#emc-devel | Logs for 2010-07-28

[01:06:48] <mozmck> jepler: you around?
[01:41:29] <jepler> mozmck: I am now
[01:41:52] <jepler> mozmck: I think you can ignore that warning
[01:42:11] <mozmck> ok. I didn't know if a different directory name would be better
[01:42:41] <jepler> the source package will be created with the directory name it said, linux-2.6.32
[01:43:20] <jepler> the standard debian/ubuntu way is for the source package name and the source directory name to be the same
[01:43:22] <mozmck> if I look in the .tar.gz the root directory is ubuntu-lucid
[01:43:27] <jepler> oh really?
[01:43:40] <mozmck> yup
[01:44:46] <jepler> ah, it's the extraction process that gives it the right name
[01:44:47] <jepler> I never knew that
[01:45:04] <jepler> so yeah I guess it would be a tiny bit better if you renamed the source directory to linux-2.6.32
[01:46:03] <mozmck> another question: how in the world is the 99-rtai.rules file created in the rtai package??? I used the debian directory from the rtai y'all did with hardy, and I can't find any reference in the 4 files to 99-rtai.rules, but it creates it anyhow!
[01:46:59] <mozmck> Ok, I'll look at changing the dir name. Will it mess up git or does it care?
[01:47:09] <jepler> git shouldn't notice
[01:47:17] <mozmck> ok.
[01:47:55] <jepler> oh, I forgot something. you probably actually want to build the source package with dpkg-buildpackage -S -i
[01:48:03] <jepler> otherwise it includes the .git dir which is generally not what is desired
[01:48:10] <jepler> actually, check the tar to make sure that's right..
[01:48:25] <jepler> anyway, the 99-rtai.rules file is created by 'make' I think
[01:48:25] <jepler> GNUmakefile.am: grep -q RTAI:`basename $$b .udev` $(DESTDIR)/etc/udev/rules.d/99-rtai.rules || \
[01:48:27] <jepler> GNUmakefile.am: $$sudo sh -c "( echo ; cat $$f ) >> $(DESTDIR)/etc/udev/rules.d/99-rtai.rules" ; \
[01:49:18] <mozmck> The reason I ask about the 99-rtai.rules file, is that I get this error " trying to overwrite '/etc/udev/rules.d/99-rtai.rules'" etc if I try to install two versions of rtai-modules (for 2 different kernels)
[01:49:47] <mozmck> so it's part of rtai?
[01:49:50] <jepler> yeah, I think you'll have to remove the old kernel before installing the new one
[01:50:01] <jepler> er, old rtai-modules
[01:50:12] <jepler> those lines I pasted were from the rtai makefiles
[01:51:00] <mozmck> well, since rtai is installed in separate directories named after the kernel version, I would think you should be able to have multiple copies
[01:51:17] <jepler> there's no place you can put a udev rule besides /etc/udev/rules.d
[01:51:43] <jepler> we've always had one realtime kernel per distro so it's never been a problem, and I don't think it will be on most users' machines
[01:52:25] <mozmck> ok. I'll just force it or remove the old. just a minor nuisance.
[01:52:40] <jepler> an actual problem for a few people, like those building the kernels
[01:53:36] <mozmck> Yeah, if I could tell dpkg not to error out on that file, like only install it if it's not there, that would be great
[01:53:56] <mozmck> here's the line I used to build source like you posted earlier: dpkg-buildpackage -S -us -I.git -I.gitignore
[01:55:40] <jepler> I think that covers it too
[02:09:22] <mozmck> when using update-dch-from-git, how do I tell it to use my actual name and email instead of my login name.
[02:47:41] <jepler> mozmck: hmmm let me check
[02:48:35] <jepler> mozmck: set the environment variables DEBEMAIL and DEBFULLNAME to the values you want
[02:48:57] <mozmck> I guess I should do that in .bash_profile?
