#linuxcnc-devel | Logs for 2014-03-11

[01:15:58] <linuxcnc-build> build #50 of 4017.deb-wheezy-armhf is complete: Failure [4failed apt-get-update shell_3] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/4017.deb-wheezy-armhf/builds/50 blamelist: Dewey Garrett <dgarrett@panix.com>
[14:41:04] <KGB-linuxcnc> 03Dewey Garrett 05dgarr/demo_appsv4 ce1debd 06linuxcnc 10(41 files in 10 dirs) configs/apps: new directory and .demo support * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=ce1debd
[14:42:13] <KGB-linuxcnc> 05dgarr/demo_appsv3 3386dc6 06linuxcnc 04. branch deleted * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=3386dc6
[15:29:05] <andypugh> I have been thinking about the toolchanger that the guy hoped to have working in 2 weeks. (He has gone quiet, so I suspect that he has been persuaded it is a 2-day job in Mach and is now ripping everything out to fit a Smoothstepper and Geckos)
[15:29:45] <andypugh> Anyway, be that as it may, it seems to me that there is no obvious way for a HAL component to move a G-code controlled axis.
[15:30:37] <cradek> I've done that with a component that's just like limit3 but I don't recall the details
[15:30:59] <andypugh> A component can supply an alternative source of input for the spindle PID, and can lie about the axis position to the system.
[15:31:05] <cradek> the component knew the offsets and had handshaking in and out like "go up now" and "ok I made it go ahead"
[15:31:22] <cradek> then the sequencing was trivial in CL
[15:31:56] <cradek> this was a 5 axis machine so this goop was built right in to the kins
[15:32:10] <andypugh> That would certainly be one solution.
[15:32:11] <cradek> kins is not the worst place to put it
[15:32:21] <cradek> well it worked fine
[15:32:35] <andypugh> I have been thinking for a while that a per-axis pin for injecting external offsets wouldn't be a terrible idea.
[15:32:53] <cradek> using M6 remap and writing all the sequencing in gcode instead would be tempting today
[15:33:23] <andypugh> It's not too hard to insert an offset using (for example) the "offset" comp.
[15:33:42] <andypugh> But I was thinking in terms of moving to absolute machine positions, and that seems harder.
[15:34:04] <cradek> yeah that's not really an offset then
[15:34:23] <cradek> at the kins level you have cleaner numbers
[15:34:38] <andypugh> (At the very least the component needs to be able to read in the "axis.N.motor-offset" pin to be able to calculate machine position from the axis.N.motor-position-cmd pin.
[15:34:55] <cradek> I drove to pensacola 1.5 yrs ago and wasn't anxious to do it again or I might've helped the guy
[15:35:34] <andypugh> For the toolchanger in question I guess a "toolchange at G33" (or whatever it really is) plus a fixed clearance offset would work.
[15:35:36] <cradek> do you think M6 remap isn't the way to go for him? I'd sure try it first.
[15:35:56] <andypugh> I think he is too dumb to figure out a remap in 2 weeks.
[15:36:12] <andypugh> Not tha most people aren't.
[15:36:19] <cradek> yes his umbrella would use tool change position, and one simple upward joint offset
[15:36:32] <cradek> oops brb
[15:37:11] <andypugh> I guess that there are people who could learn all that they need to know from scratch, and implement it, in 2 weeks. But those are much cleverer people than me.
[15:42:23] <linuxcnc-build> build #51 of 4017.deb-wheezy-armhf is complete: Failure [4failed apt-get-update shell_3] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/4017.deb-wheezy-armhf/builds/51 blamelist: Dewey Garrett <dgarrett@panix.com>
[16:13:43] <cradek> andypugh: clearly that's why he wanted to hire an expert. I don't see why people had such a negative reaction to that... it's unlikely it would work without flying someone there, but that's the only thing I thought was unrealistic about his request
[16:14:52] <andypugh> No, it seemed like the only likely way to make it work. Unfortunately I think that most of the folk who could help are more short of time than money.
[16:15:12] <archivist> or too far away
[16:15:23] <cradek> I bet you are right
[16:15:31] <andypugh> And I dobt he would want to fly Visteurs in from Latvia.
