical failure Dec 07 23:07:05 thats <95% yield rate, thats unacceptable Dec 07 23:08:03 Right, I agree it appears to not be ideal, but the ease of use and programming makes it desirable if we can find a way to improve reliability Dec 07 23:08:08 the entire system is incorrect for space. you want a rad-tolerant, fault-tolerant SoC. you want all ram to be ECC enabled, etc Dec 07 23:08:35 my recommendation as a professional engineer? don't go down that road Dec 07 23:09:14 if you want a space-capable ecosystem, you're not going to have an easy time, end of story Dec 07 23:09:59 I'm assuming this isn't for a particularly critical application, or using commercial non-rad-hardened hardware wouldn't even be considered? Dec 07 23:10:15 if you want to design some replacement board that fits the same cape, that's one thing Dec 07 23:10:41 Yeah I think you're right, but I should clarify we're in a relatively low radiation space environment (~1-2 kRad or less) and it's a demo mission. Dec 07 23:10:51 then you can use the BBB as a ground-site dev platform until you have a better situation Dec 07 23:10:54 mmmmmmm Dec 07 23:10:56 hrm Dec 07 23:11:39 sam_: anyway, regardless of the application, configuring eMMC into SLC mode seems like a good idea. It probably also increases the eMMC's resilience to bit-errors of any sort Dec 07 23:11:42 I guess if you really, really wrap it in copper shielding? maybe you'd be mostly ok if its low rad? Dec 07 23:11:47 yea Dec 07 23:11:50 *yeah Dec 07 23:11:59 this is of course only one component of many Dec 07 23:12:13 SLC mode is going to be a lot safer, and it should give you some nicer ECC options Dec 07 23:12:19 but if it doesn't boot, presumably eMMC corruption would be the cause Dec 07 23:12:28 ECC is always handled internally Dec 07 23:12:44 in SLC NAND mode, its usually handled on the SoC Dec 07 23:12:56 this isn't NAND flash, it's eMMC Dec 07 23:12:58 SLC is latch logic, theres next to no processing on the flash chip Dec 07 23:13:23 SLC is a NAND standard, eMMC is an interface standard Dec 07 23:13:50 Ah got it, yeah we haven't actually experience these boot issues ourselves, I have just read about it online. For instance, this article (http://www.mikini.dk/2014/10/beaglebone-black-periodic-boot-failure-establishing-failure-rate-and-possible-cause) was particularly concerning with a 4% boot error Dec 07 23:14:08 the eMMC can be configured to store only one bit per cell instead of two bits per cell (its default) Dec 07 23:14:20 i.e. SLC mode Dec 07 23:14:23 ooooooh Dec 07 23:14:26 i see Dec 07 23:14:42 so you're still driving it via emmc Dec 07 23:14:55 just, you have lower density Dec 07 23:14:56 of course, there's no direct access to the underlying NAND Dec 07 23:15:04 yes, you sacrifice 50% capacity Dec 07 23:15:33 yeah Dec 07 23:16:00 what, no love for raw slc nand interfacing? Dec 07 23:16:03 in exchange for like 10x increased lifetime, along with better reliability and much faster write speed Dec 07 23:16:17 (i have done said things, it is absolutely painful) Dec 07 23:16:25 using raw nand is a pain in the ass Dec 07 23:16:26 Sorry, why would the lifetime increase? Dec 07 23:16:36 sam_ nand flash is complicated Dec 07 23:16:54 flash memory uses a weird kind of technology called "floating gate MOSFET" Dec 07 23:16:59 sam_: in SLC mode, the nand only has to distinguish whether a cell is erased or programmed Dec 07 23:17:17 sam_: in MLC mode is also uses two intermediate slightly-programmed levels Dec 07 23:17:25 and has to distinguish those again on read Dec 07 23:17:29 basically, a floating gate mosfet is a weird capacitor that can, in theory, store a charge for an indefinite amount of time Dec 07 23:18:03 its a lot more complicated than that but to save the semiconductors lesson, you can think of it as a capacitor Dec 07 23:18:15 its not, its really a transistor, etc etc Dec 07 23:18:40 the way flash distinguishes whether a bit is a 1 or a 0 is by a shift in the voltage required to "turn on" the transistor Dec 07 23:18:47 so reading an SLC cell is just more much error-tolerant than reading an MLC cell, since it only requires distinguishing 2 voltage levels instead of 4 Dec 07 23:18:56 yeah ^ Dec 07 23:19:27 basically, to store 2 bits, you need 4 discrete voltage levels (1 for 00, 1 for 01, 1 for 10, 1 for 11) Dec 07 23:19:35 I'm not sure exactly what happens to a cell as it ages, but presumably it loses the ability to reliably hold charge Dec 07 23:19:42 a few things Dec 07 23:19:48 Ah I see, sounds like there's a lot more to learn haha. Do you know roughly the system lifetime? Dec 07 23:19:57 it depends on the implementation Dec 07 23:20:02 and how much you write Dec 07 23:20:19 writing data is what causes NAND flash to age Dec 07 23:20:21 in general, SLC nand can usually handle.... 