[02:49:11] <jepler> I think that'll make it stick in all terminals
[02:49:28] <jepler> export DEBEMAIL="..."; export DEBFULLNAME="..."
[02:50:29] <mozmck> thanks. I wasn't even sure where to look to find that info.
[02:52:19] <jepler> update-dch-from-git uses the program dch and it's documented in the dch manpage
[02:52:55] <mozmck> I see. Thanks.
[08:11:08] <CIA-4> EMC: 03cmorley 07v2.4_branch * rae79deea7e24 10/src/emc/usr_intf/pncconf/pncconf.py: create a desktop entry in the right Desktop directory
[11:58:03] <moopy> anyone have suggestions what is best way to multiplex position commands from axis?
[13:45:19] <mozmck_work1> jepler, cradek, etc, I put a complete new set of packages up at http://www.linuxcnc.org/mozmck/newest
[13:46:07] <mozmck_work1> The only change to the kernel is the version number from 2.6.32-22 to 2.6.32-122, and then a recompile and rebuild of packages.
[13:46:36] <cradek> sweet!
[13:46:40] <mozmck_work1> I had to rebuild rtai and emc for that kernel as well, and there are source packages for everything there as well.
[13:47:27] <cradek> do you want me to try them on that dual p3?
[13:47:40] <mozmck_work1> I built a livecd with these packages and did some preliminary testing and it all seems to work.
[13:48:01] <mozmck_work1> I guess so! did the previous ones work?
[13:48:03] <cradek> cool. if you can upload it, I can try it on several machines
[13:48:18] <cradek> yes the previous ones are fine - that's what it's currently running (this is max, the little mill)
[13:48:58] <mozmck_work1> It will take about 4 or 5 hours for me to upload, so I'll probably have to do that tonight.
[13:49:55] <cradek> did you get the cd install to point you at the linuxcnc apt repository automatically?
[13:50:34] <mozmck_work1> Someone here mentioned removing openoffice and putting the devel packages on the livecd, so since I had already removed openoffice I put on all the dev packages and had a little room to spare.
[13:51:19] <mozmck_work1> As a result, I was able to compile and use a couple of .comp components directly from the livecd
[13:51:20] <cradek> you mean all the build deps fit?
[13:51:49] <cradek> (or just enough to run comp?)
[13:51:49] <mozmck_work1> yup. and the gnumeric spreadsheet
[13:52:18] <mozmck_work1> g++, gcc, build-essentials, etc. I think everything needed to build emc
[13:52:54] <cradek> I'm shocked it all fits in place of open office (with lyx etc)
[13:53:07] <mozmck_work1> no lyx :)
[13:53:27] <mozmck_work1> That's only needed to build the docs.
[13:53:28] <cradek> oh ok, not full build-dep for the package rebuild then
[13:53:47] <mozmck_work1> not for package rebuild, but it should be fine for emc2 rebuild
[13:53:55] <cradek> ok cool.
[13:54:01] <cradek> I can't wait to try it.
[13:54:23] <jepler> that is excellent news
[13:54:28] <mozmck_work1> but at least people can build components without having to install a bunch of extra stuff.
[13:54:36] <jepler> I'll work on putting those packages in the right form for an apt repository
[13:54:38] <cradek> yes
[13:55:20] <mozmck_work1> I didn't get the linuxcnc repo in apt-sources on the livecd yet. I forgot last night, but that is minor.
[13:56:29] <cradek> it stinks that minor things cost you another 4 hour upload :-/
[13:56:38] <cradek> working with cd images is no fun
[13:57:00] <mozmck_work1> cradek: I did get your truetype tracer on there, and the firmware packages from hardy
[13:57:31] <cradek> cool! did you use the one from http://timeguy.com/cradek/01276453959
[13:57:46] <mozmck_work1> yes
[13:58:21] <mozmck_work1> that's one reason I didn't upload the cd image yet. I can make a few changes in 5 minutes.
