#linuxcnc-devel | Logs for 2017-02-08

Back
[09:55:01] <jepler_> wjw. https://www.seeedstudio.com/Nose-LED-Kit-p-1274.html
[10:05:34] <archivist> sneeze, one flying sticky LED
[10:10:45] <pcw_home> Well thats something I didn't know I needed
[10:15:15] <skunkworks> zlog
[10:45:30] <KGB-linuxcnc> 03Dewey Garrett 05aux_gladevcp_apps c4e6a83 06linuxcnc 10scripts/linuxcnc_var.in 10src/configure.in 10tcl/linuxcnc.tcl.in aux_gladevcp_apps: Intro * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=c4e6a83
[10:45:31] <KGB-linuxcnc> 03Dewey Garrett 05aux_gladevcp_apps 2b3a591 06linuxcnc 10(5 files in 3 dirs) pickconfig.tcl: support auxiliary example configs * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=2b3a591
[10:45:31] <KGB-linuxcnc> 03Dewey Garrett 05aux_gladevcp_apps 451c803 06linuxcnc 10docs/src/gui/gladevcp.txt 10lib/python/gladevcp/__init__.py gladevcp: support auxiliary applications * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=451c803
[10:45:33] <KGB-linuxcnc> 03Dewey Garrett 05aux_gladevcp_apps 7e7a3f3 06linuxcnc 10tcl/bin/pickconfig.tcl pickconfig.tcl debugging prints for test * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=7e7a3f3
[11:01:58] <seb_kuzminsky> the state of arm is super depressing: https://wiki.debian.org/InstallingDebianOn/NVIDIA/Jetson-TX1
[11:03:26] <jepler_> any remarks I make here about ARM may be used against me in the court of public opinion.
[11:04:17] <seb_kuzminsky> yeah :-(
[11:23:39] <pcw_home> None of the ARM server boards seem to be available _at all_ (Gigabyte, AMD, Cavium all have boards but they are either 1000's of $ or just unobtanium )
[11:24:05] <seb_kuzminsky> yeah
[11:25:19] <seb_kuzminsky> the debian.org folks were using APM arm servers for their build systems, but APM was just bought by a company that intends to get rid of the compute division
[11:25:22] <seb_kuzminsky> https://www.fool.com/investing/2016/11/28/applied-micros-x-gene-was-an-abject-failure.aspx
[11:28:32] <pcw_home> didn't mean to piss off martinjak but after Genes "simple" system setup
[11:28:34] <pcw_home> I'm not likely to recommend RPI/7I90 to anyone
[11:29:32] <seb_kuzminsky> i knew you were not trying to piss anyone off, and i think anyone who knows you knew it too
[11:29:59] <seb_kuzminsky> you can't please all of the people all of the time, no matter how good your intentions are :-/
[11:30:09] <seb_kuzminsky> occupational hazard
[11:31:10] <pcw_home> _is_ there decent ARM hardware for the buildbot?
[11:31:21] <seb_kuzminsky> if there is, i dont know about it
[11:31:58] <seb_kuzminsky> i'm currently using an Odroid U3, it gets random segfaults on maybe 5% or of the builds
[11:31:58] <pcw_home> what about the udoo? (it has SATA which is nice)
[11:32:32] <seb_kuzminsky> http://buildbot.linuxcnc.org/buildbot/builders/1405.rip-wheezy-armhf/builds/2901
[11:32:49] <seb_kuzminsky> i dont know anything about udoo other than that andypugh has recommended it
[11:34:45] <pcw_home> slower (quad 1G imx6)
[11:38:13] <seb_kuzminsky> yeah
[11:38:35] <seb_kuzminsky> 1GB isn't enough to build linuxcnc iirc, so it'd be swapping
[11:41:09] <jepler> 1GB may be OK without a desktop environment running too, but it's not enough to use "-j" parallelism in the build
[11:46:25] * skunkworks copies and pastes this conversation to the mailing list..
[11:56:19] <seb_kuzminsky> skunkworks: ok, i'll put my asbestos underpants back on
[11:56:37] <skunkworks> ;)
[11:56:48] <tinkerer> Wow! I'm really impressed. hasta la vista.
[11:56:54] <skunkworks> I rarely throw anyone under the bus
[11:57:04] <jepler> is there even a bus where you live?
[11:57:23] <skunkworks> school buses? yes
[11:57:33] <jepler> oh I always assumed it was about the city bus
[11:57:42] <jepler> but yeah I suppose you could throw someone under the school bus
[11:58:28] <skunkworks> If you go to lacrosse - they have city buses. you would just have to tie the person up and transport him about 20 minutes.
[11:58:57] <skunkworks> tinkerer?
[12:01:08] <skunkworks> Maybe I crossed the line of assholeness? Seemed all in fun.
[12:07:56] <archivist> andy did not recommend the udoo, he discovered the usb intermediate comms and gave up, although I read a bit of the udoo page and was confused about the comms "GPIO's can be accessed as Arduino pins, GPIO's or as additional S\PDIF, FlexCAN,I2S, SPI"
[12:08:46] <pcw_home> is it too painful to setup a cross build environment on X86 and just use test the resultant code on a connected arm platform?
[12:08:54] <jepler> The hypothetical machine for the build farm doesn't need to be capable of any I/O
[12:09:10] <archivist> but user expect :)
[12:10:35] <jepler> pcw_home: full system emulation is too slow. partial system emulation via qemu-user can't successfully run the testsuite. but yes if our goal was just to ensure that .deb packages were available, and not to run the testsuite, it's a possibility.