10,000 writes? Dec 07 23:20:29 *SLC flash Dec 07 23:20:42 meh that's the industry "rule of thumb" Dec 07 23:20:45 I've seen varying numbers, but that order of magnitude yes Dec 07 23:20:47 nand emmc flash does some fancy stuff though to distribute those writes Dec 07 23:21:21 yeah I have no idea how to estimate write-amplification Dec 07 23:21:30 theres a really fancy alogrithm about distributing writes that can extend lifetime of a flash chip to "1 billion writes" Dec 07 23:21:41 at least, thats the intel claim Dec 07 23:21:49 or was, on their first batch of SSD's Dec 07 23:22:08 but for standard cheap eMMC nand, you're probably not exposed to that Dec 07 23:22:17 sam_: wear-out might not be your biggest concern though. in your case it would mostly be to increase reliability in general. I can't quantify that though, it's just my assumption the eMMC will be more reliable and fault-tolerant in SLC mode than in MLC mode Dec 07 23:22:34 thing is, in a high-radiation environment, nand memory will die much faster Dec 07 23:22:42 zmatt it is Dec 07 23:23:05 so, here's some wild facts: Dec 07 23:23:27 floating gate mosfets are basically a mosfet sandwich Dec 07 23:23:44 sam_: and like I said, as extra bonus it also increases write speed (I've seen a 3x increase when I measured it) because it's much easier for the controller to fully program a cell than to carefully partial-program it Dec 07 23:23:56 sam_: the price you pay is losing half the disk space Dec 07 23:24:12 if you look at the way a mosfet is designed normally, its a metal gate layer, then an oxide layer, then your silicon Dec 07 23:24:26 space-qual dev platforms are super underserved Dec 07 23:24:50 a floating gate mosfet goes metal gate, oxide, electrically isolated (floating) metal layer, oxide, silicon Dec 07 23:24:58 so you have 2 metal layers Dec 07 23:25:06 That seems reasonable - we will look into using SLC mode. Do you think there's anything else we might want to consider maybe on the hardware side? Dec 07 23:25:12 the middle metal layer, the "floating gate" builds up charge Dec 07 23:25:24 i mean sam_ i wouldnt even use a single flash chip Dec 07 23:25:45 if you want to use flash? i'd use 4, and use an fpga to determine correct data Dec 07 23:26:01 it really depends on your budget etc Dec 07 23:26:05 i know that sounds nasty, but thats the right way to do space resilient systems Dec 07 23:26:33 i guess you could do a 4x emmc raid controller? Dec 07 23:26:44 sam_: configuring the eMMC into SLC more with reliable writes enabled is kinda tricky, since it involves multiple error-prone commands and these settings are One Time Programmable only. so we patched mmc-utils to add a command that performs a whole bunch of sanity checks and then performs all steps of the reconfiguration Dec 07 23:26:56 Ah yeah I see, I mean we have baselined the BBB enhanced industrial for now, but we could feasibly add something extra or include redundant BBBs Dec 07 23:27:11 sam_: see the 'custom' branch of https://github.com/dutchanddutch/mmc-utils Dec 07 23:27:23 Nice, thanks! Dec 07 23:27:30 hahahaha really really weird Dec 07 23:27:34 the last commit adds the relevant command Dec 07 23:27:45 developing a fault tolerant system out of essentially consumer grade hardware Dec 07 23:27:59 it's never going to be particularly fault tolerant Dec 07 23:28:03 zmatt so, rq, even in SLC, the long story short is that stray radiation effects can build up a charge in the floating gate Dec 07 23:28:13 go ahead and drop $2000 on your rad-hardened 20MHz uC Dec 07 23:28:24 hedgeberg: but more charge will be required to toggle it Dec 07 23:28:32 than in MLC Dec 07 23:28:35 because thats how flash storage works by default Dec 07 23:28:48 i did find a $600 space qual m0 Dec 07 23:29:04 yeah, but its still not ideal since in order to toggle it you need to drain more charge Dec 07 23:29:25 basically, flash is especially prone to errors compared to other systems, is my point Dec 07 23:29:47 and if youre going to use flash storage in space, you not only want SLC, you also want significant redundancy of some kind Dec 07 23:29:59 well flash already involves a lot of redundancy Dec 07 23:30:21 generally it has ECC able to fix dozens of bit errors per 1KB page Dec 07 23:30:35 yeah, because thats necessary for it to function within the ionosphere Dec 07 23:30:36 Right, it's not too uncommon actually, just to have a few of the same COTS parts for redundancy. We have a huge benefit of using hardware that's well known with a good support community Dec 07 23:31:09 without that redundancy flash dies remarkably quickly Dec 07 23:31:28 or, as someone amusing said: flash doesn't really store your data, but it's generally able to recover it Dec 07 23:31:36 yeah Dec 07 23:31:52 i mean, funny and a touch frightening, but also accurate Dec 07 23:32:56 point being SLC mode will help a lot compared to MLC, but you should assume lifecycle and reliability of a single nand chip drop by 2 to 3 orders of magnitude Dec 07 23:33:32 sam_: btw, I just looked at the link... it is indeed true that the BBB requires sufficiently fast ramp-up of input power Dec 07 23:34:25 though the 500us some post claims is nonsense Dec 07 23:34:42 it requires ramp up to occur within 50ms Dec 07 23:34:47 