IRC logs for #automotive for Tuesday, 2016-12-06

*** JordonWu has quit IRC00:55
*** JordonWu has joined #automotive00:58
*** mdurnev has joined #automotive01:33
*** AlisonChaiken has quit IRC03:04
*** waltminer has quit IRC03:33
*** rdale has quit IRC03:34
*** rdale has joined #automotive03:34
*** AlisonChaiken has joined #automotive05:00
*** ashwasimha_ has joined #automotive06:14
*** praneeth has quit IRC06:40
*** praneeth has joined #automotive06:40
*** ctbruce has joined #automotive08:16
*** fredw has joined #automotive08:19
*** jlrmagnus has joined #automotive08:20
*** rajm has joined #automotive08:24
*** toscalix has joined #automotive08:41
*** NathanI__ has joined #automotive08:51
*** NathanI_ has quit IRC08:51
*** kbroulik has joined #automotive09:02
jbocklage123:)09:18
*** fredw1 has joined #automotive09:24
*** fredw has quit IRC09:26
*** fredw1 is now known as fredw09:26
*** JordonWu has quit IRC09:44
*** jlrmagnus has quit IRC09:53
*** jlrmagnus_ has joined #automotive09:53
*** ctbruce has quit IRC10:32
*** CTtpollard has quit IRC11:08
*** CTtpollard has joined #automotive11:11
*** sanjeev has quit IRC11:24
*** caiortp has joined #automotive11:32
*** JordonWu has joined #automotive11:46
*** fredw has quit IRC11:49
*** fredw has joined #automotive11:50
*** fredw has quit IRC11:59
*** fredw has joined #automotive11:59
*** jlrmagnus_ has quit IRC12:11
*** mdurnev has quit IRC12:17
*** fredw has quit IRC12:29
*** fredw has joined #automotive12:41
*** jlrmagnus_ has joined #automotive12:51
*** waltminer has joined #automotive12:59
*** fredw1 has joined #automotive13:00
*** fredw has quit IRC13:02
*** fredw1 is now known as fredw13:02
*** JordonWu has quit IRC13:31
*** ToxicGumbo-work has quit IRC14:17
*** ToxicGumbo-work has joined #automotive14:24
*** jlrmagnus_ has quit IRC14:34
*** jeremiah has joined #automotive14:49
ToxicGumbo-workWhen I look at event announcements like the following….14:50
ToxicGumbo-workhttps://wiki.automotivelinux.org/doku.php?id=agl-distro/dec2016-f2f14:50
ToxicGumbo-work…I'm left wondering about the timezone.  Is 10 am - 6 pm relative to Yokohama?14:50
CTtpollardI would presume so surely?14:50
*** ctbruce has joined #automotive14:51
*** ctbruce has quit IRC14:52
*** ctbruce has joined #automotive14:54
jeremiahHmm, gunnarx around?15:50
CTtpollardI've not seen him online today jeremiah15:51
jeremiahCTtpollard: Thanks15:52
jeremiahI'll try and reach him via email15:52
*** kbroulik has quit IRC15:59
*** kbroulik has joined #automotive15:59
smurrayToxicGumbo-work: yes, the f2f is 10 - 6 in Yokohama time16:05
*** chbae has joined #automotive16:08
NathanI__I'm trying to compile the agl-demo-platform from stable branch Blowfish 2.0.3 for the raspberry pi 3. But I'm getting a "no target" error for xproto-native/1_7.0.28-r016:16
NathanI__Any ideas?16:16
jeremiahHmm. I haven't seen that error, sorry.16:17
CTtpollardNathanI__: can you paste the full error into a pastebin?16:18
NathanI__pastebin, that's the one! trying to use pastebucket... it's broken16:20
CTtpollardplenty out there :)16:20
NathanI__http://pastebin.com/z7sCHUCz16:21
CTtpollardNathanI__: as a starting point for any recipe failure I recommend doing bitbake clean & cleansstate on the package16:23
NathanI__-c cleanall? did that16:24
CTtpollardcleanall is one step further yes16:25
NathanI__hmm, although now that you saw that, i did cleanall xproto-native and not simply xproto, seems to be working now16:26
NathanI__oops, care to explain the difference between the -native and the non native?16:26
CTtpollardnative is a host package, non native is deployed to target16:27
NathanI__why is the host package required?16:28
CTtpollardas a build dependency probably16:29
NathanI__ok, i guess i could see that. Thanks CTtpollard for the help!16:30
CTtpollardIt's just a method to ensure your build host has the right package & versions without having them as explicit bitbake dependencies16:30
furythat's pretty cool. so not only is it compiling the embedded code itself, but it's literally compiling the tools it needs to use to compile them16:32
*** ashwasimha_ has quit IRC16:33
furyI do still have to do some apt-get installing to get an out of the box Ubuntu 16.04 setup to compile it tho16:33
furyi guess there are some dependencies you kinda have to have on the host system before it can start compiling the rest :P16:33
furySkynet can't build itself yet16:34
NathanI__yeah, you need some python and gcc at min. But I see that it's compiling an older version of python right now, haha16:34
