#linuxcnc-devel | Logs for 2013-10-31

[10:39:27] <seb_kuzminsky> i want to try to use the sourceforge bug tracker a little more to track 2.6 issues, can someone add me as an admin on the sf emc project?
[10:39:39] <cradek> I think I can do that
[10:45:56] <cradek> somehow
[10:48:13] <cradek> I can't figure out how
[10:48:26] <jepler> I'll look
[10:52:50] <seb_kuzminsky> holy wow, sf's documentation is not very helpful here
[10:53:18] <cradek> they made it all pretty, and now I can't figure out how to do anything
[10:53:28] <cradek> but all the corners are rounded
[10:53:47] <seb_kuzminsky> there's this: https://sourceforge.net/apps/trac/sourceforge/wiki/Add%20a%20project%20developer
[10:54:01] <cradek> you're already a project developer
[10:54:27] <seb_kuzminsky> i am, but i think i need to be a project admin in order to do things like create milestones for our tickets
[10:54:42] <cradek> I think that was not previously the case
[10:54:46] <jepler> having admin permission on the ticket system isn't enough?
[10:54:55] <cradek> it was formerly
[10:55:02] <cradek> but I couldn't figure out how to do that
[10:55:07] <jepler> (which you didn't have before but should now)
[10:55:13] <cradek> yay
[10:55:32] <jepler> if not I'll go ahead and make you a project admin as well
[10:56:32] <jepler> also added "configure" permission, which is described as "edit milestones"
[10:59:58] <jepler> seb_kuzminsky: any luck?
[11:09:27] <seb_kuzminsky> yes, i see an 'edit milestones' button now, thanks
[11:13:39] <seb_kuzminsky> yep, it's working like i want now, thanks!
[11:14:15] <seb_kuzminsky> i made a '2.6' milestone, for tickets i want to close before 2.6, and a '2.next' for stuff that can wait until later
[14:36:34] <skunkworks> seb_kuzminsky, ! Sqashing my dreams!
[14:36:43] <skunkworks> ;)
[14:37:50] <skunkworks> (It was mainly a phishing scam to get an update on his work)
[14:44:15] <seb_kuzminsky> heh
[14:44:28] <seb_kuzminsky> don't get me wrong, i'm all for it, just not in 2.6
[14:49:58] <andypugh> I see 2.6 as more an infrastructure release.
[14:50:27] <andypugh> But then how I see things isn't all that relevant
[14:51:05] <andypugh> One question, I am converting to JA4 and I am wondering why it is net jogwheel joint.0.jog-counts ? I would expect to be jogging an axis, not a joint.
[14:51:22] <cradek> axis wheel jogging isn't done yet
[14:51:34] <cradek> there will eventually be both
[14:51:50] <andypugh> OK.
[14:51:52] <cradek> I believe micges is working on it
[14:52:05] <cradek> seb_kuzminsky: should that be on your list?
[14:52:13] <andypugh> This is a trivkins machine, so I can live with it.
[14:52:22] <andypugh> And I think I saw it on the list
[14:52:28] <seb_kuzminsky> cradek: i think it is
[14:52:39] <seb_kuzminsky> 4.4
[14:53:01] <seb_kuzminsky> and 4.6.2
[14:53:07] <seb_kuzminsky> my list is not very well organized
[14:55:26] <jepler> happily, I just now noticed that most of those "All Rights Reserved" files are #ifdef __ppc__ so we can just delete them
[14:55:29] <jepler> at least two are not, though
[15:10:44] <jepler> pretty sure I have no systems where I would test those source files
[15:15:04] <cradek> I have a ppc that still works, running dapper
[15:17:04] <jepler> I think you also need a xenomai kernel
[15:30:16] <andypugh> I think we are going to need to offer some sort of script to change configs to the new format.
[15:30:30] <andypugh> I know what I am doing and find it troublesome.
[15:30:41] <cradek> or hackery that can read the old format
[15:32:34] <andypugh> It's the HAL pin name changes.
