emctest: Delete the scripts/rtapi.conf, do an update, run make again, and report back - Just done a small commit. just reading your emails so if the RTOS is RTAI, it comes with suitable loading scripts? Paul, no change in load failures. cat scripts/rtapi.conf then copy/paste the MODULES= line email the rtapi.conf over... MODULES='adeos rtai rt_mem_mgr rtai_sched rtai_fifos rtai_shm rtai_libm' emctest@SCS-laptop:~/emcdev/emc2$ paul_c: I would have expected your fix to work for the "start" but fail for "stop" hmmmm. rtai_up isn't there time to look at Makefile.inc steve: yes...... readlink /lib/modules/`uname -r`\rtai\rtai_ksched.o silent readlink /lib/modules/`uname -r`/rtai/rtai_ksched.o oops, my mistake hed.o and rtai_up.o on seperate lines what the heck is hed.o? is this for rtai 3.0? yes i finally got to compile emc2 on my system (the problem was a missing tcltk rpm) SuSE has tcltk split into 2 rpm-s, one normal, one for developers the dev one was missing? but now it compiles without any warnings (cvs version from about a week ago), cool Iam checking out the latest version right now. to give that one a try too * paul_c finds a typo in the makefile compile ok emctest: try a cvs up and go through the process again... i ran emc_run from scripts and it locked my machine that's not good i hope any messages before it locked up? nope just that it inserted 2 rtai modules what rtai are you using? 24.1.13 I just see : Starting emc... ... /bin/sh: line 1: RTAI_MOD: command not found Hi! Pretend this message is a nice popup graphic Using /lib/modules/2.4.21-rthal5/rtai/rtai.o and then it hangs entire machine is hung? completely I'll try a modprobe on the modules to see if there's a problem with rtai maybe how can you do that if it's hung? after reboot ;) ok same thing that msg about RTAI_MOD shouldn't be there modprobe rtai .... machine dead ;) so... relax, it's not a EMC2 problem ;) lol what gcc version did you use ? gcc -dumpversion emctest: Try cvs up again. 3.3 with rtai-24.1.13 ? yes not good? Bad move. rtai-24.1.13 was never modified to use gcc-3.x actually it's not quite a 3.3 gcc --version gcc (GCC) 3.3 20030226 (prerelease) (SuSE Linux) but it woked for me on another machine uhgnnn... pre-release versions are never good (for RT work) i know ;) Try upgrading to a recent 2.4.xx kernel and RTAI-3.x well it's a 2.4.21 I'll get rtai 3.x what minor should I choose? 3.0r4 ? or 3.1testx Paul, I need to run in a few minutes, any more requests after my compile failure. what compile failure? emctest: Can't think of anything... John.... Generate /home/emctest/emcdev/emc2/scripts/rtapi.conf Yields..... /bin/sh: line 1: RTAI_MOD: command not found you mean rtapi.conf contains that msg, or when you try to run emc you get that error? Both from the make that Paul requested after the commit he just did. found another typo... mail me (us) rtapi.conf and Makefile.inc committed a change. or not cvs up and try again.. noticed something strange about cvs cvs diff was unable to find either of your last 2 changes Still getting errors on compile steve: open src/rtapi/Makefile with your favorite editor search for "rtapi.conf" remove the leading @ from every line after "This rule actually makes rtapi.conf" try make again, capture the results, and mail alex_joni2: Use a 2.4.27 kernel and RTAI from the kilauea cvs branch at www.gna.org (dont' forget to remove rtapi.conf before doing make) paul_c: I think I see the problem mod_dir=$$(find /lib/modules/`uname -r` -name "mbuff.o" -o -name "rtai_sched.o") ; it's looking for rtai_sched to identify the modules directory sched isn't there for rtai-3 rtai-3.0r4 not good? paul_c: should I patch with Adeos? I have been deleting rtapi.conf most times, but also used "make clean, make" Guys, I'v gotta go now... we gotta be getting close possible to come back later? Send me email with status and I will try to be back this evening. I will not be logging, so assume nothing. duh - I have an RC46 HD install on one of the farm PC's I'll try it there Later... i don't get it why rtai doesn't work :( compiled with gcc-3.3 ? yes but it's the same setup I used on another machine, where it worked that one was a SBC (Pentium class, 300 MHz) this one is a AthlonXP gawd, RC46 is dog slow on a 200MHz box Both using exactly the same distros ? yes and the same kernel config ? what's bugging me is that it simply hangs no errors, no nothing anyways... I'm building a kernel now patched with adeos from rtai-3.0 which CPU did you select during kenel config ? athlon should I go more genric? Use the generic i686 586/K5/5x86/6x86/6x86MX ? 