[16:15:37] <JT-Shop> I've got one guy shipping a machine to me to convert
[16:19:10] <cradek> I have done remote retrofit help work like that, but with a pretty-smart-about-stuff-that-isn't-mathy-or-programming guy on site
[16:20:06] <cradek> if I was short on money or long on time and energy, I may have tried this job
[16:20:20] <cradek> but, eh
[16:20:35] <andypugh> I am probably going to write a toolchanger comp for him. Hust because it looks easy. :-)
[16:21:08] <cradek> I would sure use ladder
[16:21:11] <andypugh> (Or, rather, it looks like something easy for me and hard for him)
[16:21:23] <cradek> I would be really reluctant to write sequencing in comp, because I've seen the results many times
[16:21:33] <andypugh> He probably should too. But I have never learned Ladder, and now is not the time.
[16:21:37] <cradek> writing state machines sucks
[16:21:40] <cradek> heh
[16:22:01] <andypugh> I quite like writing state machines. I don't know why.
[16:22:09] <cradek> the hard part is figuring out the sensors on the thing and tracking its position
[16:22:31] <andypugh> And catching the potential error states without horrible things happening.
[16:22:38] <cradek> ideally it would have a prox for "pocket facing the spindle is empty" to avoid crashyness (I had my guy add one)
[16:22:52] <cradek> yep exactly
[16:24:11] <cradek> it should be able to home and tell whether the special pocket is empty
[16:25:36] <JT-Shop> andypugh, have you seen my classicladder tool changer sim?
[16:26:06] <andypugh> I think his arm is pneumatic, so in this case it is probably not so bad if the in-spindle tool alnd selected tool touch. They will just prevent the "arm etended" prox from triggering.
[16:26:45] <cradek> ah maybe so
[16:27:51] <cradek> so it's not destructive, but it's harder to recover from than erroring and never starting
[16:29:57] <andypugh> I had plans for a toolchanger, but I really have no need for one.
[16:30:33] <cradek> hint for you: if it's not homed yet, assume the pocket facing the spindle is where the loaded tool came from, put the tool away, THEN home
[16:32:34] <cradek> you don't need any knowledge about the past other than that
[19:07:17] <memleak> andypugh, you said you compile native on ARM right?
[19:08:06] <andypugh> I have been doing so, yes.
[19:08:19] <memleak> any troubles so far minus speed / performance?
[19:09:08] <memleak> i've heard rumors about compiling native on ARM is bad practice but i have yet to actually see such a case
[19:09:10] <andypugh> The compile goes as well as can be expected. The Udoo is quad-core, so it's no slouch.
[19:09:34] <memleak> buggy code output / toolchain problems
[19:10:24] <andypugh> No, no issues there.
[19:10:38] <memleak> thanks!
[20:25:45] -kornbluth.freenode.net:#linuxcnc-devel- [freenode-info] why register and identify? your IRC nick is how people know you. http://freenode.net/faq.shtml#nicksetup
[21:22:04] <skunkworks> how come there isn't circular blend rc3 debs?
[21:22:04] <skunkworks> http://buildbot.linuxcnc.org/dists/lucid/scratch-rt/binary-i386/linuxcnc_2.6.0~pre~circular.blend.arc.rc3~12a6c8b_i386.deb
[21:51:55] <KGB-linuxcnc> 03Dewey Garrett 05dgarr/demo_appsv4 68046f8 06linuxcnc 10scripts/linuxcnc_info 10scripts/linuxcnc_var.in linuxcnc_var: support option for all * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=68046f8
[22:22:55] * seb_kuzminsky hides from skunkworks
[22:23:47] <seb_kuzminsky> err
[22:23:56] <seb_kuzminsky> scratch debs older than a week are removed
[22:24:23] <seb_kuzminsky> memleak: i'm building natively on arm too now
[22:24:36] <seb_kuzminsky> debian wheezy on odroid u3
[22:24:38] <seb_kuzminsky> works fine
[22:25:37] <skunkworks> seb_kuzminsky: what does that mean? will they build again?
