#emc-devel | Logs for 2009-06-06

[04:02:00] <mozmck> any idea why I'm getting unknown symbols in rtapi.ko?
[04:02:53] <mozmck> I'm running on a custom kernel with rtai 3.7 and a new compile of emc trunk
[04:30:57] <mozmck> logger_dev
[04:35:19] <mozmck> logger_dev: bookmark
[04:35:19] <mozmck> Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emcdevel/2009-06-06.txt
[07:13:24] <CIA-48> EMC: 03micges 07TRUNK * 10emc2/src/emc/task/ (emctask.cc emctaskmain.cc): improve/cleanup debugging interpreter in task
[07:13:35] <CIA-48> EMC: 03micges 07TRUNK * 10emc2/src/emc/nml_intf/interpl.cc: improve/cleanup debugging interpreter in task
[07:30:45] <micges> alex_joni: around ?
[10:21:25] <alex_joni> micges: now
[10:25:34] <micges> are going to fix bug in task we found ?
[10:26:06] <micges> (SS_ENABLED in wrong case in pre or post condition)
[10:29:00] <alex_joni> we probably should, but atm I don't really have time to look at it :/
[10:30:03] <micges> ok
[10:54:58] <alex_joni> micges: if you feel confident you can fix it right, I will look at it later ;)
[10:57:13] <micges> I think not (yet) ;)
[10:59:08] <micges> sorry for disturbing
[11:08:52] <alex_joni> micges: no problem at all
[12:16:09] <micges> on dapper after sudo apt-get build-dep emc2, when I wan't to --rip --enable-simulator it show error that libpth-dev is not installed
[12:16:28] <micges> (I don't know if it ^^ is important or not)
[12:19:06] <BigJohnT_> BigJohnT_ is now known as BigJohnT
[12:55:11] <mozmck> I have rtai running on a patched vanilla linux with ubuntu 9.04...
[12:55:20] <mozmck> but emc won't start
[12:59:10] <mozmck> I can run the rtai tests fine, but if I try to run the emc latency-test I get a "-1 Unknown symbol in module" for rtapi.ko
[12:59:24] <SWPadnos> check dmesg to see which symbol it is
[12:59:42] <mozmck> symbols...
[13:01:15] <mozmck> nano2count, rt_get_time, rt_shutdown_irq, rt_task_make_periodic, rtf_get, start_rt_timer
[13:01:29] <mozmck> and a lot more...
[13:02:06] <SWPadnos> hmmm
[13:02:08] <mozmck> for each one it first says there is no symbol version
[13:02:30] <SWPadnos> oh, did you enable symbol versioning in your kernel?
[13:03:09] <mozmck> I thought I did. Let me check again...
[13:09:56] <mozmck> module versioning is enabled.
[13:10:20] <mozmck> there is an option for "Forced module loading" that is not enabled
[13:10:39] <mozmck> it says it allows loading of modules without version information
[13:10:44] <mozmck> maybe I need that?
[13:13:58] <mozmck> here the dmesg output: http://www.pastebin.ca/1449698
[13:21:24] <bjt-plasma> if I route axis.2.motor-pos-cmd through my thc and add to or subtract from the position command then send it to my hm2.stepgen.02.position-cmd
[13:21:45] <bjt-plasma> I get a joint following error when I reach the ferror level
[13:21:48] <bjt-plasma> brb
[13:27:48] <mozmck> I notice the documents say to turn off high memory support in the kernel config
[13:28:30] <mozmck> does this mean that you won't have use of more than 1 gig of ram?
[13:32:56] <SWPadnos> mozmck, those instructions might be old, if you're using a recent version of RTAI
[13:33:19] <SWPadnos> there was a bug, which is supposedly fixed, which prevented RTAI-enabled kernels from using more than 1G
[13:37:23] <bjt-plasma> nm, I know what I missed
[13:37:30] <skunkworks> It think it was fixed - I tested it may of last year.. http://www.electronicsam.com/images/KandT/testing/pent4-26ghz.png
[13:38:03] <skunkworks> something to do when you pushed the machine into swap - it caused issues.'
[13:39:15] <SWPadnos> strangely, I haev 2.6.28+RTAI working on 9.04 right now
[13:39:17] <SWPadnos> have
[13:39:37] <SWPadnos> what's strange is that I didn't recompile it, and it's been up for several days now
[13:39:44] <SWPadnos> but it used to crash
[13:40:24] <SWPadnos> I have run the latency test, and have done a little bit of 3D stuff (flipping the desktop around with "Desktop Cube")
[13:41:39] <pcw> Any one know Sebastians Email? Tried to send him a set of updated bitfiles
[13:41:41] <pcw> with the HM2 stepgen bug fixed, but its stuck (retrying on name server lookup failure)
[13:41:57] <SWPadnos> seb at highlab dot something
[13:42:12] <pcw> ok that what I have
[13:42:19] <pcw> (thats)
[13:42:46] <skunkworks> how big are the files? maybe too big for his service?