6x86 I used i686 here on Knoppix I've got a 2.2 athalon * paul_c boots another box... What are we chasing? rtai-3/emc2 module loading bug got a cvs download, in the process of compiling How may I help? the problem only happens on RC46 Got that. wierd make is failing, but no error messages are printed must be more typos in the makefile yep, better now * paul_c gives up on the makefile for the timebeing.. emc2 won't run if halmeter isn't compiled... the ini loads three halmeters - that's left over from testing can be removed comment out or remove the lines that start with USRMOD1, 2, 3 paul: am I just confused, or are there two sets of RT modules in RC46 one at /usr/realtime/modules one at /lib/modules/`uname -r`/rtai jmkasunich_4: are there any parameters that must be passed to motmod.o at insmod time ? let me check yes, the thread periods no default settings ?? nope shit! the run script reads them from the ini file not having defaults really sucks. I thought I had checking on them, but looks like I forgot it the insmod should fail unless they are provided a default that is too fast could lock the machine if the insmod doesn't fail, that's a bug, and I will fix it Oh it fails alright I meant fail cleanly segfaults and produces a kernel oops. init_module should check and bail - sorry about that at the "HAL: creating servo thread" trying to create a thread with a period of zero actually, rtapi should reject that... * jmkasunich looks A 1mSec rate is fine on just about all boxes. yeah, the servo rate is no big deal... the base rate is the one that can really load things down 50uSec is a reasonable figure for the base rate. except if you're running servos (or jon's board) you can get away with 1mS if you feel strongly about defaults, feel free to change them from 0 to whatever 50uSec for the step generator. but it still should not crash if the value is zero * paul_c goes to get a coffee. found the bug committed seems that rtai is finally running rtai 3.0 on 2.4.21 but emc doesn't run it complains about rtapi and hal_lib any ideas? yes need rtai_up and rtai_sem in rtapi.conf and maybe rtai_hal too still not it opens: fifos, shm, up, sem, hal rtai_math rtai_math is already in the list alex_joni: I'm working on the same problem here on a RC-46 box with a diversion into period error checking ;-) i see... alex_joni: try "sudo scripts/realtime -v start" well at least I'm not alone I'm root don't think sudo will help then you can skip the sudo... ;-) the -v will show insmod error messages are you sure it's -v? if I use -v it just explains the syntax I just added that option last night are you using a recent checkout? guess it's not on anonymous checkout yet guess not as far as I see it, it doesn't run through the whole list of modules or... there's a error from modprobe which I don't see I'll try loading them by hand right - each module has dependencies on prior ones, if one insmod fails, the rest will also fail normally errors are suppressed, because some of the modules on the list aren't present on all systems so some insmod failures are normal well.... adeos is not known hmmm... guess I need to specify the path for rtai 3 ? right? insmod adeos should find it but I only know what happens with BDIs, not your install nope I've seen from previous discussions that you? have some symlinks in lib/modules/ to /usr/realtime you mean my question about two sets of rtai modules? on this box, I can't find any place where one set is symlinked to the other well... for rtai 24.x it actually seems like two distinct directories there was a dir inside /lib/modules called rtai * jmkasunich_4 does a check which held the modules and modprobe or insmod found those from the standard rtai 3.0 install that one's missing (or I did something wrong :-?) here it is /lib/modules/2.4.25-adeos/rtai/ well I don't have that one wonder why .... ???? and also usr/realtime/modules/ yes should I try to ln -s? or cp? what do you think? insmod automatically looks in /lib/modules/`uname -r` and it's subdirs yes but it doesn't look into /usr/realtime/modules it would need to be explicitly told to look in /usr/realtime and because there is no /lib/modules/`uname -r`/rtai it fails locating the modules what is uname -r on your box? 2.4.21-adeos but there is not lib/modules/2.4.21-adeos at all? or no rtai subdir in it? rtai subdir make the rtai subdir