#linuxcnc-devel | Logs for 2016-09-15

[09:20:55] <skunkworks_> zlog
[13:23:30] <andypugh> jepler: Have you any useful insights on: https://forum.linuxcnc.org/forum/10-advanced-configuration/31580-writing-modules-for-preempt-rt#80472
[13:25:43] <jepler> I don't have any interest in developing that kind of device. With preempt-rt, you can make any linux syscalls you want in RT but if it breaks you get to keep all the severed limbs.
[13:26:27] <andypugh> So, what he suggests ought to be easy, if he can find USB syscalls?
[13:32:17] <jepler> I think anybody who wants "a laptop and possibly a virtual machine [using] USB" needs to be an expert in their own right before starting
[13:48:23] <tinkerer> jepler: Hi Jeff, I would be pleased if you can test my image with your latest achievement.
[13:49:45] <jepler> my latest achievement? I haven't achieved anything of note yet with the pi3.
[13:49:55] <jepler> but I hope to download that image tonight and give it a quick try at any rate
[13:50:37] <tinkerer> the pi3 is the achievement :)
[13:52:43] <tinkerer> ok, suggestions and criticism are welcome
[13:57:18] <andypugh> jepler: Within the context of the limited demo I don’t think that the USB device is a terrible idea,
[14:15:45] <jepler> well, you could just put GRBL on the USB device. run a linuxcnc sim configuration, and your own userspace component that samples the position and writes "G0X%fY%fZ%f\n" to USB serial as frequently as possible.
[14:16:32] <jepler> .. awfully similar to what seb_kuzminsky hacked together for his robot arm, it is a stuttering mess and not a good impression of what linuxcnc is capable of
[14:16:46] <andypugh> That is the risk, I guess.
[14:17:09] <andypugh> An unflattering demo might be worse than no demo
[14:27:56] <cradek> I bet a smooth-running simulated machine with cool 3d graphics is better than a finicky and terrible hardware demo
[14:28:45] <andypugh> Although the demo is about making a machine run hardware, which is much cooler than just watching a screen.
[14:28:56] <cradek> I wonder what happened to ray's hardware demo boards
[14:29:03] <jepler> not that axis or vismach are cool 3d graphics by 2016 standards anyway
[14:29:08] <cradek> they had a little stepper, some switches and lights, an encoder, etc
[14:29:16] <cradek> iirc
[14:29:22] <jepler> best case they're gathering dust somewhere
[14:29:32] <cradek> probably so
[14:33:25] <CaptHindsight> Axis runs smoothly enough using software rendering on the multicore ARM soc's
[14:34:10] <CaptHindsight> 1 core for Linuxcnc with prempt_rt and the other 3-7 cores for software rendering
[14:34:19] <jepler> I noticed somebody posted an ITP for this, I've been meaning to see if it gets acceleration on the common ARM boards: https://github.com/lunixbochs/glshim
[14:38:38] <jepler> .. linuxcnc axis doesn
[14:39:02] <jepler> .. linuxcnc axis doesn't start with glshim, 'invalid enumerant' at the call glPixelStorei(GL_UNPACK_SWAP_BYTES, 0)
[14:49:45] <jepler> after a bit of hacking, you can see the lack of stippled lines and the lack of (or maybe black on black) DRO: https://media.unpythonic.net/emergent-files/sandbox/axis-glshim-hacked.png
[15:04:09] <andypugh> Looks black-on-black. It occludes the bounding box
[15:04:37] <CaptHindsight> memleak will look at it later, he just mentioned that llvmpipe for ARM might also work really well
[15:04:38] <andypugh> Am i right to beleive that linuxcnc.stat does not contain a reference to the current INI file path?
[15:10:28] <jepler> export INI_FILE_NAME="$INIFILE"
[15:10:40] <jepler> I think that's right, but the linuxcnc runscript does put the path of the inifile in the environment if I am reading it right
[15:11:13] <andypugh> Yes, i was just thinking that importing os into glcanon was possibly a step too far
[15:14:31] <jepler> "import os" is not particularly heavyweight in Python in general
[15:17:57] <andypugh> OK
[16:30:32] <andypugh> Yay! It is now possible to format the Axis DRO in hex. :-)
[16:38:52] <pcw_mesa> That's nice :-) now you just need a foot with 16 inches
[16:39:28] <seb_kuzminsky> ooh, Fern has been working on new features for Features and is nearing release!
[16:45:33] <KGB-linuxcnc> 03andypugh 05master 7253001 06linuxcnc 10docs/src/config/ini-config.txt 10lib/python/rs274/glcanon.py glcanon.py: Allow easy customisation of the DRO number format * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=7253001
[17:44:41] <jepler> the pi3's built in wifi is flaky with their standard kernel, sigh
[17:45:05] <jepler> I spent minutes and minutes grousing about the low speed of their package mirror, but no, it was hardware
[17:45:23] <jepler> the -rt kernel from machinekit just doesn't boot at all
[17:45:39] <jepler> just a test pattern on the screen
[17:46:09] <jepler> no serial console
[17:46:52] <pcw_mesa> thats better than most of my 4.6.x RT kernel attempts (usually nothing but a black screen)
[17:47:48] <jepler> this test pattern is the pi equivalent of a black screen
[17:48:41] <jepler> https://goo.gl/images/1ohcAf
[17:49:11] <pcw_mesa> Nice looking anyway
[18:23:28] <tinkerer> jepler: enable_uart=1 in /boot/config.txt
[18:24:26] <tinkerer> and console=serial0,115200 in /boot/cmdline.txt
[18:43:13] -linuxcnc-github:#linuxcnc-devel- [13linuxcnc] 15zultron opened pull request #163: `g10-l1-10` test updates (062.6...06dup-g10-test-2.6) 02https://github.com/LinuxCNC/linuxcnc/pull/163
[18:55:53] <CaptHindsight> memleak thinks it would be better and easier to just LLVMPIPE it using all the extra cores
[18:56:33] <CaptHindsight> unless you have a team of pros to write all the wrappers
[18:57:07] <CaptHindsight> the pandora only had a single core ARM plus GPU
[18:57:23] <CaptHindsight> now you get 8 cores for $5
[22:18:25] <jepler> tinkerer: I found the instruction I missed, didn't make all the right edits in /boot
[22:19:49] <jepler> seems the machinekit kernel deb doesn't have the corresponding source code in the normal debian way. in fact they don't seem to have source at all in their debian repo.
[22:20:07] <jepler> too bad that seems to be the ""easiest"" way to get a realtime kernel on this thing (same with bbb)
[22:34:20] <jepler> I wonder if I should do hm2_serial
[22:34:57] <jepler> this RS-422 connector on the 7i90 just stares up at me
[22:35:07] <jepler> saying "hook me to a serial port on one of these damned arm boards"
[22:35:12] <jepler> "maybe latency will be better than spi"