sam_ if i were you, i'd go with a bbb for now and find a linux-capable space=-grade SoC + hardened components for later in dev cycle Dec 07 23:34:55 oh I see - we're planning to run at least 2 BBB for just a few months in low earth orbit so hopefully that lifetime wouldnt be an issue. Ah got it, so that's completely solvable, right? Dec 07 23:35:28 and slap those onto a board in the beaglebone form factor Dec 07 23:35:53 hhhhh i'd honestly be surprised if your BBB lasted a few /days/ in LEO Dec 07 23:36:27 maybe im being paranoid but when you can't service something, reliability requirements become much higher Dec 07 23:36:35 sam_: another big concern is memory errors, especially since it has no ECC. to some extend you'll have to manage that by making the software as robust as possible... double-check things, and just deliberately crash and recover if sanity checks fail Dec 07 23:36:41 and use a watchdog obviously Dec 07 23:36:51 ^ Dec 07 23:37:08 tbh, i wouldnt even want to run linux on this, i'd want it to be single-threaded system Dec 07 23:37:30 you know what? just realized AVR might actually be a good platform for space stuff? Dec 07 23:37:44 I'm trying to stay within the outline sam_ gave Dec 07 23:38:08 Yeah those are very good points - we could look into building a custom board later, as you say hedgeberg. We are planning for at least one external watchdog Dec 07 23:38:17 there's an internal one too Dec 07 23:38:20 yeah i know, i guess im just saying if i were them, i'd be telling my boss that we need to go back to the drawing board, but im an asshat Dec 07 23:39:04 doing it Right may require more time and money than available Dec 07 23:39:12 Haha nah good point. Our timeline is a bit too tight at this point to make any major changes Dec 07 23:39:44 then your best solution may be to stay the course, us a /lot/ of copper tape, and pray Dec 07 23:39:47 hoping the ducktape will hold for two months is not ideal, but hey... as long as you're aware that that's what you're doing ;) Dec 07 23:39:48 *use Dec 07 23:40:26 sam_ is this cubesat stuff? Dec 07 23:40:58 Yeah we're generally aware - we've just recently launched a space-grade satellite that was a 3U cubesat. The next one isn't a cubesat but is roughly a 2U cubesat size Dec 07 23:41:28 if you really want to get fancy you could have two boards performs the same tasks, inform each other of what they're doing and cross-check it Dec 07 23:41:49 i'd even say 3, that way you can generate consensus Dec 07 23:41:52 and force recovery on mismatch Dec 07 23:42:02 Good point - we've definitely been considering that, then it just becomes a matter of space Dec 07 23:42:05 yeah, or 3 to be able to use consensus and recover for a dead board Dec 07 23:42:10 *from Dec 07 23:42:44 2 suffices for transient errors if they're rare enough and the time it takes to recover (e.g. by reset) is acceptable Dec 07 23:43:19 or 9, and then you can 4 liberal boards and 4 conservative boards and 1 "swing" board who everyone knows really just votes conservative every time making the whole thing moot Dec 07 23:43:26 *cough* sorry what was that Dec 07 23:43:40 So we are planning to have 2 currently, but they aren't running the same processes. We could add extra Dec 07 23:43:51 Haha yes that's perfect Dec 07 23:44:38 >politics jokes in a tech irc channel Dec 07 23:44:42 inb4 im get ban Dec 07 23:45:08 and overall minimize eMMC writes, especially metadata writes. perhaps periodically flush disk cache Dec 07 23:45:39 since the worst possible thing that could happen is for disk-related data structures to get corrupted in memory and then written back to eMMC Dec 07 23:45:47 yeah Dec 07 23:46:55 i'd probably do the emmc raid thing, + use an mcu with no OS to minimize unknown i/o, + a hyperram sram type dealio so that we wouldnt need to worry about things like dram refresh etc Dec 07 23:47:13 also ensure that the boards will try their best to recover from filesystem errors... you don't want one to end up hanging at an "Ok to fix errors? [Y/N]" prompt on the serial console Dec 07 23:47:37 hedgeberg: you're back to pink unicorn land Dec 07 23:47:40 ;) Dec 07 23:47:41 and then in an ideal world, do 3 of those in parallel and generate consensus Dec 07 23:47:43 i know Dec 07 23:47:55 let's try to focus on reliality Dec 07 23:48:00 *reality Dec 07 23:48:01 im saying if i were a magic overbudget and overtime engineer Dec 07 23:48:06 typing is hard Dec 07 23:48:39 (which, i am, i work for a well funded nonprofit, so my sense of priorities is off) Dec 07 23:49:18 zmatt how much can you restrict i/o in a linux build? Dec 07 23:49:41 yeah I work for a young company, so I'm used to shipping things that are held together by ducktape and then hope you get an opportunity to replace it with something better before the ducktape falls apart Dec 07 23:49:58 mmmmm so you live in technical debt hell Dec 07 23:50:16 One thing I forgot to mention is we have only about 1-2 months of time to pull off any adjustments to standard BBB with essentially 