a symlink to /usr/realtime/modules, perhaps? I still expect problems loading rtai_ksched, that's what I'm investigating now I can't find a module called adeos do you have such a module? did you enable adeos as a module, or built in during kernel config ? built in i remembered now.. silly me then you should be able to move to the next module in the list I can confirm that ln -s in lib/modules/kernelname/ does not really work nor does insmod find the modules nor does depmod I just copied the modules by hand, and they began to be found insmod uses a very complicated file tree walk algorithm, to avoid following a build symlink and walking the entire kernel source tree exactly when it does and does not follow symlinks is hard to understand ;) anyways... I did a insmod rtai_shm it finds the module, but I get a lot of unresolved symbols others need to be loaded first for example: rt_get_name, rt_get_adr I believe the order is: adeos (you have built in) rtai rtai_hal rtai_ksched then the remaining ones, in any order rtai_ksched is a symlink that insmod may or may not follow it didn't follow so I did insmod rtai_up my rtapi.conf had rtai_sched instead of rtai_ksched that seems to have been the problem here sched is used by rtapi 2.x, ksched is 3.x under 2.x sched is _not_ a symlink, that's why 2 works and 3 doesn't (yet) I'm working on the script that generates rtapi.conf so it gets that right I seem to be missing halmeter :-? I think it's because of GTK as far as I see from the Makefile two choices... remove or comment out the lines in emc.ini that invoke halmeter, or figure out why halmeter didn't compile when you did ./configure, it should have displayed something about the version of GTK that it found under rtai-24.1.xx, rtai_sched was a hard link. you may need a gtk-dev package do hard links show up when you do ls -l? should I did remove halmeter references from emc.ini and now TkEMC started ;) it really took me by surprise :)) I have rtai-21.1.10... either rtai_sched.o is a real file, not a link, or hard links don't show up in ls -l Be buggered if I can get emc2 to run on this Live install. emctaskmain can't read motion status. I haven't tried to run it yet, still mucking about with rtapi.conf I hacked rtapi.conf by hand I just made a hard link for testing and it doesn't show on ls -l so how is one supposed to know if a file is a real file or a hard link? ls -li shows also the inodes the idea for hard links is that both files (or more) point to the same inode so you cannot say that the new file is a copy and the old file is the real file they are both likewise real ok, got it now this install of RC46 has adeos.o in yet another place... /lib/modules/2.4.25-adeos/kernel/adeos/adeos.o shouldn't matter if depmod finds it... should be ok check /lib/modules/2.4.25-adeos/modules.map it doesn't matter when I run insmod adeos manually but when I run the realtime script, I get: insmod: adeos: no module by that name found ah-ha what? the script is setting MODPATH to point to /usr/realtime/modules but adeos isn't there, and setting MODPATH over-rides the normal search path I see messing with modpath is not the right way to do this you could append to MODPATH ? normally MODPATH is undefined, and insmod uses a built in default /lib/modules/`uname -r` is the default yes but... what's in /etc/modules.conf should overwrite MODPATH committed another fix to rtapi.conf This one is working on my Live install. /etc/modules.conf is symlinked to /MorphixCD/etc/modules.conf what is MorphixCD? hmmm Live CD? trying to cat /etc/modules.conf says no such file it was the Live CD base mod this is a hard disk install, the CD isn't mounted modules.conf is depreciated with the current modtools they now use /etc/modules ok, so I should ignore that red herring define current modtools modprobe/insmod/lsmod I meant version ok, this much I know... the makefile (thru rtapi.conf) is setting MODPATH to /usr/realtime/modules, which prevents insmod from seeing adeos bc. I have 2.4.22, and it still uses /etc/modules.conf (on SuSE 8.2) jmk: right so back to studying makefile where can I find a nc-file? the 3d-chips emc/programs they haven't been copied over to the emc2 tree yet do you have the emc tree? I found it on another machine strange.... when shutting down emc it doesn't unload rtai_hal that's another manifestation of the symlink problem... you have to install it using rtai_up, but remove it using rtai_ksched rmmod rtai_hal works without problems