[22:31:51] <seb_kuzminsky> if someone asks it to
[22:32:28] <seb_kuzminsky> skunkworks: say linuxcnc-build: force build --branch=circular-blenc-arc-rc3 0000.checkin
[22:32:51] <seb_kuzminsky> err, but spell blend right
[22:33:01] <skunkworks> linuxcnc-build: force build --branch=circular-blend-arc-rc3 0000.checkin
[22:33:14] <skunkworks> linuxcnc-build: force build --branch=circular-blend-arc-rc3 0000.checkin
[22:33:20] <linuxcnc-build> The build has been queued, I'll give a shout when it starts
[22:33:24] <skunkworks> yay
[22:33:32] <seb_kuzminsky> you have to wait for dgarr's push to finish building
[22:33:38] <skunkworks> thanks seb_kuzminsky - do I have to do that every week? :)
[22:34:09] <seb_kuzminsky> heh yeah
[22:34:14] * seb_kuzminsky sobs quietly
[22:34:32] <seb_kuzminsky> i suppose we could bump the timout up to a month or 6 or something
[22:37:25] <seb_kuzminsky> i never anticipated the scratch stuff to be used, really, but maybe it should be more usable than it is
[22:38:00] <skunkworks> it's just space :)
[22:40:13] <seb_kuzminsky> http://dilbert.com/strips/comic/1995-06-24/
[22:40:24] <seb_kuzminsky> you all knew what cartoon that was before you clicked on it
[22:40:54] <seb_kuzminsky> i
[22:41:18] <seb_kuzminsky> i've been looking at the docs build system, trying to root out the parallel make failures in there
[22:42:16] <KGB-linuxcnc> 03Robert W. Ellenberg 05circular-blend-arc-rc3 a5662c4 06linuxcnc Merge branch 'feature/velocity-limits' into circular-blend-arc-experimental3 * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=a5662c4
[22:42:17] <KGB-linuxcnc> 03Robert W. Ellenberg 05circular-blend-arc-rc3 e74559f 06linuxcnc 10src/emc/tp/tp.c Removed old comments * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=e74559f
[22:42:17] <KGB-linuxcnc> 03Robert W. Ellenberg 05circular-blend-arc-rc3 e777904 06linuxcnc 10src/emc/tp/tp.c TP: removed extraneous rtapi_print * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=e777904
[22:42:18] <KGB-linuxcnc> 03Robert W. Ellenberg 05circular-blend-arc-rc3 d8ae75c 06linuxcnc 10src/emc/tp/blendmath.c 10src/emc/tp/blendmath.h 10src/emc/tp/tc.c 10src/emc/tp/tp.c TP: Possible fix for circular segment slowdown w/ parabolic blends * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=d8ae75c
[22:42:23] <KGB-linuxcnc> 03Robert W. Ellenberg 05circular-blend-arc-rc3 c548c44 06linuxcnc 10src/emc/task/emccanon.cc canon: Slight refactor of ARC_FEED to isolate velocity / accel calculation * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=c548c44
[22:42:27] <KGB-linuxcnc> 03Robert W. Ellenberg 05circular-blend-arc-rc3 51ca8be 06linuxcnc 10src/emc/task/emccanon.cc Revert "Fix arc acceleration constraint" * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=51ca8be
[22:42:31] <KGB-linuxcnc> 03Robert W. Ellenberg 05circular-blend-arc-rc3 63e7135 06linuxcnc 10src/emc/task/emccanon.cc canon: comments and whitespace fix * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=63e7135
[22:42:35] <KGB-linuxcnc> 03Robert W. Ellenberg 05circular-blend-arc-rc3 7470721 06linuxcnc 10src/emc/tp/tc.c Actually store the new maxvel... * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=7470721
[22:42:39] <KGB-linuxcnc> 03Robert W. Ellenberg 05circular-blend-arc-rc3 ef43567 06linuxcnc 03tests/trajectory-planner/circular-arcs/nc_files/quick-tests/arc-slowdown.ngc 03tests/trajectory-planner/circular-arcs/nc_files/quick-tests/cradek_accel_issue.ngc Added missing test files * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=ef43567
[22:42:44] <KGB-linuxcnc> 03Robert W. Ellenberg 05circular-blend-arc-rc3 676b844 06linuxcnc 10(5 files in 2 dirs) Added symlinks back and misc octave scripts * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=676b844
[22:42:48] <KGB-linuxcnc> 03Robert W. Ellenberg 05circular-blend-arc-rc3 2940838 06linuxcnc 10(15 files in 2 dirs) Additional test files for arc cases * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=2940838
[22:52:23] <linuxcnc-build> build #52 of 4017.deb-wheezy-armhf is complete: Failure [4failed apt-get-update shell_3] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/4017.deb-wheezy-armhf/builds/52 blamelist: Dewey Garrett <dgarrett@panix.com>
[22:56:57] <linuxcnc-build> build forced [ETA 1h19m31s]
[22:56:57] <linuxcnc-build> I'll give a shout when the build finishes