#emc-devel | Logs for 2010-05-15

[01:02:48] <Dave911> jepler: The Mesa LED patch: It is exactly as Andy described. I tried out his patch and the pins appear and work as he described. I think those LEDs could be quite useful for diagnostics and error messaging.
[02:48:30] <cradek> mozmck: how's it going? what's the latest news?
[03:15:50] <mozmck> Hey cradek: haven't had time to do much again! I compiled with the latest patch and rtai 3.8.1, and still have intermittent boot problems.
[03:16:18] <cradek> they seem convinced it's grub doing it - sounds so bizarre.
[03:21:27] <mozmck> I saw that, but I'm not sure how downgrading grub will work on 10.04
[03:21:44] <mozmck> I may try that though.
[03:21:45] <cradek> yeah, possibly badly
[03:21:57] <mozmck> When it boots though it seems to work great!
[03:22:07] <cradek> that sounds very close...
[03:22:34] <mozmck> Compared latency on one machine with 8.04 (dual booted on same machine), and it was very close.
[03:22:50] <mozmck> Even though my 10.04 kernel is compiled for SMP.
[03:23:08] <mozmck> 10.04 was a little worse though.
[03:23:46] <cradek> with isolcpus it's probably a lot better
[03:24:01] <cradek> at least that's how my dual p3 works
[03:24:06] <mozmck> I'm using the CONFIG_PREEMPT instead of the normal desktop setting, but I don't know if this hurts or helps.
[03:24:23] <mozmck> yes, but this machine was single core.
[03:24:44] <cradek> oh I see
[03:25:15] <mozmck> I haven't upgraded my dual-core machine yet. It's still on 9.10
[03:25:37] <cradek> mine's dapper I think... :-/
[03:25:55] <mozmck> I'm tempted by the 6-core chips to build a faster kernel builder!
[03:26:08] <cradek> I bet mine is the oldest type of smp machine you'll find people wanting to use - I'll be sure to test on it.
[03:26:30] <cradek> hmm, I think I still have a working smp P2 machine, too
[03:26:39] <mozmck> You can see in my notes on the wiki the changes I make from the ubuntu generic config.
[03:27:21] <mozmck> not much. there may be some other settings I should make that I haven't found yet for performance...
[03:28:51] <mozmck> Is there a way I could get access to put my packages up somewhere directly? SWPadnos has done it before, but if I had a spot I wouldn't have to bother someone.
[03:29:42] <mozmck> Then other folks could start testing them...
[03:30:11] <cradek> you will need to ask SWPadnos
[03:33:28] <mozmck> ok.
[03:35:28] <cradek> this did not help? https://mail.rtai.org/pipermail/rtai/2010-April/022931.html
[03:36:45] <mozmck> nope. was worse on one computer.
[03:36:54] <cradek> mozmck: what would it cost to update you to a fast build machine? I would be able to help a bit. Maybe others would too?
[03:37:50] <cradek> when I did the breezy build each try ran pretty much overnight - it was so frustrating - I went and bought a new motherboard/processor/ram then.
[03:38:49] <mozmck> ooh, that would be neat, but I'd almost hate to do that! I think the stuff I was looking at would be about $600 or so. That's for the 3.2 gz 6-core ($300) and a motherboard to handle it and ram and etc.
[03:39:15] <mozmck> A quad core may do as well though and would be cheaper.
[03:41:43] <mozmck> That's what I've done several times now (compiled overnight) but I use that same machine for writing firmware and that compiler has to run in windows (blech!) I been real busy trying to get some firmware finished and debugged so haven't had the time to reboot and have the machine compiling as much.
[03:41:59] <cradek> $600 is a lot for one person but not much for 10. I'll send $60 if it would help.
[03:42:44] <cradek> (wish I could do more - I'm not feeling very rich right now with cnc workshop coming up)
[03:43:23] <mozmck> That would certainly help. I might have some more junk you could use too.
[03:44:01] <mozmck> I know the feeling. I won't be able to make the workshop for time as much as money.
[03:44:14] <cradek> oh you aren't coming? darnit.