[13:43:04] <pcw> 4M total
[13:43:12] <mozmck> hmmm. does the dmesg output I posted give you any clues?
[13:43:28] <SWPadnos> mozmck, it might if I had looked at it :)
[13:43:36] <mozmck> what version of rtai are you running?
[13:43:40] <mozmck> :)
[13:43:47] <SWPadnos> err - lemme check
[13:44:11] <SWPadnos> looks like 3.7test2
[13:44:39] <SWPadnos> just to be sure, have you done a make clean / configure / make cycle?
[13:45:09] <SWPadnos> and if you're using run-in-place, did you source emc-environment?
[13:47:00] <mozmck> I'm not doing run-in-place, i did a make install
[13:47:19] <SWPadnos> oh, ok
[13:47:20] <mozmck> I did a fresh compile of everything
[13:47:55] <mozmck> I'm running rtai 3.7 from cvs, updated a day or two ago...
[13:50:14] <SWPadnos> I'd have to think a lot to help, so I think I'll get breakfast instead ;)
[13:50:30] <SWPadnos> which might help in the long run :)
[13:50:33] <SWPadnos> bbl
[13:50:34] <mozmck> ok, thanks
[13:52:19] <BigJohnT> * BigJohnT wanders out to mow the grass while it is cool outside
[14:17:50] <skunkworks> mmmmm bacon and eggs.
[14:17:55] <alex_joni> mozmck: what rtai sched are you using?
[14:27:34] <alex_joni> mozmck: my guess would be that you have versioning enabled for the kernel build, but not for the rtai build
[14:35:50] <mozmck> let me check
[14:38:42] <mozmck> I don't see any option for versioning in rtai
[14:42:22] <mozmck> under scheduling options I have "Display RTAI task execution time", "Allow round-robin scheduling", and "Use Linux syscall mechanism for RTAI calls from user space"
[14:43:12] <alex_joni> try modinfo on the built module
[14:47:09] <mozmck> filename: rtai_ksched.ko
[14:47:09] <mozmck> srcversion: 4F3928353F3273B929E7E8C
[14:47:09] <mozmck> depends: rtai_hal
[14:47:09] <mozmck> vermagic: SMP preempt mod_unload modversions K8
[14:47:09] <mozmck> parm: rtai_global_heap_size:int
[14:47:09] <mozmck> parm: rtai_kstack_heap_size:int
[14:47:11] <mozmck> parm: OneShot:int
[14:47:13] <mozmck> parm: Latency:int
[14:47:15] <mozmck> parm: SetupTimeTIMER:int
[14:47:17] <mozmck> parm: Reservoir:int
[14:47:19] <mozmck> parm: SpareKthreads:int
[14:49:16] <mozmck> looks like vermagic is the information on the kernel, not this module
[14:50:24] <mozmck> alex_joni: did you see my dmesg output?
[14:51:25] <alex_joni> yes, saw it
[14:51:51] <alex_joni> would be interesting to see one from running the kern latency test
[14:52:44] <mozmck> just a sec.
[14:54:21] <mozmck> http://www.pastebin.ca/1449757
[14:56:51] <mozmck> I have to run it through sudo. I guess this is normal?
[15:01:21] <alex_joni> right, it is
[15:01:53] <alex_joni> mozmck: the latency test looks ok, but I did mean the dmesg while you do a latency test
[15:02:21] <mozmck> oh, let me look
[15:04:08] <mozmck> http://www.pastebin.ca/1449764
[15:12:00] <lerman_> Has anyone looked at porting EMC to Solaris?
[15:12:17] <lerman_> (Solaris supports hard real time.)
[15:14:12] <mozmck> bbl
[15:30:35] <bjt-plasma> does stepgen.position-fb ever return something other than position-cmd?
[15:32:05] <alex_joni> it always does
[15:32:08] <alex_joni> return something else
[15:32:15] <alex_joni> it returns how far it has stepped
[15:32:25] <alex_joni> and if the stepgen enable is low for instance, then fb will never change
[15:32:36] <alex_joni> (of course you'll get a ferror from emc2)
[15:32:57] <bjt-plasma> like a step or two behind?
[15:33:40] <bjt-plasma> is there any reason not to just connect motor-pos-cmd to motor-pos-fb?
