#linuxcnc-devel | Logs for 2014-09-19

[12:27:37] <skunkworks> A week at 16us.. I think we are good
[12:30:16] <CaptHindsight> http://store.linksprite.com/pcduino3-nano/ $39 dual core 1GHz ARM in *duino form factor
[12:36:18] <Tom_itx> will the HDMI kill it for our purpose?
[12:43:18] <skunkworks> the device that jepler was playing with for spi had hdmi...
[12:44:49] <kwallace2> CaptHindsight, thanks for the pcDuino link. That looks perfect for a webcam application I'm thinking about.
[13:03:56] <seb_kuzminsky> skunkworks: Odroid U3, from hardkernel.com
[13:10:03] <CaptHindsight> kwallace2: Linuxcnc runs on it with preempt_rt
[13:10:49] <kwallace2> Cool.
[13:11:11] <CaptHindsight> hm2_eth builds but we haven't tested it yet
[13:14:25] <kwallace2> I was hoping to put a picture server inside my camera box: http://wallacecompany.com/E45/Q68_cam2-1a.jpg , camera -> USB -> server(gPhoto, ImageMagic) -> Ethernet -> WRT54 -> WiFi
[13:33:02] <jepler> ugh on the pcduino schematic, SPI0_CLK is level-shifted with a single-transistor inverter
[13:33:23] <jepler> sheet 15 in https://cdn.sparkfun.com/datasheets/Dev/PCDuino/PCDUINO3Schm.pdf
[13:36:13] <jepler> this doesn't make a lot of sense to me, as SPI0_CLK is a signal direct from the A20 chip on sheet 6 too
[13:36:30] <jepler> oh I read that backwards, duh
[13:36:39] <CaptHindsight> page #16
[13:37:18] <CaptHindsight> SPI is on a 6 pin header with some esd protection
[13:39:15] <CaptHindsight> IIRC all that IO is 3.3V on the pads
[13:41:33] <jepler> that would be convenient
[13:42:27] <CaptHindsight> ethernet MAC is integrated so hm2_eth might work
[13:44:28] <CaptHindsight> I have to do some soldering on the a20 board here since the 5V is flaky, we have problems booting from the sd-card and SATA gets corrupted
[13:48:21] <CaptHindsight> kwallace2: just so you're aware the GPIO on the PCDuino is not as fast as the AVR's on the Arduino
[13:52:42] <jepler> seb_kuzminsky: do the scratch-branch docs go somewhere? I didn't look too hard.
[13:53:21] <jepler> ah looked a little harder
[13:57:51] <KGB-linuxcnc> 03Jeff Epler 05jepler/proposed-2.6 62ca1b4 06linuxcnc 10docs/src/quickstart/stepper_quickstart.txt how do our platforms do on latexmath today? * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=62ca1b4
[13:59:33] <seb_kuzminsky> jepler: you probably already found it, but here you go: http://buildbot.linuxcnc.org/doc
[14:09:57] <jepler> well just from local testing on debian wheezy, even the supposed example from asciidoc.org doesn't work to put math in html
[14:19:16] <CaptHindsight> maybe I should mirror psha's camview packages since they might disconnect Russia from the net
[14:22:53] <jepler> it's a2x that is busted, not asciidoc itself.
[14:23:07] <jepler> this thread is years old but it seems to touch on the same problem: https://groups.google.com/forum/#!msg/asciidoc/QQF7NlhE5rs/F8NZhpfJKVQJ
[17:40:26] <memleak> jepler, any idea what to replace math.h with in RTAI? base/math/e_*.c
[18:06:23] <seb_kuzminsky> Ted Ts'o reiterates what makes a patch contribution mergeable: http://thread.gmane.org/gmane.comp.encryption.general/20735
[18:06:28] <seb_kuzminsky> the PS: is the TL;DR
[18:31:09] <CaptHindsight> http://emc2arduino.wordpress.com/downloads/hal2arduino/ is this worth trying to use?
[18:33:13] <CaptHindsight> heard of it but haven't tried it yet
[18:33:45] <mozmck> seb_kuzminsky: what does TL;DR mean?
[18:33:52] <CaptHindsight> I'd like to try controlling several hobby type RC servos
[18:35:27] <seb_kuzminsky> mozmck: too long; didn't read
[18:36:20] <mozmck> huh, not sure I understand what that means then relating to the PS.
