#linuxcnc-devel | Logs for 2015-03-19

[00:02:57] <mozmck> You are supposed to be only able to get 100 mA (IIRC) though unless the device requests more power
[00:06:19] <pcw_home> It maybe that that only makes a difference if the port has fancy current sense hardware
[00:08:37] <pcw_home> 'nite
[09:26:41] <seb_kuzminsky> o
[09:27:27] <seb_kuzminsky> morning
[09:30:43] <skunkworks> o
[09:30:44] <skunkworks> morning
[09:30:51] <skunkworks> I know that song
[09:35:55] <seb_kuzminsky> i was singing a different song
[09:36:24] <seb_kuzminsky> o *canada*
[10:07:40] <skunkworks> PCW, have you seen or used TI lv1023a or lv1224b - seems to be serial lvds..
[11:54:12] <jepler> andypugh/memfrob: the one thread I've seen about pi2 and preempt-rt is that it freezes, and signs point at USB -- http://mid.gmane.org/3671201425478878%40web10j.yandex.ru
[12:10:45] <jepler> I did get a pi2 but I didn't use it for anything hard-realtime.
[12:11:07] <jepler> It's running a light alarm clock with a separate microcontroller to do phase control of the lamp
[14:18:55] <skunkworks> yay peter! http://www.cnczone.com/forums/tormach-pathpilot-/263130-hey-stole-pci-cards-35247-pathpilot-upgrade.html
[14:27:05] <seb_kuzminsky> yep, that's a good problem to have :-)
[14:50:44] <skunkworks> andypugh, did you see jeplers comments?
[14:53:58] <andypugh> On what and where?
[14:56:55] <skunkworks> andypugh, http://tom-itx.no-ip.biz:81/~tom-itx/irc/logs/%23linuxcnc-devel/2015-03-19.html#11:54:12
[15:34:13] <Tom_itx> which card is the Tormach using.. the 5I25?
[15:41:25] <mozmck> seb_kuzminsky: Hopefully I'm getting closer to getting a xubuntu 14.04 set up with a preempt-rt kernel. When I do I would be willing to set up a buildslave for that if you would like.
[15:42:15] <Roguish> hey all. is there a convenient way to monitor the latency while lcnc is running?
[15:53:30] <skunkworks> Roguish_shop: you can view thread times...
[15:57:35] <Roguish_shop> ok
[15:59:01] <Roguish_shop> i keep getting bit by the watchdog
[16:00:43] <andypugh> Ah. Oh.
[16:01:02] <andypugh> Well, I do see it freezing…
[16:01:23] <andypugh> But put that down to being my fault for kernel-building incompetence
[16:03:24] <skunkworks> Roguish: what hardware?
[16:04:47] <Roguish_shop> an intel D945gclf AND A 5I20
[16:05:25] <skunkworks> what is your servo thread?
[16:05:43] <Roguish_shop> acpi is off in the grub startup. now running the vesa vid driver.
[16:06:32] <Roguish_shop> servo thead is 1,000,000
[16:06:40] <skunkworks> any base thread?
[16:06:44] <Roguish_shop> no
[16:07:13] <skunkworks> what do you get when you run the latency test?
[16:07:22] <skunkworks> and do you run glxgears and such?
[16:08:21] <Roguish_shop> definitely less than 20000. usually around 15000
[16:08:57] <skunkworks> huh.
[16:09:12] <skunkworks> are you running a large classic ladder?
[16:09:29] <Roguish_shop> none
[16:09:59] <skunkworks> you get a watchdog bite - do you get a realtime error also?
[16:11:05] <Roguish_shop> yes
[16:11:12] <Roguish_shop> task 1
[16:11:51] <skunkworks> how long did you run the latency tests?
[16:12:39] <Roguish_shop> 1 min, 3 glxgears and other stuff
[16:12:50] <Roguish_shop> sorry 15 min
[16:13:43] <skunkworks> wow - that is a single core atom..
[16:14:09] <skunkworks> I have run linuxcnc on the 330 version of that board - but a long time ago...
[16:14:21] <Roguish_shop> using lastest debian with updated master 2.8.0 pre1 551
[16:15:22] <Roguish_shop> didn'tused to get errors back with ubuntu and 2.5
[16:16:04] <skunkworks> is this rtai?