2 full time engineers. Do you guys have any thoughts of if that's feasible? Dec 07 23:50:41 i mean, its a lot more feasible with BBB than it is with a lot of other dev platforms sam_ Dec 07 23:51:05 you've got the linux kernel source, and youve got the BBB debian patch source Dec 07 23:51:25 you're already gambling on whether or not the whole thing will keep working long enough, so sure, why not ;) Dec 07 23:51:35 hahaha Dec 07 23:51:36 yeah good point, just wanted to confirm it's not absolutely insane haha Dec 07 23:51:44 just mostly insane Dec 07 23:51:50 it's not really an answerable question, since it depends very much on the engineers Dec 07 23:51:52 nobody really knows what your project requirements are or penalties for failure eh Dec 07 23:51:56 no its absolutely insane Dec 07 23:52:20 you're sending consumer grade hardware to space and praying that hacking some stuff together will get the job done Dec 07 23:52:46 but, sometimes thats your only option, and if thats the case then, well, you don't have other options Dec 07 23:53:00 i mean there are circurmstances where even thats okay Dec 07 23:53:23 its not like sam_ came in her saying that they're gonna use the bbb to control flow control valves on some life support system Dec 07 23:53:24 yep exactly, but if you can prove that a system like that does the job, then that's pretty useful for future space developers. Custom space-grade hardware sucks to work with in my experience Dec 07 23:53:34 i disagree Dec 07 23:53:36 100% Dec 07 23:54:02 when the choice is between bootstrapping an insnane project together, and not shipping, i say choose the insane project 100% of the time Dec 07 23:54:24 no, if it does the job then it means luck was in your favor :P Dec 07 23:54:28 can someone point me to an space-quality equiv of the bbb? Dec 07 23:54:39 because i'm not aware of any Dec 07 23:54:47 nah, which is a shame Dec 07 23:54:48 t Dec 07 23:54:50 right so Dec 07 23:54:53 *that would be a fun project Dec 07 23:55:00 if you're tlaking about "proofs of concept" for furture space developers Dec 07 23:55:08 i think you're at best developing on M0's right now Dec 07 23:55:12 there might be better Dec 07 23:55:23 ive done the equivalent of one bathrooms trips worth of research Dec 07 23:55:33 hedgeberg: do you have any way to simulate the radiation environment the bbb will be exposed to, for testing? Dec 07 23:55:45 not that it matters hugely I guess, since you don't have enough time for reliability testing anyway Dec 07 23:56:03 you could send a bbb up on a baloon rolfl Dec 07 23:56:04 rofl* Dec 07 23:56:11 maybe you could increase expsure that way :_p Dec 07 23:56:32 /24/24 Dec 07 23:56:36 hah Dec 07 23:57:57 uhhhhh Dec 07 23:58:00 here? Dec 07 23:58:02 yes i do Dec 07 23:58:17 can i use that for some rando's h/w? probably not without gov't approval Dec 07 23:58:48 i can however ask the satellites people downstairs if there are good public test tools that someone could get access to on a rush? Dec 07 23:58:56 For analysis of radiation we've been using spenvis (https://www.spenvis.oma.be/) assuming a 5 mm aluminum shield, the total dose is pretty low for the BBB - somewhere around 1 kRad per year. There are a few facilities that run proton testing - like UC Davis in California and others around the world, usually like a couple hundred dollars per hour of testing Dec 07 23:59:39 not sure how much an hour of testing is going to tell about reliability during two months of operation Dec 07 23:59:47 yeah... Dec 07 23:59:54 well, I mean, if it crashes instantly then it tells enough Dec 07 23:59:55 ;) Dec 07 23:59:55 radiation effects are probablistic Dec 08 00:00:14 True, generally its a toss up where you blast the board and see the effects of high total dose in short time periods Dec 08 00:00:40 anyway i gotta head home, night all Dec 08 00:00:45 that's not really the same thing as a low dose over a long time of operation Dec 08 00:00:56 Thanks for the inputs hedgeberg! Dec 08 00:01:01 (where by "not really" I mean "not remotely") Dec 08 00:01:13 ill be back on in like... 4 hours maybe? Dec 08 00:01:17 yeah agreed, it's a hard thing to accurately test Dec 08 00:02:24 low dose over long time != high dose over short time because of variable system state? Dec 08 00:02:31 I need to head out as well, so thanks for your inputs zmatt. I really appreciate it Dec 08 00:02:35 because of variable system state while powered on? Dec 08 00:02:50 because i'm pretty sure that the manufacturing testers more or less test lifetime under that assumption Dec 08 00:04:08 ayjay_t: yeah, it might work for testing how well the hardware survives physically, but it's not a good model for what happens to the operational system Dec 08 00:04:30 introducing a large pile of errors into a running system probably results in a crash and watchdog reset Dec 08 00:05:15 fewer errors over longer time increase the risk that software will stay running and act on erroneous data Dec 08 00:06:21 its an interesting though Dec 08 00:06:23 thought* Dec 