since ksched is still there, it is using symbols exported from hal I think that's what happened the current version of rtapi.conf is working fine for me. load rtai_up, unload rtapi_up lsmod shows just rtai_hal no rtai_up, or rtai_ksched hmm... steve reported otherwise last night, but who knows my head is starting to spin i did insmod rtai_up and rmmod rtai_ksched rtai_ksched notloaded :D ok, I must be mis-remembering paul: I replicated the result I remembered from last night rtapi.conf has rtai_up and not rtai_ksched everything loads properly, but lsmod shows rtai_ksched, not rtai_up and rtai_ksched is not removed (which prevents rtai_hal and adeos from being removed) The long term solution is to use the RTAI module load/unload script. where is that script? /usr/realtime/bin rtai-load Gotta go - Food. strange things happen when running scripts/emc.run it fails any msgs? however rtai_hal and rtai_math don't get unloaded and if I run it the second time (without doing anything) it starts maybe wrong order in rtapi.conf? could be you have the latest (I did a commit about 10-15 mins ago) stand by, I want to make another change I can't see differences right now, beeing on anonymous cvs checkout it works now my order is: hal, up, fifos, shm, math, sem the last four should work in any order does everything get removed when you shut down emc? yes lsmod shows no more rtai modules after shutting down if you run scripts/realtime start, then lsmod, does it show rtai_up, or rtai_ksched? rtai_up but!!!! I don't have a symlink rtai_ksched I replaced it with a copy of rtai_up sorry... ok, I think that explains it * jmkasunich_4 tries a test in /usr/realtime/modules it is as a symlink in /lib/modules/kernel/rtai it was a copy but I changed it back to initial ok, figured it out... as it is now: the modules get loaded from /usr/realtime/ but lsmod sees it from /lib/modules/ insmod is looking for rtai_up, but if it finds it thru the rtai_ksched symlink, it loads it with the name rtai_ksched that is really wierd and why should it find first through the symlink? and not as it is? that sounds not very sane :D if I have a module foo, and a symlink to it called bar and insmod foo it loads bar? insmod foo will load foo, but call it bar now that is weird because it loads by the symlink name, but searches by the resolved name it's worse insmod bar won't load anything! and the search is alphabetical if the module is bar, and the symlink to it is foo, the results are different insmod bar will load bar, and call it bar insmod foo will fail there is even a possible security hole suppose the OS loads a module called "important.o" at startup what modutils do you have?? 2.4.26 2.4.22 here attacker has no way to overwrite important.o with an evil copy but he can make a symlink !? but he can put a symlink dummy.o in the modules path, that points to /home/attacker/important.o and it shows up as important it will be listed by lsmod as dummy i did an insmod rtai_ksched no module by that name found right, cause it is a symlink to up - it resolves the symlink before testing to see if its the right module but when insmod rtai_up lsmod shows rtai_up hmmm only thing will be ... that .. there are more ksched types do insmod -v rtai_up redirect to a file (makes thousands of lines) then grep for rtai_ Using /lib/modules/2.4.21-adeos/rtai/rtai_up.o Symbol version prefix '' that's it all there is in the file wow, I get about 150K of searching AFAIK there are 7 types of sched in rtai right so we need to make it work with symlink it should be the proper way to do it... right? rtai_up is too specific the latest rtapi Makefile builds rtapi.conf by following the symlink at make time, determing the proper sched, and adding it to the rtapi.conf nice alex_joni, And after that? :) jmk: unless it's a user that swaps his sched from time to time then he needs to make again right... but that's too specific now jmk: does /usr/realtime/bin/rtai-load work for you? didn't try it... looked at it only looks complicated it doesn't start here complains about .runinfo missing I'm not even gonna try I was wondering how it behaves with the insmod there is code in there that prepares a list of modules... it probably figures out which one is the proper scheduler wow... nine commits to src/rtapi/Makefile in the last 15 hours I think we were guessing a little bit ;-) ;) well.. my emc2 is running 3D_chips nicely cool I don't really get to adjust the backplot running in backplot? or running