[16:17:27] <skunkworks> 2 things. You could try idle=poll in the kernel line.. Also - try rt_preempt.. (it is in the debian repository)
[16:18:00] <Roguish_shop> must claim i am not sure. from the hybrid iso
[16:18:30] <memfrob> i think i might have found what's causing such failure in my new dev branch of RTAI. proc / rtdm acts differently (again)
[16:19:44] <memfrob> thankfully linuxcnc removed the proc interfaces, at least for newer kernels
[16:34:10] <PCW> My experience is that even the dual core atoms are near the edge at a 1 KHz servo thread
[16:34:33] <mozmck> is that with rtai or preempt-rt?
[16:34:37] <mozmck> or both?
[16:36:05] <dgarr> Roguish: a method for measuring latency while running (requires a recent 2.7 checkout (rip or buildbot deb)), it uses the latencybins comp for the latency pin and hal-histogram to display, values are in cpu clocks: http://www.panix.com/~dgarrett/stuff/latency_while_running.jpg
[16:36:49] <cradek> there's also a hal pin output from motion that gives time since last run (I think?)
[16:38:16] <Roguish_shop> i will look into boh of those, pus look or another stronger inexpensive SBC. any suggestions?
[16:38:23] <Roguish_shop> plus
[16:40:24] <Roguish_shop> just finished my current machining project so i have a bit of time.
[16:41:00] <PCW> Both (servo thread jitter is pretty terrible if the servo thread does anything, the latency test does not show this ( since it doesnt "do" anything )
[16:43:33] <PCW> i just ran the latency-histogram ( thanks dgarr! ) on a dual core 1.6 GHz atom and its about 50 usec peak latency doing nothing (dispatch latency only)
[16:43:41] <PCW> (RTAI)
[16:44:17] <PCW> The J1800/1900s are pretty good
[16:44:22] <Roguish_shop> is that good or sucky
[16:44:36] <PCW> pretty sucky
[16:45:31] <PCW> probably OK for Mesa hardware with simple hal files but thats about it
[16:46:08] <Tom_itx> yep, i'm thinking about getting another J1900 ASRock
[16:46:20] <Tom_itx> probably a mini itx this time
[16:46:56] <Tom_itx> they're cheaper than the atoms were at the time
[16:47:00] <Roguish_shop> newegg asrock j1900 $75.99
[16:47:06] <Tom_itx> 679
[16:47:08] <Tom_itx> 69
[16:47:22] <Tom_itx> or 59 with 10 rebate now
[16:47:30] <Tom_itx> depending on which one you get
[16:48:43] <Roguish_shop> ECS brand at $55. any particular reason for ASROCK?
[16:48:44] <Tom_itx> http://www.cnczone.com/forums/tormach-pathpilot-/263130-hey-stole-pci-cards-35247-pathpilot-upgrade.html
[16:48:58] <Tom_itx> http://www.cnczone.com/forums/tormach-pathpilot-/263130-hey-stole-pci-cards-35247-pathpilot-upgrade.html
[16:49:10] <Tom_itx> it's what PCW suggested :D
[16:49:45] <PCW> I dont have a ECS is all, might be fine
[16:49:47] <Tom_itx> i got the Q1900B and am thinking about getting the Q1900M
[16:50:03] <cradek> > The flash doesn't matter because PP flashes it's own firmware when you start it.
[16:50:05] <cradek> orly?
[16:50:07] <Tom_itx> err vise versa
[16:51:01] <Roguish_shop> PCW: guess u r the leader of the pack.....
[16:51:11] <Tom_itx> fearless
[16:53:49] <jepler> hope the endurance of the 5i20's flash is high then
[16:53:54] <jepler> er whichever one it is
[16:57:52] <mozmck> maybe it checks to see what is there first?
[17:08:16] <Roguish> how about an Intel D2500HN, dual core
[17:12:57] <dgarr> PCW: fyi some results for an older HP7800 SFF (ebay jun14): screenshots for preempt-rt 3.2 and 3.18 with/without isolcpus: http://www.panix.com/~dgarrett/stuff/7800/
[17:56:50] <KGB-linuxcnc> 03Dewey Garrett 052.6 627bb49 06linuxcnc 10configs/sim/axis/xhc-hb04/README 10configs/sim/axis/xhc-hb04/xhc-hb04.tcl xhc-hb04 sim configs: typo fix * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=627bb49
[17:57:43] <KGB-linuxcnc> 03Dewey Garrett 052.7 33bad26 06linuxcnc 10configs/sim/axis/xhc-hb04/README 03lib/hallib/xhc-hb04.tcl Merge branch '2.6' into 2.7 * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=33bad26
[18:11:37] <PCW> dgarr: strange, wonder what the difference is ( vs freeby.mesanet.com/7800sff.png, other than slightly different CPU )
[18:13:02] <wholepair> I have setup 2 steppers in a H-Frame config and am using core-xy to drive it. How do I get jog mode to command movement in cartesian using kinematics instead of joint mode?