[13:58:35] <cradek> hm, since I don't have those in an apt repository, I wonder how to point people to the corresponding source
[13:58:50] <mozmck_work1> heh, ask on irc?
[13:58:51] <cradek> (other than just letting google do it...?)
[13:59:15] <mozmck_work1> why not put it in the lucid repo?
[13:59:34] <cradek> good idea, I will
[14:01:51] <skunkworks> I can test it on a bunch of computers also. (when the image is uploaded)
[14:02:55] <jepler> cradek: I'll put it in the lucid package repository on linuxcnc if that's OK with you
[14:03:02] <jepler> truetype-tracer, that is
[14:03:03] <cradek> yes please do
[14:03:27] <cradek> then people will automatically have access to the source in the most natural way
[14:04:37] <jepler> mozmck_work1: oh, have you put the emc2 board signing key on the cd as a key trusted by apt-get?
[14:04:51] <mozmck_work1> uh, I don't remember.
[14:05:17] <mozmck_work1> Where do I check that again?
[14:05:53] <mozmck_work1> I'll make a list of these things and do them before I upload.
[14:06:59] <cradek> sudo apt-key list
[14:07:05] <mozmck_work1> ok.
[14:07:36] <jepler> pub 1024D/8F374FEF 2008-04-15
[14:07:36] <jepler> uid EMC Archive Signing Key <emc-board@linuxcnc.org>
[14:07:37] <jepler> sub 2048g/17BAA37C 2008-04-15
[14:07:40] <mozmck_work1> One thing I did do is a full-upgrade so all the packages on the CD are the latest from ubuntu as of last night.
[14:12:46] <jepler> I'll probably have the package repository working within 20 minutes
[14:13:16] <jepler> it is important that this works right after installation from the live cd
[14:13:31] <jepler> mozmck_work1: will you please test that before uploading?
[14:16:24] <mozmck_work1> ok, I may not have time tonight, but I can tomorrow anyhow.
[14:17:13] <jepler> ok
[14:17:57] <mozmck_work1> The emc2 packages on there right now are from last nights git. I think you mentioned that we may want another point release before the final CD?
[14:19:11] <jepler> yes, the problem being I won't have time before this weekend
[14:19:42] <mozmck_work1> I don't mind uploading more than once either. I'll upload what I have for testing tonight or tomorrow.
[14:19:59] <jepler> ok, that works for me too
[14:20:14] <mozmck_work1> no problem. That will give a little more time to test the pre-release cd.
[14:22:33] <cradek> where can we get together to celebrate? looks like Tulsa is about midway. I've never been there.
[14:25:08] <mozmck_work1> :) It's only about 4 hours from me.
[14:26:05] <cradek> hm, it's 7 from here. wichita?
[14:26:17] <mozmck_work1> about 6
[14:26:50] <cradek> with stuart's long list of machines needing retrofits, we should go to wichita anyway and kick him in the pants to get started
[14:29:45] <skunkworks> ouch
[14:38:30] <cradek> this one especially: http://mpm1.com/enshu.html
[14:39:11] <skunkworks> hey - that looks like yours
[14:39:29] <cradek> yeah it's similar. very slightly bigger I think.
[14:39:40] <cradek> carousel is much bigger - it's a long chain
[14:40:06] <skunkworks> so - now that you have done 1 jr.. this one should take about 3 days. ;)
[14:40:08] <cradek> he says this is one of the machines they used constantly, and it is dead
[14:40:16] <cradek> haha, yeah right
[14:40:40] <skunkworks> do they know what the issue is?
[14:40:42] <cradek> his jr's currently dead too
[14:41:06] <cradek> all the I/O is flaky - it is fiber optic (??)
[14:41:24] <cradek> fanuc doesn't have any clue and wants to randomly try replacing boards at great expense
[14:41:48] <skunkworks> yeck
[14:42:24] <jepler> the enshu's I/O is fiber optic?