motors? backplot no motors at home :( I have motors, but not connected to a machine yet how do you change the setup for backplot in order to get something usefull? what's that parameter size? that's the one make it 2, I think, for chips 2? yep what does that mean? I haven't the slighest idea ;-) i see ;) I tried to make it bigger, smaller but it just behaved strangely but Ray said 2 makes chips fit on the screen, so 2 is what I did ;) I think it only scales new points, doesn't scale what's already been drawn oh... ok i usually do a reset after adjusting the value well.. guess I'll try it on my motors too still just motors here, no machine :( you have steppers? DC with geckos nice I just fitted my geckos to some CPU copper coolers my test setup is old weak NEMA 23 steppers with a Xylotex drive what is there left to do on emc2? on emc2 itself, _lots_ of testing and cleanup and documentation on hal, lots of drivers - STG, jon elson's boards, and anything else I can get my hands on I have one of these machines (without the CNC stuff) http://www.shoptask.com/images/f6.jpg interested in the DRO driver? that is what my conversion project is... NEMA34 motors and geckos DRO driver? yes there was a DRO board ISA for encoder feedback the kaluga (spelling?) board yes i think ;) yep, should make a HAL driver for that but would need one to test because I made a similar board to that one so writing a driver would be more or less the same does it use a 7266 or something? 7166 pretty much the same but it's PC104 format good for SBC'S yep here's a thought.... how about a modprobe -a -t rtai ? that should load all modules from /lib/modules/kernel/rtai adeos is elsewhere you don't have to worry about that cause you compiled it in, but others will i know ;) also ... i want to try out what happens in this case bc it loads rtai_ksched, and rtai_up i think :-? you know, flexibility is both a strength of linux, and one of it's greatest weaknesses oh, that would be interesting probably whichever one tried to load second would fail doesn't work because it takes them alfabetically so dependencies are not ok Only for information: I am also writing a driver for a ISA Board, with everyting onboard fpr a servo system. If it runs it will be very cheap i have made out of Johns parport driver a sKeletor driver I think it will be helpfull also for emc2 jep, I only kno how to write drivers for emc2 because John and paul told me :-) emc1 is a little bit easier *G* no offence John :D easier to write drivers? but less flexible The Board: 3x 16 Bit DAC out 3x encoder input 21x digital out 46x digital in encoder input on 7266? or 7166? no a special asic i see, sounds nice jep and how cheap? I got two for 8 EUr each :-) very nice can you get more? and for how long? and how many more? that's the problem with ASICs - product lifetime is only a couple years there is somebody how sells them on ebay. I think a company will empty it's storage oh... so when they're gone.... that's it he said something about 50-100 pieces I think I saw a link in last weeks log... right? link to what? jep so buy some cards more, and you will have no trouble if you blow some !! to an ebay auction but no picture thre there anyways... I hope this week I'll get some time to look over the skeleton and maybe write a driver for my card http://cgi.ebay.de/ws/eBayISAPI.dll?ViewItem&rd=1&item=5914265290 if you have any questions about drivers, just let me know sure are you still interested in that step-by-step install of rtai ? I wrote it along with my last rtai install I wanted to finish the driver this week, but i was too lasy :-) Maybe next weekend, and then I have to test every function. I hope the card has no bug don't think there are any breaking news I personally won't have time to try anthing like that soon kay ;) but it's still good to have. maybe we can put it at linuxcnc.org mybe I have to tune the skeletor driver a little bit more well.. that's it for me If the ISA-Card is ok, then I will talk to that guy about the other cards. Maybe that is a good thing for EMC I'm calling it a day sure wouldn't mind to get my hand on some I'll be quiting soon too... my wife has been away since Thursday, due back within an hour :-) Imperator: you're from germany? jep cool ;) and you ? a little bit more to the east romania ah I was there last year really? where? in Bukarest I was on the way to rusee/bulagien I'm on my way tomorrow to bulgaria ;) but bukarest is a shitty town :D anyways but