[18:13:05] <PCW> I think(and hope) they only flash it at install time
[18:13:23] <wholepair> TELEOP?
[18:13:37] <cradek> wholepair: yeah you have to switch to world/teleop mode
[18:13:51] <cradek> assuming your kins are right that'll do it
[18:15:05] <wholepair> cradek: when I run a program it looks right - and when I jog it shows diagnals for motion along a single axis which makes sense for a H-Frame - I am currently just testing with motors in free space -
[18:16:33] <wholepair> cradek: so for TELEOP do I put that in my HAL file, INI file, or run a hal command while running the machine. I have never used linuxcnc at the level and these things are new to me -
[18:16:51] <cradek> wholepair: usually you'd just poke $
[18:17:04] <cradek> switching between world and joint mode is on a menu
[18:18:16] <andypugh> Those RPi guys are no help…
[18:18:35] <cradek> andypugh: your battle sure seems uphill with that thing
[18:19:35] <andypugh> Would #include <linux/netfilter_ipv4/….> look in include/linux/netfilter_ipv4 in a typical kernel build of a git repo?
[18:20:09] <cradek> sorry I have no idea
[18:20:29] <seb_kuzminsky> wholepair: you said you're using core-xy to drive your machine, is that something you wrote?
[18:20:35] <andypugh> cradek: No worse than all my previous failures at building kernels for x86, Udoo, or anything else for that matter.
[18:21:19] <andypugh> I just assume that building kernels is nearly imposible, and I have so far not been disappointed.
[18:21:25] <cradek> yeah.
[18:21:40] <wholepair> seb_kuzminsky: no I got it from http://www.cnc-club.ru/forum/download/file.php?id=8251&sid=c6c9d08011061a1f26c3cb47458d65d2
[18:22:07] <seb_kuzminsky> neat!
[18:23:04] <seb_kuzminsky> if you have a working kins+config, i'd love to include it in linuxcnc as a sample config
[18:23:32] <cradek> yes!
[18:23:38] <seb_kuzminsky> there's a guy at my local hackspace who's working on an H robot, this would help him out
[18:23:39] <KGB-linuxcnc> 03Dewey Garrett 052.7 4daf65b 06linuxcnc 10lib/hallib/xhc-hb04.tcl xhc-hb04.tcl: err_exit for missing inifile stanzas * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=4daf65b
[18:24:47] <wholepair> I am still looking for the menu switch to change between world and joint - it's not on the axis gui - I think I missunderstood
[18:24:49] <cradek> how do you home yours?
[18:25:16] <cradek> oh it doesn't exist yet, right :-)
[18:25:33] <wholepair> cradek: right :-) spinning in free space now -
[18:28:09] <wholepair> I found it -
[18:30:54] <wholepair> cradek: poke$ - I get it now - thanks I am sure I will be back when I start working on the homing
[18:33:41] <seb_kuzminsky> cool
[18:37:33] <wholepair> the latest videos from pocketnc using fusion360 to cut a duck call are pretty sweet - anyone seen them?
[18:54:39] <andypugh> cradek: I am just exploring the possibility that this is a really strange problem related to the fact that my git repo is actually on my Mac.
[18:58:40] <mozmck> so is there a way to test latency with preempt-rt without some hardware to talk to?
[18:59:21] <memfrob> latency-test talks to the APIC timer, if that counts as hardware
[19:02:00] <mozmck> I need to test this on a different computer I think. The nvidia driver won't install with preemt-rt, so it booted with vesa which does not have glx it says.
[19:03:30] <mozmck> zlog
[19:06:27] <mozmck> Is there an example for using the latency pin and hal histogram that dgarr mentioned? I'm also not sure I understand the output.
[19:08:07] <dgarr> mozmck: see the halcmd example in the image: http://www.panix.com/~dgarrett/stuff/latency_while_running.jpg
[19:09:28] <mozmck> oh, I missed that - thanks.
[19:09:59] <mozmck> What do the numbers on the left mean? I'm not familiar with reading histograms.