[14:42:57] <cradek> yes that's what he said
[14:43:40] <skunkworks> high tech
[14:44:08] <cradek> he said it is 3 boards - one is $10k - then he stopped looking
[14:44:45] <cradek> good retrofit candidate. nice console for touchy. nice random toolchanger.
[14:45:01] <skunkworks> all of the k&t i/o is optical.. ;) (opto22's)
[14:45:20] <cradek> heh, that's slightly different from fiber optic...
[14:45:26] <skunkworks> he
[14:46:39] <skunkworks> cradek: have you played with the amc drives yet?
[14:47:43] <skunkworks> I would be interested in how the encoder feedback works for velocity control. (initally we are going to use the tachs - but it would be cool to know the encoder feedback works well also)
[14:48:06] <skunkworks> (our big amps will do both)
[14:48:09] <cradek> skunkworks: yes - I tuned one to a little 10A servo motor I have - it works great
[14:48:19] <cradek> encoder velocity feedback seems to work great
[14:48:25] <skunkworks> nice!
[14:48:49] <skunkworks> did you tune the velocity loop first? with a battery or such and scope?
[14:49:12] <cradek> when giving a step change in velocity command I get a velocity overshoot of maybe 15% but then it settles perfectly. no oscillation
[14:49:43] <cradek> I could not get rid of that - it is surely integrator windup - if I ramp the velocity so it can always keep up, I bet it doesn't do it
[14:49:54] <jepler> OK, I think the lucid package repository is live now
[14:50:24] <skunkworks> cradek: interesting.
[14:50:54] <cradek> skunkworks: yes just scope on velocity monitor output and ear
[14:50:56] <jepler> it seems to think it's able to install emc2-dev on this lucid system that's never had emc2 installed yet: http://pastebin.ca/1910158
[14:51:08] <cradek> it took a while. lots of fiddly switches and knobs and not very good instructions.
[14:51:30] <cradek> jepler: yay!
[14:51:40] <jepler> linuxcnc.list: http://pastebin.ca/1910159
[14:51:43] <cradek> libgfortran3?
[14:52:19] <jepler> cradek: yeah, that's legit. It's required by python-numpy
[14:52:23] <cradek> ah
[14:52:58] <jepler> emc2 (image2gcode) -> python-numpy -> libblas3gf -> libgfortran3
[14:56:10] <mozmck_work1> neat! is the repo address the same as you gave me yesterday?
[14:58:44] <jepler> mozmck_work1: the repo address is shown here:
[14:58:45] <jepler> http://pastebin.ca/1910159
[14:58:51] <jepler> I think it's the same as I had told you
[14:59:05] <cradek> it matches my memory
[14:59:27] <mozmck_work1> yep. thanks.
[15:01:28] <jepler> bbl
[15:03:27] <jepler> haven't tested it, but I also put this up: http://www.linuxcnc.org/lucid/emc2-install.sh
[15:03:46] <jepler> and I noticed the hardy install.sh was still pointing at 2.3 and changed it to 2.4.
[15:20:45] <moopy> is anyone interested in developing a universal machine configuration gui?
[15:31:47] <JT-Work> I'm interested in coding up a dxf to g-code generator for lathes and or plasmas :)
[15:31:59] <JT-Work> in c++
[15:34:12] <skunkworks> why not python? I think a ton of work is already done here... http://sourceforge.net/projects/vec2ngc/
[15:36:20] <JT-Work> dunno why, I kinda thought c++ would be faster somehow I guess
[15:36:33] <mozmck_work1> it is :)
[15:36:41] <mozmck_work1> not to write though...
[15:37:33] <JT-Work> I didn't know micges had done that
[15:37:44] <mozmck_work1> moopy: I've been rolling around ideas in my head for a universal machine configuration gui, but haven't had time to do much with it.
[15:38:17] <JT-Work> I did write some of the g-code generators on the wiki in python
[15:39:54] <JT-Work> I'll take a look at his work this evening thanks for the link skunkworks