furythe whole process takes about 45-50 minutes on my new build machine - 20 cores at 2.2 ghz and an NVM.e SSD16:35
fury32 gigglebytes of ram16:35
NathanI__nice... i got the RAM, but only 6 cores, haha. Takes a few hours for me.16:36
CTtpollardfor instance, quilt-native is the first package built by bitbake and is patched via your host git. After that bitbake defaults to quilt native16:36
furyit spends quite a bit of time at idle tho, so i just fire up 2 or 3 in parallel when i need to do any rebuilding (build for porter, pi, and wandboard all at the same time)16:36
NathanI__i'm also compiling inside a virtual machine though, which slowes things doesn considerably.16:36
furyyeah, virtual machine isn't all that great at handling multiple cores in my experience16:37
NathanI__speaking of which, it freezes on me regularly. The likely cause of that last error. Anyone have a suggestion?16:38
furyi was half tempted to order a second drive to just dual boot my primary desktop to linux when they said we should just get a new machine for the big build jobs16:39
NathanI__I have 32GB, Xeon E5-1650. I dedicated 16GB and 6 cores to the virtual machine. Maybe I should only dedicate 5?16:39
furythe thing with virtual machines and multiple cores is, as far as i know, they don't properly know how to handle hyperthreading, so unless you can bind it specifically to physical cores, it actually slows things down to have it try to emulate multiple cores16:40
NathanI__it is definately faster than when I had it set to 1, but the stability is a problem.16:42
furyhm16:42
furymight be worth limiting it to 516:42
furymoar ram the better, too, btw, things like qt are huge memory hogs16:43
furyas far as compiling goes16:43
furymaybe the vms have figured out how to do multicore... i didn't get far in my experimentation, i had ample numbers of machines to try it on natively so i freed up the 80 gigs on my primary desktop that i had been using for the VM16:44
fury"only" a 256 gig SSD in that one, so space is at a premium16:44
NathanI__I think I will switch to native, but I need to keep one PC on windows unfortunately. Compiling on an SSD would probably be a big boost as well. I don't have enough room on my SSD currently.16:47
CTtpollardSSD vastly improves the process, as does sharing sstate & downloads/mirrors across build targets16:48
*** mdp has quit IRC16:50
*** mdp has joined #automotive16:50
*** rajm has quit IRC16:50
furyI share a single downloads directory but my understanding of sstate is that it's unique to the build target (so that which was built for Pi cannot necessarily be used on porter, etc)16:51
CTtpollardfury: to some extent yes, especially for deployed packages. But a lot of the native build env packages are shareable (if same yocto version)16:52
furyah. so if my site.conf were modified to put all sstate in a single directory instead of a per-machine directory, I might have less duplicate building going on?16:53
furyI might give that a shot and see how it goes. The machine is sitting there idle right now anyway so doesn't hurt to try16:53
smurrayfury: you may get some reduction for the -native targets built to bootstrap things. I think the arch's for the various boards are different enough that you won't see much beyond that16:54
furyMakes sense16:54
furyI wonder, if I'm doing 3 in parallel, and one of them finishes one of the shareable native builds, do the others automatically pick up on that or do they figure that at the time of starting the bitbake (parsing the recipes)16:56
dl9pffury: its exactly as smurray said,  our board cflags are not unified enough to profit from a global sstate-cache.16:56
furyyeah16:57
dl9pfthats why I recommended the -per machine folders ... you can run multiple builds for different machines at the same time with that. just not the same machine in parallel16:57
smurrayfury: I believe building multiple users of the same sstate-cache at the same time is a no-no16:57
dl9pfaccess to the sstate-cache needs to be atomic, not shared.16:57
furyAh16:58
dl9pfbut you can have a sstate-mirror prepopulated and share known good sstate through that while still having per-build or per-board sstate-cache16:58
dl9pfso if you want to try out the sstate-mirror, just mv (or copy) a sstate-folder and point sstate-mirror to it.16:59
dl9pfthat way you can have pre-built sstate and the "atomic" access to the "local" sstate-dir17:00
furyCool. Maybe it doesn't matter too much, I'm just looking to shave off as much as I can and whenever there's a few minutes to be saved is valuable. Even with all that hardware it's still a good chunk of time spent waiting on the bitbake every time something changes early on in the dependency chain17:01