[19:12:44] <dgarr> the latencybins comp latency pin is (now - last_timer - period), the histogram horizontal axis are bins for the values of the input, the vertical axis is in counts, so the latency (jitter) mean is zero and the 'width' is measured by the standard deviation (in simplest terms), the min and max values in the observation interval are shown too
[19:14:05] <dgarr> the wikipedia page for histograms may be helpful
[19:16:18] <dgarr> hal-histogram is very recent (hal-histogram --help for options) it may be useful for looking at distribution of a hal component's calculation time (.time pin) for example
[19:22:52] <Roguish> tom_itx: hey, are there any problems with Linux and the ASRock Q1900B ?
[19:23:33] <Tom_itx> i booted linux with it but have been configuring a hdd with windows
[19:23:50] <Tom_itx> haven't had a chance yet to try it with the mill
[19:24:21] <Roguish> but to your experience, it runs linux OK, like Wheezy?
[19:24:22] <Tom_itx> part of the stuff was a week late arriving due to the bad weather
[19:24:58] <Tom_itx> i was running 10.04 but iirc i booted wheezy with it but didn't do much with it
[19:25:35] <Tom_itx> if windows update ever gives it back to me i plan to try a few things
[19:26:10] <Tom_itx> been doing updates since early afternoon...
[19:29:01] <Roguish> that's ok, you'll get windows 10 for free.......
[20:03:17] <andypugh> Well, that is one rather interesting quirk of my decision to keep all my git archives local to my keyboard and chair…. OSX is case insensitive, and some linux kernel folders have files distinguished only by case, so I lose one…
[20:03:51] <andypugh> How, err, fascinating.
[20:04:55] <mozmck> I'm surprised by that - is BSD case insensitive?
[20:06:33] <andypugh> _you_ Are surpised? I am astonished. And annoyed
[20:07:15] <mozmck> I don't know much about macs, but heard they are based on BSD these days.
[20:07:33] <andypugh> Yes, that is why I am very susprised by this quirk
[20:07:50] <andypugh> Typically OSX and Linux feel very much the same
[20:24:33] <skunksleep> pcw_home: the rt is up patch 5
[20:25:32] <skunksleep> I was applying it to the Asus 1900 but ran out of time
[20:31:58] <PCW> Yeah I noticed but have not tried yet either
[20:33:41] <PCW> I did find that I can run the 7I92 from the laptops USB power so I have a pretty portable linuxcnc setup (+ ISE so I can now make bitfilles at home)
[20:38:36] <PCW> Arghh I feel bad the Marius's problem was old firmware and I didn't know there was any issue with Ethernet firmware versions
[20:39:30] <PCW> Driver needa to warn to update firmware
[20:39:35] <PCW> needs
[20:40:19] <PCW> or muddle through without hanging
[20:40:40] <PCW> BBL Dinner!
[21:26:43] <cradek> bsd is absolutely not case-insensitive unless someone does something very strange to it
[21:26:56] <cradek> andypugh: glad you figured it out, anyway
[21:30:09] <andypugh> Well, I figured _something_ out, no guarantees that is all that is wrong.
[21:31:17] <andypugh> And as is traditional when building kernels, it is now 2am and I just started again.
[21:32:14] <cradek> :-P
[22:27:23] <memfrob> the changes to rtdm / proc and lxrt inlining still get me nowhere. :/
[22:28:11] <memfrob> its definitely magic deep within rtai. also another thing that's weird is that there are more warnings during the build, but in my tree, the warnings aren't there, so certain CFLAGS must be changing.
[22:28:33] <memfrob> ^between magma and my dev branc
[22:28:37] <memfrob> h
[23:14:47] <KGB-linuxcnc> 03Dewey Garrett 052.7 58de619 06linuxcnc 10src/hal/components/latencybins.comp latencybins.comp fix ref to using script name * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=58de619
[23:14:47] <KGB-linuxcnc> 03Dewey Garrett 052.7 bc20bf9 06linuxcnc 10scripts/hal-histogram hal-histogram: minor display improvements * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=bc20bf9
[23:43:47] <linuxcnc-build> build #1256 of 1405.rip-wheezy-armhf is complete: Failure [4failed compile runtests] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1405.rip-wheezy-armhf/builds/1256 blamelist: Dewey Garrett <dgarrett@panix.com>
[23:52:18] <linuxcnc-build> build #3079 of 0000.checkin is complete: Failure [4failed] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/3079 blamelist: Dewey Garrett <dgarrett@panix.com>