08 00:06:30 its definitely true in certain circumstances Dec 08 00:07:28 conversely, in the presence of redundancy and error-checking/recovery, a low error rate may be required to be able to recover Dec 08 00:07:32 think e.g. ECC memory Dec 08 00:07:46 not 100% on how ECC works Dec 08 00:07:49 on refresh? for volatile? Dec 08 00:07:54 or on read/write Dec 08 00:08:07 becasue my thought is that if you're speeding up the "environment" you can likely speed up the internal environment too Dec 08 00:08:23 to some limited degree maybe Dec 08 00:08:41 i should have said "speed up the internal system" Dec 08 00:08:46 but you figured out what i meant Dec 08 00:08:50 there are limits to how fast you can scrub ECC Dec 08 00:09:03 especially if you still need a working system too Dec 08 00:09:33 your testing environment in either case is becoming a conservative one Dec 08 00:09:39 man i wish i had cool space equipment Dec 08 00:09:40 wtf Dec 08 00:11:04 ECC works by being able to repair a certain number of bit errors per memory-word on read, and if it's been detected that such a correctable error has occurred the memory location needs to be written again to repair it before it has a chance to accumulate even more errors Dec 08 00:11:35 this is paired with a background "scrubbing" process that sweeps all memory with reads to ensure errors can't accumulate in infrequently-accessed memory Dec 08 00:12:10 anyway, food! Dec 08 00:13:33 :-p i have one more "system" to build Dec 08 00:13:44 a redis database and it's API for my server Dec 08 00:15:57 Psyche I'm back I guess Dec 08 00:16:01 Missed my bus Dec 08 00:16:26 Side note can someone remind me how to query irc flags so I can ensure I'm logged in with SASL Dec 08 00:17:24 anyway what have I missed Dec 08 00:18:50 afaict there's no usermode that specifically indicates you used SASL auth Dec 08 00:19:10 does come in on logs tho on the network window? Dec 08 00:19:36 yeah it should show up in the server messages when you connect Dec 08 00:19:38 -!- SASL authentication succeeded Dec 08 00:20:41 oh actually I don't know if that's a server message, it might be a message from irssi itself Dec 08 00:20:59 tbh i still have no idea how to use irssi Dec 08 00:21:10 \me discovered \help tho that was nice Dec 08 00:21:16 oops wrong slash see waht i mean Dec 08 00:21:21 ANYWAY i'm behind like 25 hours this week fml Dec 08 00:21:49 I mean nickserv didn't get mad at me and I'm authenticated, so I guess it worked? Dec 08 00:22:33 irc is normally plaintext? Dec 08 00:22:53 is sasl something that implements tls? Dec 08 00:23:37 Sasl is a separate authentication thing handled on initial connection I think? Dec 08 00:23:42 /whois shows i'm using a "secure connection" and i don't honestly know how many possible security mechanisms there are Dec 08 00:23:45 I'm not 100% sure how it works Dec 08 00:23:59 ayjay_t: "secure connection" just refers to the use of TLS Dec 08 00:24:41 yeah i think you're right yeah Dec 08 00:25:09 sasl is a replacement for "/msg nickserv identify " that may or may not be more secure Dec 08 00:25:16 sooo dumb Dec 08 00:26:32 it also has the benefit that authentication happens much earlier Dec 08 00:27:24 ah it's not more secure, the sasl mechanism used is "PLAIN" Dec 08 00:27:37 I'm still upset that /msg nickserv identify is the default Dec 08 00:27:48 Wait are you fucking kidding me Dec 08 00:27:53 yeah Dec 08 00:27:55 thats the case. Dec 08 00:27:59 zmatt is CURRECT Dec 08 00:28:04 Welp Dec 08 00:28:11 TLS is responsible for keeping your password private Dec 08 00:28:26 cryptography is pretty fun Dec 08 00:28:27 I just.... We could use hashes... Dec 08 00:28:30 all these, finite sets Dec 08 00:28:33 key exchange trickery Dec 08 00:28:41 entropy Dec 08 00:28:46 koolest computer science topic ever Dec 08 00:28:55 the purpose of SASL is mostly for the benefit of freenode: the authentication happens before you acquire the nickname and can do anything Dec 08 00:29:06 while old nickserv lets you connect and do stuff before you authenticate Dec 08 00:29:23 freenode was getting riggety-wrecked by bots for a while eh Dec 08 00:30:18 and yes it's an embarrassment that there's still any password-based authentication that doesn't use some flavor of password-authenticated key agreement Dec 08 00:30:27 in fact, pretty much nothing uses password-authenticated key agreement Dec 08 00:30:52 password authenticated key agreement like diffe-helzmeners type somethingerother? Dec 08 00:31:18 even though it lets you establish a secure connection based on a mediocre-entropy password Dec 08 00:31:35 as opposed to sending passwords over tls Dec 08 00:31:55 diffie hellman Dec 08 00:32:01 ayjay_t: yeah they're protocols that sort of blend passwords into diffie-hellman key agreement in a way that results in mutual authentication based on the password Dec 08 00:32:13 i <3 the wiki page for diffie-hellman Dec 08 00:32:19 that paint analogy blew my mind Dec 08 00:32:29 i was a bit drunk at a dive bar yesterday and i _really_ wanted to tell everyone about it Dec 08 00:32:32 had told hold back Dec 08 00:32:36 to hold back* Dec 08 00:33:02 so you wouldn't even need a server-side TLS certificate... a fake server that doesn't have a "verifier" for your password stored wouldn't be able to successfully accept a connection from you Dec 08 00:33:32 Right, now I remember why I hate irc Dec 08 00:33:35 and would only be able to perform one guess of your password per connection attempt Dec 08 00:33:55 Disconnect from server = pain to reconnect Dec 08 00:34:01 oh? Dec 08 00:34:08 _i'm never logging off_ Dec 08 00:34:27 there have been proposals for "session resume", but none have been implemented Dec 08 00:35:30 i haven't spent enough time looking at security protocols Dec 08 00:35:35 i'm kind of interested in it Dec 08 00:35:38 hedgeberg: btw, freenode does support client-side TLS certificates as alternative to password authentication Dec 08 00:35:43 hedgeberg: see https://freenode.net/kb/answer/certfp Dec 08 00:35:44 it sure does Dec 08 00:36:27 Yeah I know, and if I weren't on mobile I would use it Dec 08 00:37:03 But I'm hesitant to do anything involving client-side cert-based auth on Android. Dec 08 00:37:09 why? Dec 08 00:37:18 Because I work in infosec Dec 08 00:37:34 And as such I'm paranoid about Android in all regards Dec 08 00:37:39 It's a security nightmare Dec 08 00:37:57 okay, and? you think a password stored on, or typed in on, your android phone is better secured than a client-side cert? Dec 08 00:38:28 No absolutely not but that's the point, this way it's clear that the line is insecure Dec 08 00:38:30 * ayjay_t munch munch munch Dec 08 00:38:49 hmm Dec 08 00:38:55 I have no illusions that my current connection scheme is secure Dec 08 00:38:59 hedgeberg: you're not making sense. using a client-side cert would only have benefits over a password Dec 08 00:39:40 (from a security point of view) Dec 08 00:39:45 It's like the argument Matt blaze uses for why his gpg key is the shortest possible length Dec 08 00:40:16 never heard of it, don't think I care to Dec 08 00:40:18 Yeah, but it also has the added implication of a secured and well-identified line Dec 08 00:40:47 it doesn't imply anything Dec 08 00:41:13 That's literally the implication of diffie Hellman auth scheme's Dec 08 00:41:16 it's just an alternative to password-auth with the benefit of being able to remove a cert from your account without having to change the password (and as a result having to change settings on your other clients) Dec 08 00:42:05 I'm too tired to get into this, if we want to have this conversation let's wait until my adhd isn't kicking my ass Dec 08 00:42:42 I think I'm fine without it too Dec 08 00:44:00 Point is yes, client side tls auth is more secure, but is there any real point if I connect over an easily compromised endpoint Dec 08 00:44:01 brutal zmatt Dec 08 00:44:28 more secure > less secure Dec 08 00:44:33 and to zmatts earlier point Dec 08 00:44:39 you were complaining about sending password across the link, so I pointed out that you don't have to Dec 08 00:44:42 if one certificate is compromised it's usually not as big of a deal as one password Dec 08 00:44:45 then you seemed to pretend doing so was worse Dec 08 00:44:50 that too Dec 08 00:44:57 I'm clearly not explaining well Dec 08 00:45:22 see i just type the numbers really fast so any hackers couldn't read what i'm trying anyway Dec 08 00:45:35 my typings not to good tho so i make typos like that Dec 08 00:45:50 (also, TLS doesn't support DH authentication I think, instead an rsa or (ec)dsa key is used to sign the ephemeral public key afaik) Dec 08 00:47:27 I had forgotten that irc is a mostly plaintext communication mechanism because I forget a lot of things, so that reminded me of it. past that, I don't really care that much if I'm using my phone to connect since my phone is probably already compromised and probably will remain compromised Dec 08 00:48:08 btw what app for irc on your phone Dec 08 00:48:15 * ayjay_t is never logging off Dec 08 00:48:19 I guess what I'm saying is: assume if it's connected to the open internet, it's pwned. Also, this client app doesn't support client-side auth anyway, so point is moot Dec 08 00:48:55 ayjay_t don't remember tbh. Some beta thing someone was looking for testers on Dec 08 00:49:40 It has a nice color scheme and a not horribly clunky UI, so I'm sticking with it for now Dec 08 00:49:45 COMPROMAT Dec 08 00:49:58 K* Dec 08 00:50:10 Ah yes Cokpromat Dec 08 00:50:26 Everyone's favorite Slavic term Dec 08 00:50:48 there's a couple Dec 08 00:51:26 Yeah I know, if I really want to get back into irc tho I'll set up znc again Dec 08 00:51:43 And then I will set up cert auth and a bunch of other actually useful features Dec 08 00:52:55 i mean i'm not sure what good auth on irc is other than a small barrier to keep out the laziest bots Dec 08 00:53:04 And then idk, find some way to make the link to ZNC secure. Until then? Meh. Dec 08 00:53:15 i mean its the difference between securing