it looks nice I was there only for two days :-( do you have the email address? for this seller? jep, but you can easily contact him over the ebay system. But I suggest wait a week. then I can say you if it realy works. The card is a special development for siemens, and the doku (german only) is very basic. Maybe the card hase some problems and siemens don't wantet to buy them ???? hmmm could be what are you doing in romenia, are you working at the university ? I'm an engineer at a company that sells robots (welding robots) ah it's the romanian CLOOS representative maybe you heard of them they're located about 100 km north of Frankfurt no sorry anyways... I'm gone now bye guys is paul still around? yes any thoughts on why task can't talk to motion? on BDI-Live ? yeah cos something is well and truly stuffed. lol where is the address of the emcmot struct suposed to be set ? commands are getting to the motion controller (it prints them to the kernel log) but either the commands are failing or (more likely) the status isn't getting back init_com_buffers() in motion.c does the setup on the RT side to be honest, I don't recall where it is done on the user side I didn't do anything (much) with that code here we go... usermotInit() in usrmotintf.c which doesn't appear to be called - Or atleast the printf statement isn't appearing. which printf? in usrmotInit? I don't see one * jmkasunich puts a few in miss-read the fprintf as a printf... Not having a good day. init gets called, the alloc succeeds /proc/rtapi says that block gets allocated (658664 bytes) and has one RT and one user space owner * jmkasunich_4 wonders if the fact that the block exceeds 64K has any effect not likely - both command and status are in the first 64K jmkasunich: At a guess, I'd say the emcmotStatus pointer is corrupt. printing the head & tail counts from usrmotReadEmcmotStatus() shows wild miss-matches. anyway I wanted to talk to robin about a possible commission to do a really big plasma for a local stamping firm coo. really big? well at least 5x10 ft perhaps larger right .. thats not that big haha 1.5 by 3m is sort of standard big plasma head though client wants to go 200 ipm or so at least in what thickness 'bout as many amps for say 5mm steel? 200ipm in 5mm ... one moment ...we have to mix our units here you need around 100 amps to get that sort of speed in 5mm ok thanks big $$$$ mmm .... about 8k has to be big $$$ or I can't take it on for the plasma and torch and leads I mean for the whole machine oh, right .. ;) 50K+ right ... well, if theres that much in the budget, then you can go for a 'proper' machine, rather than one of my cheapies :) yes industrial grade or ... I design and they build you could do it that way, the mechanics are not exactly hard I wonder about timing belt drive in an often used industrial machine one of these: http://www.innerlogic-inc.com/sr_100i_plasma_cutter.htm, plus the torch height controller ... http://www.innerlogic-inc.com/atc_arc_voltage_control.htm handles the plasma part of the job, inner-logic would cut you in for around 25% ... theres about 10~12K USD there yeah... don't know if I could get a price break from our Miller or Hobart the belts work fine, but you need to keep them out of the crap, and keep the stick/slip friction low or its judder city yeah well I just would not want to go hardened rack and forget using linear rails ... * robin_sz nods that plasma dust kills it in a month oh yeah... rails would die real quick I use plain cold-rolled steel bars, and 30mm cam roller bearings ... I see no problems I have to cost this up and see if I want to do it yeah. Might be a real money loser for a one up machine you could probably just refer to inner-logic ... get them to cut you in 10% and they will sell a plasma, and probably have a machine-maker that they work with over there ... tbh, mine are not 50k sorta machines .. more like 10k sounds like a plan cheapies for the low end, not 50k nice ones just be a sales rep yeah, middleman can be a nice position a lot less work * robin_sz nods and I am booked up pretty good here anyway engineering work and that silly turkey call production yep, give the guys at inner-logic a ring, they make a real nice plasma, and will talk turkey. you should be able to get a cut for a refferal oh yes, the turkey calls :) cool and, I dont ship mine stateside anyway :) I was wondering.... my product liability insurance is on condition I dont sell any over there they couldnt even