dl9pffury: absolutely ...17:01
dl9pfI use exactly that combination above.  sstate-mirror + local sstate-cache17:01
furySounds good :D17:02
dl9pfit works very well for 'stable' releases.  less efficient for a master where tthings change daily.17:02
*** chbae has quit IRC17:03
furyI read some interesting posts about yocto bitbakes with massive numbers of cores (64)...it really kinda doesn't improve after a certain amount because of the dependencies17:05
dl9pfyou have like 2 big waves ... -native (aka toolchain)  and the target files  plus big stuff like qt  that tends to run serialized and thus act as bottleneck17:06
dl9pfin total like 4 waves and some bottlenecks is pretty usual also for other builds/buildsystems17:07
smurrayqtwebkit is a bit of a pita, bottlenecks things badly17:07
dl9pfabsolutely, but thats webkit / chromium  fault as it has the "a build all" approach hardcoded in the makefiles17:08
furyYeah. The amount of time it spends in the configure process just sitting there idle, especially when it's the only step running...maddening :P I saw someone try a configure cache for a similarly annoying long wait in a configure step, but I don't think it proceeded out of the experiment stage17:09
furyI would say out of the 50 minutes, it's only pegging all 20c/40t half the time17:10
furyI keep meaning to time it to see exactly17:11
*** NathanI__ has quit IRC17:15
furyall those threads will come in super handy for my own CIAT tho17:15
furyI'd be thrilled if I could come up with something that builds it automatically, loads it into an SD card, installs the SD card in the board, boots it up, and starts fiddling with the UI to make sure stuff works :)17:17
furyI wanted to lurk in the CIAT call this morning but was running late, had to get to the office17:19
furyand then an all employee meeting, so I missed the dev call too, boo17:19
furyhow far does the AGL Jenkins go? Does it actually flash a board and boot it up and whatnot or does it just build?17:21
*** AlisonChaiken has quit IRC17:29
*** toscalix has quit IRC17:51
murphy sup fury17:52
furynot much, stuffing my face, how's it hangin?17:53
murphystill waiting on trying to get my TX1 discount17:53
murphymight just use a raspi in the meantime to start playing with AGL17:54
murphyi have all the ideas i want to try17:54
furyYeah, I'd start with the Pi. Haven't heard of anybody using Nvidia stuff on AGL yet.17:54
murphythe elinux page has some examples but i think someone here said they are hit or miss17:55
radiofreeit'll be a walk in the park compared to the renesas boards17:55
murphyi just wish the raspi was a bit more powerful17:56
murphythe 3 was pretty weak compared to the cubox and hummingboard specs that had been out for a while once it was released17:56
murphyoh i guess it was the TK1 that was used with AGL17:58
furyI guess I'm having a hard time imagining quad core 1.2ghz not being enough. Then again, for the longest time, all I had was a single 120 MHz core and 256 KB of RAM :P I'm swimming in gigglehertz now17:58
murphymainly it was the 1GB of ram17:58
murphyi think it should be 2GB min of youre using X17:58
murphymy copy of pentesting for carhackers should arrive today17:59
furymaybe I'm not thinking big enough. Self driving tech needs some pretty beefy processing...but I'm guessing motorcycles/powersports are in no danger of becoming self driving any time soon...people buy them to enjoy driving them17:59
murphyi want to replace my jeep's head unit and cluster and add image recognition among all the other little things i want to have18:01
murphyTX1 is pretty future proof as of today for SBC's18:01
murphyif i cant get a discount ill get the new minnowboard18:02
murphydigitial cluster with nav, bluetooth connectivity with tables and phones for media etc, car to car communication, SDR, autostart, maintenance reminders18:04
murphy*tablets not tables18:05
murphyadd a ZED to the front18:07
murphyhttps://www.stereolabs.com/18:07
murphyor maybe the back first as my reverse camera for sensing objects18:08
murphyshits expensive though18:08
*** AlisonChaiken has joined #automotive18:08
murphyi should see if there is some car hackers meetup i the SF bay area for help18:09
furyyou can autostart? I thought that kinda stuff had to be baked in at the factory18:10
murphyit takes some CAN bus hacking but its totally possible18:12
murphyhttp://www.canhack.org/board/search.php?keywords=remote+start18:14