every channel or you know Dec 08 00:53:25 i mean i don't put any sensitive info on irc Dec 08 00:53:29 I mean, as a way to verify your co-conspirators it would be nice Dec 08 00:53:33 But yeah, exactly Dec 08 00:53:50 It's not like irc is a secure medium Dec 08 00:54:06 but still, password auth is clearly Dec 08 00:54:07 not the best. Dec 08 00:54:19 indeed, better to use slack instead ;) Dec 08 00:54:22 ugh Dec 08 00:54:28 ugh. Dec 08 00:54:32 i've been looking at ubikey Dec 08 00:54:35 discord then? ^_^ Dec 08 00:54:37 whats that other little usb sized uc that came out? Dec 08 00:54:55 the really small open source one, i think that has symmetric too Dec 08 00:55:03 err asymmetric rather Dec 08 00:55:11 although they gave it a fancy new name Dec 08 00:55:51 no idea what you mean... especially since your description seems to mix hardware and software Dec 08 00:56:06 TI has microcontrollers with public-key crypto acceleration Dec 08 00:56:16 yeah i'm talking about integrated solutions along the whole stack Dec 08 00:56:21 yubikey is a commerical and comerically supported one Dec 08 00:56:36 and then the community has been leveraging a new coin-sized uC dev board as an open source solution Dec 08 00:56:39 but i forget what its called Dec 08 01:01:06 TOMU! Dec 08 01:01:08 it's called the tomu Dec 08 01:02:26 Tomu and tomu fpga are neat Dec 08 01:03:02 oh yeah the tomu fpga could make it even easier to get into fpgas Dec 08 01:03:06 I like yubikey a lot tho considering it has a lot of prebaked support for common cryptography toolkits Dec 08 01:03:07 i'm a huge fan of tinyfpga Dec 08 01:03:30 I wouldn't start with tomufpga tbh, it has exactly one I/o mechanism... Dec 08 01:03:48 haha stop Dec 08 01:03:53 what is it like, a greenpak? Dec 08 01:04:44 If you want an easy usb-enabled fpga that can do easy comms, go for the icebreaker once it hits CrowdSupply Dec 08 01:05:34 Or, if you hate yourself, you can do what I'm planning and implement your own usb-compliant endpoint on the tiny fpga Dec 08 01:05:44 i'd probably go for tiny fpga Dec 08 01:06:00 after that, probably hubbard's spartan baord Dec 08 01:06:10 No no no do not Dec 08 01:06:20 you don't like it because its too proprietary Dec 08 01:06:29 I love Hubbard, and I think he would be happy to know you want to use his board Dec 08 01:06:54 No, I don't think you should use it because it's not a good platform for any kind of fpga novice Dec 08 01:07:07 You want to stay as far away from vivado while learning as possible Dec 08 01:07:29 well, the issue being that, my business partner was trained on vivado in school Dec 08 01:07:34 and we're probably integrating the spartan into our product Dec 08 01:07:41 Oof. Dec 08 01:07:44 Good luck. Dec 08 01:08:11 hahah Dec 08 01:08:11 I love the Spartan fpga series, it's great for whipping up weird dsp things, which is why Hubbard chose it Dec 08 01:08:15 what about the Altera DE10-Nano? Dec 08 01:08:22 hey CoffeeBreakfast Dec 08 01:08:26 Quartus is worse lol Dec 08 01:08:30 but what hedgeberg Dec 08 01:08:47 Especially since Intel picked up altera Dec 08 01:09:01 Quartus was better than ISE, by far Dec 08 01:09:10 good on ya ayjay_t Dec 08 01:09:11 And I'd take quartus over ISE in a heartbeat Dec 08 01:09:30 But vivado beats quartus by a mile Dec 08 01:09:35 It's still /bad/ Dec 08 01:09:44 Just way way way less bad Dec 08 01:10:24 Hence why I tell people to start with OSS capable boards. Much easier to learn on, imo. Dec 08 01:10:52 If you have to use a proprietary toolkit then imo vivado is the best, but that's a low bar to cross Dec 08 01:11:23 Does that make sense? Dec 08 01:11:46 yeah i mean, i'd go OSS everytime if i could Dec 08 01:11:58 but i'm not an expert so i can't make these decisions Dec 08 01:12:04 and keelin is basically not capable of market research for whatever reason Dec 08 01:12:07 he also loves windows Dec 08 01:12:14 he got mad but i had to teach him linux Dec 08 01:12:23 weirdest thing for a guy that technically talented to not be into foss Dec 08 01:12:31 Not at all Dec 08 01:12:59 Windows was the choice OS of the professional engineer OS until... 5 years ago? Dec 08 01:13:06 Just because that's what everything was on Dec 08 01:13:18 huh Dec 08 01:13:29 it's been about 7 years since i've used it Dec 08 01:13:36 although one day i'll get a mac Dec 08 01:13:39 and maybe a windows too Dec 08 01:13:44 maybe like one of those intel usb sticks that has windows on it Dec 08 01:13:55 It was an uphill struggle to get my lab automation stuff working on Linux, 4 years or so ago? Despite the existence of pyvisa Dec 08 01:14:13 Wouldn't work with Matlab either Dec 08 01:14:38 yeahhh i've never used any of that Dec 08 01:14:43 i mean in school i did Dec 08 01:14:47 labview, visa, matlab Dec 08 01:15:08 I was working in semiconductors Dec 08 01:15:17 oh yeah i did in undergrad Dec 08 01:15:20 So, we needed to, automation is a big thing in that space Dec 08 01:15:22 not on silicon except as a substrate Dec 08 01:15:30 we didn't automate anything :-( Dec 08 01:15:31 Also ugh labview Dec 08 01:16:03 where's the money Dec 08 01:16:12 ?? Dec 08 01:16:20 In what? Semiconductors? Dec 08 01:16:39 Because if that's the question, then the answer is Taiwan. Dec 08 01:16:54 Why do you think I don't do semiconductors stuff anymore Dec 08 01:17:11 <\dev\cache> Hi does anyone here use an X15?, For some reason when i download large files my beaglebone heats up the entire board and shuts down Dec 08 01:17:16 <\dev\cache> is it because im using the OS from emmc? Dec 08 01:17:22 ???? Dec 08 01:17:36 No, that shouldn't be using enough power to brown our Dec 08 01:18:32 No, that shouldn't be using enough power to brown out the board, sounds like maybe a thermal shutdown? That's a question for zmatt Dec 08 01:18:43 gosh those backslashs on your name really bother me Dec 08 01:19:24 pretty sure all the money is in ios apps Dec 08 01:19:44 No, the money is in the hands of the rich Dec 08 01:19:54 ios apps Dec 08 01:19:57 Rise up and reclaim the fruits of your labors Dec 08 01:20:43 Something something chains Dec 08 01:21:15 i mean i'm for it as long as i don't have to do any work Dec 08 01:21:29 Anyway, ayjay_t fair warning, vivado, like ISE and quartus, is intended for windows Dec 08 01:21:37 \dev\cache: download using chat? Dec 08 01:21:57 They did, however, build the entire thing in Java, so it is more properly cross platform Dec 08 01:22:10 \dev\cache: download using what? Dec 08 01:22:12 (sorry) Dec 08 01:22:56 Anyway my commute is almost done so I'mma disappear again Dec 08 01:22:57 it does sound like thermal shutdown, but downloading a file has no reason to heat up the x15 Dec 08 01:23:00 Bye Dec 08 01:23:04 icebreaker seems promising... but I would like to get more logic cells and more pins Dec 08 01:23:32 Yeah CoffeeBreakfast you'll be hard pressed to find a particularly beefy oss capable fpga Dec 08 01:23:55 <\dev\cache> zmatt: I was downloading an OS Iso from the web Dec 08 01:24:01 8k slices is usually plenty for small projects, but that's why Luke is using the ECP5 for the TinyFPGA EX Dec 08 01:24:10 \dev\cache: using? Dec 08 01:24:18 Which will be neat Dec 08 01:24:35 Anyway now a actual bye Dec 08 01:24:37 <\dev\cache> zmatt: the default browser. Dec 08 01:24:46 hedgeberg: yeah luckily uh, keelin is the fpga guy and also the windows lover Dec 08 01:24:54 ohhhh Dec 08 01:26:27 \dev\cache: oh, I'm sure chrome is just being insane. it seems nowadays web browsers are among the heaviest applications you can run on a computer :P Dec 08 01:26:50 <\dev\cache> zmatt: you suggesting i should use fetch or wget to grab the iso? Dec 08 01:27:37 <\dev\cache> zmatt: sort of silly, because I used to use Raspberry Pi Model B, and I could download these very same ISO's using its less powerfull processor, easily.. Dec 08 01:27:48 <\dev\cache> i dont understand why the beagleboard would shutdown from just a download.. Dec 08 01:28:00 <\dev\cache> Im also wondering at this point if I got the fan if it would be any help or good at all to begin with. Dec 08 01:29:46 I'm testing a download right now to see what happens Dec 08 01:44:49 <\dev\cache> zmatt: I was graabig a 4GB iso FWIW Dec 08 01:44:49 cpu heated up from 57-58 to 61-62 but doesn't increase any further **** BEGIN LOGGING AT Sat Dec 08 01:45:08 2018 Dec 08 01:47:41 <\dev\cache> ok lets try it Dec 08 01:51:02 <\dev\cache> zmatt: can i flash the sdcard from debian on beagleboard>? Dec 08 01:51:22 <\dev\cache> i.e can i just download it flash the card make changes and reboot, or does the flashing of sd card need to happen on a seperate pc? Dec 08 01:53:40 wouldn't you run into the overheating issue again while downloading the image? also, do you even have enough free space? Dec 08 01:54:03 wait, or are you booting from sd card? Dec 08 01:54:19 <\dev\cache> booting from emmc Dec 08 01:54:39 I just realized you said you were downloading a 4GB iso, but that wouldn't even fit on internal eMMC Dec 08 01:55:03 nor will the debian image fit in your free space Dec 08 02:12:40 ooh CoffeeBreakfast yeah i'm interested in getting one Dec 08 02:14:55 ayjay_t: as lab instrument is pretty bad, but as FPGA dev board, could be great Dec 08 02:16:31 40 MHz bandwidth? :/ Dec 08 02:17:06 16k samples? they couldn't add a bit of ram to the board? Dec 08 02:22:29 Yes, the review of Dave from EEVBlog is very illustrative in that respect (all those downsides, as bench instrument) Dec 08 02:22:50 well i just kind of wanted something to mess around with for audio stuff Dec 08 02:23:12 dave eh Dec 08 02:23:16 he's gonna review one of my products one day Dec 08 02:23:18 and i'm not gonna be ready Dec 08 02:25:08 The rule of thumb: make sure your product works out of the box, in windows without much effort, and without registration required Dec 08 02:56:21 hmm Dec 08 02:56:35 do you have to ask him i wonder Dec 08 02:56:39 i guess you'd have to have a bit of traction first Dec 08 02:56:48 i honestly don't think hes really my target audience **** ENDING LOGGING AT Sat Dec 08 03:00:00 2018