give me a quote to sell in the US, no one wants to quote for that I don't even carry any coo. those turkey calls could explode in someones face then I'd be in big trouble I run as a shell company, assetts held by my holding company, house etc in wifes name etc etc .. and still have insurance wow better safe than sorry yup I just have to decide what I want to do here now that things are a bit better most $$$=engineering least$$$=furniture carved signs in between if anyone sues, the shell company goes pop, and they get the phone and the paperclips, the rest of it they can sing for. signs are fun, engineering can be fun, furniture? just work huh? no I like it all I would say furniture is the most fun ok. is that not the sort of time you need an employee? actually sold some $1000 chairs too I will need a spray finisher soon but right now I am still making a bunch of those stepper based robotic test fixtures heh so I'm chewing up just metal had to do bearing stuff all last week bearing stuff= .02 mm tolerances On the Boxford with Themac tool post grinder have to do it by feel since the Boxford is not that repeatable I don't like Tool Post grinders have to carefully cover up the ways The themac runs at 42,000 rpm for the smallest id grinding nasty $$$ too...it has class 9 bearings don't recall what the ISO grade is for that It's the highest wow im supprised the boxford can hold the tolerances you need vibration must be a killer The Boxford is a good machine...everything has to be tight though best bearings ive seen were on our Ampex 2" VTR the head had air bearings ;) the last grinds have to be done by look and feel rather than dialing in numbers even measuring .02 mm is not easy quite. .02mm is tiny, get ye one of those ruby-tipped wonder things yup lost one shaft...went too far I'm back.... * paul_c hides again Hi Steve Do a cvs up Just did update and attempted make - make failed with >> Makefile:85: *** missing separator. Stop. open up the Makefile with an editor ok do you see anything that looks like ">>>>>>> 1.27" around line 85 ? no, line 85 is totally blank and there is no 1.27" anywhere nearby emctest: delete the Makefile and do another cvs up Just "makefile" and not rules or inc ?? hangon.... Which Makefile are you looking at ? the one in home/emctest/emcdev/emc2 and did you see a directory with the error message ? make was run from within same directory and error message was as posted, no mention of a directory OK... Open up the Makefile in emc2/src/rtapi line 85 of that file is "<<<<<<< Makefile" right... Delete THAT Makefile and do a cvs up make failed to link "Linking /home/emctest/emcdev/emc2/bin/halcmd" --/usr/bin/ld: cannot open output file /home/emctest/emcdev/emc2/bin/halcmd: No such file or directory from emc2 dir - D a "make clean" then ./configure and finally, make OK make is running, I had done the clean but not the configure make was successful, EMC2 seems to be running, TkEMC popped up, time for champaign? not with that spelling. hey, i'm an engineer from the responsiveness of the machine it looks like realtime is sucking up cycles so it must be running can I do anything with TkEMC? hi, I'm back Do a cvs up I've modded realtime to use the RTAI script I see that alex tried the script on his box, it errored out looking for a file Well, some good news at least... Steve has emc2 running on his laptop I saw... I'm surprized he didn't run into the comms error At least I think I do. Is TkEMC functional with EMC2? Can I do anything useful without a machine attached? run programs and watch in the backplotter open up the backplotter and play. the new version didn't work on my RC46 box looks like it failed to load adeos did you delete the scripts/rtapi.conf and re-rn make ? ran ./configure and re did make, that should do the same thing * jmkasunich_4 redoes it check that scripts/.runinfo exists it's there I like our script better than theirs You want to craft something that works on all boxes ? their's doesn't work on all boxes ours is simple - "load this list of modules" where this list = adeos + rtai_hal + rtai_ksched + + our version keeps the messiness in one place - once the module list in rtapi.conf is correct, everything else is simple instead of having "if [ RTAI = 3 ]" in several different places and eventually "if RTAI = 3.2" or "if RTAI = 4" that's what the conf file is for, to keep those if's out of the scripts The only way you are going to avoid that is to get to gtips with autotools and