[15:36:49] <cradek> hm yeah that's harder
[15:39:02] <jepler> There are C APIs hal_pin_alias() / hal_param_alias()
[15:39:05] <jepler> we could emply them in these cases
[15:40:17] <andypugh> It shouldn't be too hard to compile an equivalance list, and offer to re-make the files in a fresh config directory.
[15:41:27] <andypugh> It's probably easier if you know how to hack a text-processing language. I don't think it would be a fun thing to write in C :-)
[15:41:54] <cradek> we've done autoconversion of a tool table. this is harder because we don't already read/parse/write the file.
[15:42:08] <cradek> this would be a lot more guessy
[15:43:22] <andypugh> HAL pins is easy enough, and the INI sections are OK, I think. The fun part is fixing the HAL references to the INI.
[15:47:34] <jepler> beacuse we don't have a way in hal to say "try this, then that" like we have in the linuxcnc script?
[15:47:42] <jepler> sp
[15:48:54] <cradek> we could add a new halfile, to be loaded first, that sets up aliases?
[15:49:20] <cradek> that would be a new file and one predictable line added to the ini
[15:51:26] <kwallace2> In the 2.6 to-do list http://wiki.linuxcnc.org/cgi-bin/wiki.pl?Todo-2.6 I noticed item 10 which seems similar to a problem I am having with a 2.6.0~pre. based setup I'm working on.
[15:52:05] <kwallace2> Probing stops randomly, but mostly continues on a mouse move or key press.
[15:52:35] <cradek> a mouse move!?
[15:53:11] <jepler> that would be a new one on me too
[15:53:21] <kwallace2> Yes. if I keep the mouse moving, the probe routine will go without a hitch.
[15:53:28] <jepler> fwiw I think there were problems probing from mdi in 2.5 and under
[15:53:50] <jepler> in whatever timeframe I worked on and abandoned "high resolution probing"
[15:55:07] <cradek> I suspect the thing in that thread is a halui feature biting us
[15:55:27] <kwallace2> It seems from the item 10 references, that the problem went away, but there seemed to be no indication what the problem was or how it got fixed.
[15:55:51] <cradek> have you turned on task-issue debug and watched for nonsense?
[15:56:17] <kwallace2> No, I can give it a try.
[15:59:02] <kwallace2> I assume this is relevant? http://www.linuxcnc.org/docs/devel/html/config/ini_config.html#_task_section_a_id_sub_task_section_a
[16:00:52] <cradek> ?
[16:01:19] <cradek> oh are you asking how to turn on task-issue debug?
[16:01:55] <cradek> machine/set debug level, check "task issue"
[16:02:01] <cradek> then watch stdout
[16:02:02] <kwallace2> Sorta.
[16:07:55] <kwallace2> Oops, my application doesn't use AXIS.
[16:08:36] <alex_joni> kwallace2: you can change DEBUG in the ini file
[16:08:47] <alex_joni> I just make it 0xfffffff and it spits out a lot ;)
[16:08:51] <jepler> or you can run the debug-level-switcher, it's a standalone program whose name I forget
[16:09:03] <jepler> "debuglevel"
[16:09:05] <alex_joni> but there's info somewhere what bit you need for task
[16:10:09] <jepler> >>> "0x%x" % linuxcnc.DEBUG_TASK_ISSUE
[16:10:09] <jepler> '0x10'
[16:12:22] <kwallace2> "$ debuglevel" seems to work. I need to go to the shop and try it out. Thank you.
[16:44:35] <kwallace_tmill> Here is what I got, with a corner finding probe which moves from a near position, then +Y, +X until a probe trip, backs out to near position, then +X, +Y until probe trip. http://wallacecompany.com/tmp/task_debug.txt
[16:57:38] <kwallace_tmill> Is there a probe tripped event here?