[18:36:40] <seb_kuzminsky> CaptHindsight: looks like a neat hack, btu don't arduinos connect via usb? so you won't get good realtime performance on the communications like you would with a mesa fpga
[18:37:18] <seb_kuzminsky> mozmck: it means: if you don't want to bother reading the whole message, just read the PS:
[18:37:46] <seb_kuzminsky> CaptHindsight: if you want to plug in a bunch of non-realtime I/O, arduino seems like a fine way to go
[18:37:47] <mozmck> seb_kuzminsky: ah, I see :) good PS I thought - and whole post
[18:38:11] <seb_kuzminsky> :-)
[18:38:24] <CaptHindsight> seb_kuzminsky: I don't need real time over USB, I just want to control some motors and sync to Linuxcnc with classic ladder
[18:38:33] <seb_kuzminsky> treating clueless contributors as hostile nsa spies is a new analogy for me, but it sort of makes sense
[18:39:14] <seb_kuzminsky> CaptHindsight: then it might be worth a try, i didnt read the code at all but the overall idea seems reasonable
[18:40:37] <CaptHindsight> seb_kuzminsky: I'm controlling a robot with Linuxcnc, but need some RC servos to open and close things the robot moves things through
[18:40:38] <mozmck> best way for a hostile spy to act - clueless contributor!
[18:48:56] <KGB-linuxcnc> 03Sebastian Kuzminsky 052.6 15d8eba 06linuxcnc 10docs/src/quickstart/images/step-calc-inch-math.png docs: update stepper quickstart equation * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=15d8eba
[19:05:37] <jepler> memleak: with a rtai header that properly provides extern declarations for everything that rtai_math.ko provides? (this file doesn't exist yet afaik, but it would be in some respects similar to rtapi_math.h except without the arch-dependent asm magic)
[19:13:46] <memleak> - return log1p(t+sqrt(2.0*t+t*t));
[19:13:46] <memleak> + return __ieee754_log(t+__ieee754_sqrt(2.0*t+t*t));
[19:14:12] <memleak> that compiled... but it's probably wrong
[19:14:41] <memleak> this is over my head i think.
[19:18:36] <jepler> yes that's certainly wrong
[19:19:14] <jepler> log1p(x) is equivalent to log(1+x), except that it is more accurate when x is close to zero
[19:19:48] <memleak> i wonder how seb was able to build RTAI debs if i cant get past this stupid stubs-32 issue..
[19:20:23] <memleak> i could just modify /usr/include/stubs.h and that'd fix it (for me)
[19:21:18] <jepler> s_log1p.c has an implementation of log1p named log1p, so it's probably because of a lack of a proper declaration of log1p that you're encountering a problem
[19:21:28] <jepler> so you need a header which declares it
[19:21:37] <memleak> or modify mathP.h
[19:22:10] <memleak> http://dpaste.com/14BDJ5E
[19:22:27] <memleak> debian's userspace headers shouldnt be broken in the first place
[19:22:36] <jepler> they aren't broken
[19:22:51] <memleak> # include <gnu/stubs-32.h> is a nonexistant file
[19:23:07] <memleak> its in i386-linux-gnu/gnu/ not gnu/
[19:23:41] <jepler> $ cat are-stubs-broken.c
[19:23:42] <jepler> #include <gnu/stubs.h>
[19:23:42] <jepler> int main() { return 0; }
[19:23:42] <jepler> $ gcc are-stubs-broken.c
[19:23:42] <jepler> $
[19:23:49] <jepler> does this work for you?
[19:24:19] <jepler> I very much doubt gnu/stubs.h is broken; *including it from a kernel module* is never something that the linux kernel folks nor the glibc folks have defined as working
[19:24:36] <jepler> it may have worked for years by coincidence, but as you no doubt know that statement is true about a bunch of stuff in rtai that is better fixed properly
[19:25:11] <memleak> that .c file works fine.
[19:27:29] <memleak> +extern double __ieee754_log1p(double); to mathP.h and return __ieee754_log1p(t+__ieee754_sqrt(2.0*t+t*t));
[19:27:36] <memleak> is that right now?
[19:28:23] <memleak> if i had your skills this would be fixed in seconds
[19:28:52] <jepler> to me it looked like the source file base/math/s_log1p.c defined a function named log1p, and I did not see a file that defined a function named __ieee754_log1p
[19:30:02] <memleak> i added __ieee754_log1p to mathP.h
[19:30:10] <memleak> (just now for testing)
[19:30:13] <jepler> you added a *declaration*
[19:30:22] <jepler> the declaration is not so useful without a *definition*
[19:30:31] <jepler> however, I would expect modpost to diagnose the problem
[19:34:03] <jepler> (huh, log1p was not in C89)
[19:35:09] <memleak> jepler if you can fix a few .c files in base/math I can do the rest but its getting it started that i cant do, i hardly know any C
[19:36:08] <memleak> ill be more than happy to copypasta and sed the entire math tree once i have a baseline
[20:19:45] <jepler> memleak: I can't test it, but ..
[20:21:25] <jepler> https://github.com/NTULINUX/RTAI/pull/3
[20:31:21] <jepler> (like the pull request says, compile-tested but not run-tested)