murphyso i have a long path a head of me but im hoping i get atleast something done18:17
murphythis book is awesome - http://opengarages.org/index.php/Main_Page18:31
*** fredw has quit IRC18:37
*** smiller has joined #automotive18:37
*** waltminer has quit IRC19:10
*** ctbruce has quit IRC19:32
*** kbroulik has quit IRC19:51
*** Zebs has joined #automotive19:55
*** Zebs has left #automotive19:55
ToxicGumbo-workmurphy: The author sits in #opengarages19:56
ToxicGumbo-workCTtpollard & smurray: Thanks for the confirm.19:57
*** caiortp has quit IRC20:14
*** rdale has quit IRC20:18
*** NathanI_ has joined #automotive21:13
*** thaytan has quit IRC22:05
*** thaytan has joined #automotive22:06
*** chbae has joined #automotive22:07
*** thaytan has quit IRC22:15
*** thaytan has joined #automotive22:18
murphyoh nice22:22
murphythanks ToxicGumbo-work22:22
*** waltminer has joined #automotive22:32
furywaltminer: what happens if I hit the accept button on all these AGL meeting invites? Is it gonna spam the mailing list with "Accepted: " emails or does it filter that stuff out?22:42
furytrying to wrangle my calendar app into being something actually useful hehe22:43
waltminerIt filters them out or they just go to me.22:43
furyah ok22:43
furyprepare for incoming spam22:43
*** chbae has quit IRC22:43
furywow, the iPhone's calendar app kinda sucks. Every time I accept one of those pending invites it scrolls to the top of the list of events O_o22:45
furyLike, several times. I keep hitting the Today button and it keeps snapping back lol22:45
murphyyeah ios is stupid with calendar stuff22:45
murphyi have to deal with it all the time for work invites22:45
murphyis AGL a port of yacto?22:48
JEEBAGL is a distro that bases on openembedded/bitbake23:02
NathanI_Sorry guys, have another error.23:21
NathanI_http://pastebin.com/GcFgdsSi23:21
NathanI_I tried cleaning qtbase, no joy23:21
murphyhow is working with Qt as a newb?23:33
murphyim planning on using it for my cluster23:33
JEEBQt Quick with QML seems rather simple to prototype stuff in23:34
JEEBand then you just use your C++ functions to drive the data in it23:34
murphyyou can use python with the UI builder right?23:35
murphyi dont know C++23:35
murphyi just want to make some dials and guages from CAN bus data23:35
JEEBthe actual QML stuff has js as the scripting language, and there are python bindings for Qt that can utilize those things methinks23:36
NathanI_hmm, FYI, I appended my local.conf file with the troubleshooting steps here and it seems to have solved my problem: https://wiki.automotivelinux.org/agl-distro/agl-raspberrypi23:36
JEEBpyqt is gpl/corporate licensed and pyside was the LGPL alternative that nokia used to be doing... I don't think they got too far with that23:36
murphyyeah this is all just vapor ideas for me till i get some hardware23:37
JEEByeah... pyside's site still notes it supporting Qt 4.823:37
JEEBso for any newer Qt you'll need to be doing pyqt if  you want python23:37
murphyon the Qt site there is a difference between application and device,23:39
murphyAGL would be using qt as software right? since its not an actual qt OS or whatever they consider Qt for hardware23:39
murphythis is all new for me so sorry for the newb questions23:40
murphyive been messing with embedded for a while but ive never actually built anything23:40
murphyso many possibilities, so cool23:40
JEEBjust saying that PyQt is either under GPLv3 or a commercial license, so if you are going to be using it, just make sure that in your use case you can follow that license23:42
JEEBor one of those licenses23:42
JEEBwhat license Qt itself is under is a separate discussion23:43
murphyim just a hobbiest and everything i do will be on github23:44
JEEBalright, that makes it simpler23:44
JEEBjust follow GPLv3 and you're golden :P23:44
murphyit would be nice to make money off this but i dont see it being really profitable making custom dash clusters23:49
murphyif anything maybe since im doing it for a jeep and that world has a lot of after market stuff, i can sell a kit with the hardware23:50
murphyand the code is all free open source23:50
murphythats thinking crazy far ahead though... right now my best case is just connecting to my jeep and pulling some extra info from and displaying it on an LCD23:51
murphywhile i would love to create a whole integrated system with mapping and integrated nav, cluster, media car to car comms im just one guy with a dream23:53
*** smiller has quit IRC23:59

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!