[15:33:48] <bjt-plasma> for a stepper system
[15:38:36] <alex_joni> yes there is
[15:38:46] <alex_joni> the stepgen has some accel, vel limits
[15:39:11] <alex_joni> you can set those tight for testing if the emc2 motion controller obeys it's own limits
[15:39:29] <alex_joni> usually they are not set though, and -fb is just a safety thing
[15:43:16] <bjt-plasma> ok, thanks
[15:47:06] <bjt-plasma> I'm working on my THC and have a dial to simulate the correction and it is working :)
[15:47:20] <mozmck> alex_joni: did that paste tell anything helpful?
[15:48:12] <mozmck> bjt-plasma: what kind of THC are you using?
[15:48:26] <bjt-plasma> Mesa
[15:48:45] <mozmck> they have a torch height controller?
[15:49:01] <bjt-plasma> it's an A-D THC card that I'm testing and working on a comp for EMC for it
[15:49:35] <bjt-plasma> I don't think they are offered yet
[15:49:40] <mozmck> and the card is made by Mesa
[15:49:43] <mozmck> I see.
[15:49:55] <bjt-plasma> it is really slick
[15:50:01] <bjt-plasma> yes it is a prototype
[15:51:14] <mozmck> I work for Tom at CandCNC, and we make a THC.
[15:51:30] <bjt-plasma> I see
[15:52:06] <mozmck> I finally got Tom to allow me to release stuff for EMC for our products.
[15:52:27] <bjt-plasma> that's good
[15:52:33] <pcw> BJT: The hardware stepgen its really a kind of velocity servo
[15:52:35] <pcw> system, motor-pos-fb tells EMC where the step motor is
[15:52:36] <pcw> (how many pulses the step generator hardware has made)
[15:52:38] <pcw> this will vary from the desired position because of differences in
[15:52:39] <pcw> clock speed between EMC and the hardware, and during direction reversals
[15:53:05] <bjt-plasma> thanks pcw
[15:53:42] <bjt-plasma> I've got 8 out of 10 things worked out :)
[15:54:27] <bjt-plasma> it should not vary more than ferror if it is set correctly right?
[15:56:08] <bjt-plasma> to keep the motion controller from thinking I had a following error I patched the pos-cmd back to the pos-fb in axis
[15:56:22] <pcw> No, it shouldn't , the Stepgen drivers job is to keep adjusting the velocity
[15:56:23] <pcw> to keep pos-cmd and pos-fb as close as possible
[15:56:32] <alex_joni> mozmck: not really :/
[15:56:41] <alex_joni> either rtai or emc2 has some compile issues
[15:56:54] <alex_joni> any reason against using 8.04 packages?
[15:57:09] <alex_joni> and 8.04 :)
[15:57:32] <mozmck> I'm running 9.04 on my main machine here and I don't want to/can't downgrade.
[15:57:52] <mozmck> I don't want to have to set up another computer to work on EMC
[15:58:33] <mozmck> My router in the shop is running 8.04, but I can't really develop on that.
[15:59:34] <bjt-plasma> the offset from the THC does not show up on the DRO but that doesn't seem to matter
[15:59:43] <bjt-plasma> to me anyway
[16:00:55] <bjt-plasma> the only part left before actual testing is removing the offset as the Z raises up, after this mornings testing I don't see any problem with that.
[16:03:54] <alex_joni> bjt-plasma: you can maybe hook the offset to a pyvcp display
[16:04:00] <bjt-plasma> thanks pcw
[16:04:22] <bjt-plasma> alex_joni: I did that :)
[16:06:43] <bjt-plasma> * bjt-plasma heads back out to the garden before it gets hot out
[16:07:24] <alex_joni> mozmck: doing RT stuff on your devel box?
[16:07:34] <alex_joni> because you can do a sim only install, and that should be fine
[16:08:11] <micges> alex_joni: how can see float value from rtai module? I saw that floats are unsuported
[16:08:22] <mozmck> hmmm, I may do that. I was hoping to hook up some simple hardware here for testing too.
[16:08:46] <alex_joni> mozmck: it's surely fixable, but one would need a 9.04 machine for that ..
[16:08:58] <alex_joni> maybe SWPadnos will try to figure out running emc2 on his 9.04
[16:09:05] <mozmck> I also wanted to see how the rt would perform on this dual core processor.
[16:09:38] <mozmck> I'll be working on it too. I may try and disable module versioning in the kernel...
[16:11:08] <mozmck> one thing that was not enabled by default in the rtai config is "Real-Time Driver Model over RTAI"
[16:11:14] <mozmck> is that something I need?
[16:19:30] <alex_joni> nope
[17:06:21] <micges> alex_joni: about POSITION_FILE error checking: is it better to emc not save NAN values to this file or not load NAN values from this file ?