[16:57:49] <kwallace_tmill> Issuing EMC_TASK_PLAN_EXECUTE -- (+520,+268, +164,G38.3\032Y3.7460,) Issuing EMC_TRAJ_CLEAR_PROBE_TRIPPED_FLAG -- (+228,+12, +0,) Issuing EMC_TRAJ_PROBE -- (+229,+116, +0,6.704403,-0.010019,-2.366242,0.000000,0.000000,0.000000,0.000000,0.000000,0.000000, +5,0.166667,2.250000,15.000000,\001,) waiting for status.probing - i = 0 waiting for probe trip - i = 0 waiting for probe trip - i = 1
[17:21:27] <kwallace_tmill> Here is another which stalls on the first touch. http://wallacecompany.com/tmp/task_debug-2.txt
[17:49:48] <andypugh> Jog-While-Paused with Touchy: http://www.youtube.com/watch?v=fSYt-cRVjzw
[17:52:27] <cradek> did you have to fix it, or did it just work?
[17:53:02] <andypugh> It just worked
[17:53:31] <andypugh> But it also behave rather strangely if you try to stop and restart the spindle...
[17:53:59] <cradek> I didn't really understand what happened
[17:54:46] <andypugh> No, neither do I.
[17:55:28] <andypugh> If you run, pause, quill up, run it's all fine.
[17:55:57] <andypugh> If you run, pause, quill up, stop the spindle, run it continues, but with a stopped spindle.
[17:56:10] <cradek> that all I expect so far
[17:57:07] <andypugh> If you run, pause, quill up, stop the spindle, start the spindle, press run, you get spurious soft-limit errors on all axes, and the program won't run. But, the machine _doesn't_ e-stop. You can tell that because the lights go off when the machine e-stops.
[17:57:35] <cradek> I don't expect an automatic estop pretty much ever
[17:57:57] <cradek> ferror and amp-fault will cause it to go to machine off state, not estop
[17:57:57] <andypugh> In what sense?
[17:58:09] <cradek> but soft limit errors don't cause any state change
[17:58:26] <andypugh> Yeah, OK. Machine-off also normally turns off the lights.
[18:00:22] <andypugh> In fact, when I pressed the machine-on button there was a long pause with both buttons showing pressed, then the off-button un-pressed itself, but there was otherwise no change in machine state. Normally I would expect the lights and VFD fan to be off with machine-off, and to both go on with machine-on.
[18:01:03] <cradek> yeah those delays are weird
[18:01:15] <cradek> touchy probably needs a massage now that modal expectations are a bit different
[18:01:40] <andypugh> For something that wasn't actually meant to work, it works quite well.
[18:02:10] <cradek> yes, purely accidental :-)
[18:02:29] <andypugh> But pressing "play" with the machine paused triggering soft-limit errors looks a lot like an out-of-bounds write or soemthing...
[18:03:04] <cradek> you should check to see what mode switch commands it's sending. probably you don't want it to send any in these situations you're testing
[18:03:40] <cradek> yeah I agree that's the weirdest thing you saw...
[18:03:56] <andypugh> In the video it was just showing soft-limit on 2. But in a prior run it was rapidly flicking continuously 012012012...
[18:04:23] <cradek> yeah that ain't right...
[18:05:14] <cradek> might be a bug in the new resume state machine (it's being confused by extraneous mode switches maybe)
[18:05:20] <cradek> wild guess
[18:05:42] <andypugh> Yeah, perhaps mah will come up with an idea.
[18:06:07] <cradek> I suggest checking the messages and seeing if touchy is doing something weird
[18:07:18] <andypugh> How do I check messages in Touchy?
[18:07:32] <andypugh> (and it isn't going to happen tonight)
[18:07:34] <cradek> set DEBUG=0x10 and run in a way that you can see stdout/stderr
[18:08:35] <cradek> you don't have continuous-jog switches do you
[18:08:49] <andypugh> No
[21:57:57] <skunkworks> http://www.electronicsam.com/images/house/stellabee3.JPG
[22:24:52] <jepler> another reason I hate github: it makes copies of every original branch in every fork, so there's a lot of unneeded branches in each github remote you add
[22:25:02] <jepler> you'll end up with a dozen different */master each pointing at a different ref
[22:49:43] <seb_kuzminsky> jepler: "committed"... i dont think that means what you think it means