auto-generate the realtime script. plus I don't like hidden files (.runinfo) if that's what it takes... You want to do a KL on the RTAI list, go right ahead. KL? Kim Lux no, not at all I'm not gonna complain about their script, I was just gonna ignore it OK, after some fussing with scale factors, I have 3D_Chips.ngc showing up in backplot. John - what comms error did you expect? Is it easy to shut down all the diagnostics printed to the console, or do they not really limit speed? Tomorrow I will try to get hooked to real machine and thrash out the ini file. many of those diagnostics are printfs in the C code, can't be turned off they'll be removed as testing progresses I don't think they slow things down enough to matter paul_c: the only thing I thought about questioning (not complaining, just questioning) is the insmod search algo it's behavior is quite odd... there may be good reasons for the oddness I dunno if you read back when I was describing it to alex if foo.o is a module, and bar.o is a link to it, it behaves differntly than if bar.o is a module and foo.o is a link to it that's just strange also ( I haven't tested this yet but it follows from the behavior I've seen) if foo.o is a module in the modules tree and I have another foo.o elsewhere if I put bar.o in the tree, as a symlink to my foo when you try to insmod foo, you'll get mine, not the one in the dir * jmkasunich tests it depmod follows symlinks... it's not that it follows them, it's that it uses the name of the link target for the search, not the name of the link itself if bar points to foo, insmod bar fails insmod foo follows the link and loads foo (but it loads it as bar, you need to do rmmod bar to remove it) unless of course you are in that directory when you run insmod then insmod bar loads foo (and calls it bar) and insmod foo loads foo and calls it foo well the security risk is not too severe... it does indeed follow the symlink to my "trojan" foo, but won't load it unless it's owned by root the behavior of insmod when started from within the directory is what you would want insmod rtai_ksched follows the link and loads rtai_up, but calls it rtai_ksched, so a later rmmod rtai_ksched can remove it neither the insmod or the rmmod command needs to know that rtai_ksched is really a symlink to rtai_up but issue the commands from anywhere else, and it doesn't work I'm not going to worry about it... The current version works for me, and I suspect on any other RTAI-3 install. not on my RC46 install I dunno what is different between mine and yours did the one committed immediately prior to your most recent change work on your system? (the one that follows the symlink, and puts both the link and rtai_ksched in the MODULES list) it worked if run from the scripts dir - That's way I added a ( cd $EMC"_HOME/scripts... to realtime but not if run from say emc2? right. which is why I made another commit. So, are you using the latest files ?? did a cvs up when you told me to yeah, my realtime has the cd in it but you added the cd _after_ you changed to using their script I was asking if the lastest revision of _our_ script worked (after all my commits, but before you started using theirs) no what didn't work? adeos ? our script never failed to load adeos, that I can recall we had problems with ksched, not adeos in fact, the problem I have with their script is that if fails to load adeos I have no idea if it would load rtai_ksched cause it never gets that fa far * jmkasunich_4 's head is spinning you feel rather strongly that we should use the RTAI supplied script, don't you? I just want something that works - If it entails using the RTAI scripts, fine. what system are you testing on? RC46 or something else? rc46 that was used for development. strange that what works for me doesn't work for you and vice-versa just noticed that the new version of the script doesn't get the modules from /lib/modules/`uname -r`, it gets them from usr/realtime/modules need to burn an rc46 CD and do a fresh install. do you have a copy of adeos.o in /usr/realtime/modules? I don't it is in /lib/modules/`uname -r`/kernel/adeos/ (not /rtai/ with the rest) I think that's why it's not working for me /lib/modules/{uname -r}/kernel/adeos (as I recall) just rebooted to another flavour.. OK - 2.4.22-adeos works (pre rtai-3) right - that uses the old approach, loop thru a list of modules calling insmod need to change realtime:101 to a string comparison to head of a bash error Time to go.