[17:08:11] <alex_joni> I'd go with both
[17:12:09] <jepler> and fix the bug that gives a nan in the first place
[17:17:00] <micges> jepler: right
[17:19:19] <jepler> (I still think the other's unneeded because it only hides a bug)
[17:36:47] <micges> ok thats weird: I have 4 rip: trunk, modified trunk x 2, and 2.3 branch, and all of them shows nan in xyz position
[17:37:38] <micges> even if POSITION_FILE is disabled
[17:38:35] <micges> (in sim_mm config)
[17:44:22] <micges> last thing before: I've added local int variable to int pmCircleInit() in _posemath.c
[17:46:19] <cradek> I had no trouble with sim on 9.04
[17:53:20] <micges> cradek: can you check how manu cpu are milltask consuming ?
[17:55:21] <micges> bbl
[18:28:56] <lerman_> Have any of you tried approximating an involute of a circle by a series of arcs?
[18:37:43] <mozmck> Ok, I've gotten farther, but now I get this message:
[18:37:44] <mozmck> Error constructing axisoptions({}):
[18:37:44] <mozmck> couldn't load file "/usr/local/share/emc/tcl/emc.so": /usr/local/share/emc/tcl/emc.so: cannot open shared object file: No such file or directory
[18:37:55] <mozmck> when trying to start latency-test
[18:41:13] <mozmck> I disabled module versioning in the kernel and no longer get all the symbol errors...
[18:49:01] <alex_joni> mozmck: I'd suggest using run-in-place
[18:49:40] <mozmck> ok, I'll try that.
[18:49:54] <alex_joni> at least until you figure it out
[18:52:29] <alex_joni> and afterwards it pays to build a deb, and install that
[18:53:07] <alex_joni> check if emc.so is in /usr/local/share/emc/tcl, it may be a packaging bug
[18:53:19] <alex_joni> we usually use/test deb's and those use /usr as a prefix
[18:53:29] <alex_joni> so emc.so usually ends up in /usr/share/emc/tcl/
[18:54:15] <mozmck> I just make install, and I checked and emc.so is not in /usr/local/share/emc/tcl
[18:55:51] <alex_joni> try without the /local/ and if it's there, just symlink it
[18:56:01] <alex_joni> and file a bugreport at sourceforge ;)
[18:56:02] <mozmck> looks like it just didn't get copied.
[18:56:14] <alex_joni> hmm.. odd
[18:56:16] <mozmck> there is no directory /usr/share/emc
[19:00:36] <mozmck> aha! I just copied emc.so and hal.so from my emc2/tcl directory to /usr/local/share/emc/tcl and now latency test is running
[19:01:14] <mozmck> Max jitter so far on the base thread is 5016...
[19:07:16] <alex_joni> sounds good :D
[19:09:08] <mozmck> bleh, now glxgears and emc are telling me glx is not loaded. But the Xorg log file says it is...
[19:09:42] <mozmck> So I should file a bug report that some files are not getting copied on make install?
[19:11:00] <mozmck> bbl
[19:22:22] <cradek> mozmck: yes, or if you can figure it out, send a patch that fixes it
[20:00:09] <micges> cradek: can you look at this: http://www.pastebin.ca/1450025
[20:00:27] <micges> (resetting dtg and current_vel after aborting program)
[20:04:32] <cradek> looks good to me
[20:04:38] <cradek> also good catch in tpInit
[20:08:03] <cradek> bbl
[20:11:41] <CIA-48> EMC: 03micges 07TRUNK * 10emc2/src/emc/kinematics/tp.c: clearing DTG and current velocity after aborting program
[20:36:49] <BigJohnT_> BigJohnT_ is now known as BigJohnT
[20:54:30] <cradek> micges: in fact, it couldn't hurt to put the tpInit fix in 2.3 branch
[20:55:04] <cradek> or both...?
[20:59:51] <micges> both what?
[21:02:28] <micges> branches or fixes?
[21:04:13] <cradek> both fixes
[21:05:49] <micges> ah ok
[21:08:09] <micges> cradek: how I must point in message log that tihs is backport?
[21:08:46] <micges> this*
[21:15:49] <micges> nm I found it
[21:17:02] <mozmck> cradek: I just noticed that the install put hal.so in /usr/local/lib/emc/tcl
[21:17:08] <mozmck> is that the correct place?
[21:28:21] <CIA-48> EMC: 03micges 07v2_3_branch * 10emc2/src/emc/kinematics/tp.c:
[21:28:21] <CIA-48> EMC: from TRUNK:
[21:28:21] <CIA-48> EMC: clearing DTG and current velocity after aborting program
[21:28:21] <CIA-48> EMC: fix initialisation in tpInit()
[21:39:21] <micges> good night all