[12:11:09] <pcw_home> I was suggesting cross build and native test
[12:11:54] <jepler> sounds complicated
[12:13:17] <jepler> zlinuxcnc-devel-2016-05-24.log:20:11 <jepler> schroot is neat. I should have used it before now. with it and qemu-user-static, I can create a chroot on my x86_64 machine that identifies as being armhf (but only about 1/10 as fast, for that genuine arm-style experience :-P)
[12:33:52] <KGB-linuxcnc> 03Dewey Garrett 05aux_gladevcp_apps_v2 626ec87 06linuxcnc 10(5 files in 3 dirs) pickconfig.tcl: support auxiliary example configs * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=626ec87
[12:33:52] <KGB-linuxcnc> 03Dewey Garrett 05aux_gladevcp_apps_v2 da63f81 06linuxcnc 10docs/src/gui/gladevcp.txt 10lib/python/gladevcp/__init__.py gladevcp: support auxiliary applications * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=da63f81
[12:53:29] <seb_kuzminsky> i just realized tinkerer == matsche == W. Martinjak
[12:58:37] <skunkworks> oops
[13:11:16] <seb_kuzminsky> anyone know if the softiron overdrive 1000 is actually shipping?
[13:11:18] <seb_kuzminsky> https://shop.softiron.com/product/overdrive-1000/
[13:14:10] <seb_kuzminsky> this one uses that discontinued(?) x-gene arm64: http://b2b.gigabyte.com/Server-Motherboard/MP30-AR1-rev-11#ov
[13:15:06] <archivist> ew they fail on the about us page
[13:15:22] <pcw_home> you can't order the softiron without the case, who knows if they are in stock or not
[13:17:31] <pcw_home> another nice unavailable board: http://www.lenovator.com/product/103.html
[13:18:05] <seb_kuzminsky> there's sure a lot of tasty-looking vaporware in the arm server market
[13:19:24] <pcw_home> maybe cheaper(other than power bill) to emulate on a 2x20 core Xeon or some such
[13:20:14] <seb_kuzminsky> there's some problem with emulated arm atomics, i believe jepler looked into it and concluded it couldn't be done
[13:20:26] <seb_kuzminsky> we use atomics and barriers a lot, it's really important
[13:20:37] <seb_kuzminsky> maybe things have changed since then, or maybe i'm misremembering
[13:21:15] <pcw_home> sigh
[13:47:36] <jepler> qemu-user's emulation of atomic instructions only works within a single multithreaded program, but we rely on atomic instructions working across multiple programs that have the same shared memory segment mapped.
[13:47:59] <jepler> I think qemu-system does not have this problem, because like it says on the label it emulates the whole system; but performance is lower than qemu-user because of this
[13:48:09] <seb_kuzminsky> aha, thanks
[13:48:21] <jepler> (e.g., when you call the write syscall you're entering emulated code on qemu-system but you enter the native kernel on qemu-user)
[13:49:09] <seb_kuzminsky> the folks on debian-arm seem to like the softiron overdrive 1000, $600, https://shop.softiron.com/product/overdrive-1000/
[13:50:07] <seb_kuzminsky> https://lists.debian.org/debian-arm/2016/07/msg00080.html
[13:50:50] <seb_kuzminsky> that would let us run debian + kvm on the bare metal, and have armhf and arm64 virtual machines for the buildslaves
[14:23:11] <pcw_mesa> I can buy one right now if you wish
[14:44:11] <KGB-linuxcnc> 03Dewey Garrett 05aux_gladevcp_apps_v2 1f430bd 06linuxcnc 10(5 files in 3 dirs) pickconfig.tcl: support auxiliary example configs * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=1f430bd
[14:44:11] <KGB-linuxcnc> 03Dewey Garrett 05aux_gladevcp_apps_v2 d20bd86 06linuxcnc 10docs/src/gui/gladevcp.txt 10lib/python/gladevcp/__init__.py gladevcp: support auxiliary applications * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=d20bd86
[15:31:06] <seb_kuzminsky> that's very generous of you pcw_mesa
[15:31:31] <seb_kuzminsky> let me do some research and try to verify that it will work as well for us as i hope
[16:14:11] <pcw_mesa> one " review" http://blog.regehr.org/archives/1465
[16:23:31] <seb_kuzminsky> sounds about right: If you buy an OverDrive 1000, buy it because you want an ARMv8 machine that shows up ready to use and that isn’t an embedded systems toy like the Raspberry Pi family.
[16:26:58] <pcw_mesa> 8G of RAM should help also
[16:27:10] <seb_kuzminsky> yeah
[16:27:26] <pcw_mesa> no speed demon though
[16:28:02] <seb_kuzminsky> i think we'll want at minimum 3 guests: wheezy-armhf, jessie-armhf, and jessie-arm64, so that's comfortably 2.5 GB RAM and 1 CPU each
[16:28:43] <seb_kuzminsky> i'm guessing it wont be as fast as the 4 CPUs and 2 GB RAM of the Odroid U3, but if it's more stable it's probably still a win
[16:30:06] <jepler> hah looked into the "cello" board mentioned in the debian-arm post and found: https://linustechtips.com/main/topic/680521-still-broken-lemaker-cello-board-restarts-production/ [there's a PCI slot on it, but not even the CPU manufacturer or the board designer can figure out why it just doesn't work]
[16:30:32] <seb_kuzminsky> the cello is also not GA yet
[16:31:07] <seb_kuzminsky> ooh, comes with a free fan
[16:31:49] <jepler> does that mean they also didn't hit their thermal design goals?
[17:02:38] <skunkworks_> zlog