[03:45:20] <mozmck> no, I would like to though. If there's another one in wichita later I could probably make that though ;)
[03:46:01] <cradek> yes that is definitely tempting. stuart has a dozen projects he'd love to have us all help on...
[03:47:05] <mozmck> I better get to bed. Lots to do tomorrow. goodnight.
[03:47:11] <cradek> goodnight
[07:31:37] <ries_> ries_ is now known as ries
[12:56:39] <jepler> grumble, andy's patch doesn't apply for me with 'git am' or 'git apply'.
[12:56:51] <jepler> I guess it does with 'patch -p1 <' but that's just barbaric
[12:57:47] <SWPadnos> heh
[12:57:55] <SWPadnos> that's what you get with diff-u
[13:00:35] <jepler> it *looks* like a git patch, with a/ and b/ prefixes
[13:02:24] <SWPadnos> huh
[13:17:50] <jepler> OK, I've mostly got the whitespace problems sorted out
[13:17:59] <cliffb> SWPadnos: Couple questions about NML if you don't mind...
[13:18:17] <SWPadnos> sure, but I probably can't answer them (if you don't mind :) )
[13:19:49] <SWPadnos> this should have an NML reference: http://www.linuxcnc.org/docs/EMC2_Developer_Manual.pdf
[13:20:02] <SWPadnos> according to the page it's linked from on linuxcnc.org
[13:20:54] <cliffb> Np.. Ill take a RTFM.. If i directly communicate using NML I assume you have to link to the souce lib, is it portable across platforms?
[13:21:42] <SWPadnos> should be
[13:22:20] <SWPadnos> it uses shared memory and/or sockets (and also optionally serial, but I don't know if that ever materialized)
[13:22:42] <cliffb> k.... I'll start source reading to see how it's used + that manual...
[13:22:53] <SWPadnos> I don't know if they just use POSIX or something for the shared memory code, or if there are platform-specific headers or something else
[13:23:39] <SWPadnos> ok. sorry I can't be of much more specific help, it's been a few years since I looked at it
[13:25:20] <cliffb> Np... more than helpful... I wrote a realtime controller using RTAI, then found EMC (And love it...) but it biggest project I've started digging into..
[13:25:31] <SWPadnos> heh, yeah
[13:26:11] <SWPadnos> and don't worry if it's confusing finding things - I still don't know where in the source to look :)
[13:27:28] <cliffb> Yeah... the realtime/user code really adds a lot of overhead to browse through.
[13:37:04] <SWPadnos> don't even try to trace how a line of G-code actually gets parsed and eventually turned into some motion :)
[13:37:18] <SWPadnos> gotta run, looks like it's time for breakfast
[13:43:05] <mozmck> cliffb: I would be interested in helping on your project if it will be open source. I've sorta started something like it using FLTK but haven't had time to get very far.
[13:48:40] <cliffb> Yeah I definatly will give it out open source.. I havn't got all that far either. If you know Qt and want to help by all means I could use a hand.
[13:49:18] <mozmck> Don't know Qt yet, but I can learn it.
[13:57:45] <cliffb> Is FLTK a gui designer for the TK toolkit? How far along did you get... would it be better to focus on that?
[14:01:48] <mozmck> no, FLTK stand for "Fast Light Tool Kit" and it has a designer with it. C++ and small, it's normally statically linked.
[14:03:18] <mozmck> I mostly was working on the overall object structure and writing shell classes so far. My plan was to make a custom designer as that looked like it would be easier than just modifying an existing one.
[14:35:48] <jepler> wtf -- "git pull --dry-run" just modified my working tree
[14:35:56] <jepler> and created a commit
[14:45:48] <alex_joni> cliffb: rcslib is multi-platform
[14:45:55] <alex_joni> emc1 used that, and it's still available
[14:46:19] <alex_joni> emc2 uses libnml a small subset of rcslib which drops some unused parts, including multi-platform support
[14:46:30] <alex_joni> but rcslib and libnml can still communicate together
[14:46:50] <alex_joni> so you could write a program linked to rcslib and have it talk to emc2 over NML
[14:50:40] <mozmck> alex_joni: can you use the rcslib from http://www.isd.mel.nist.gov/projects/rcslib/?
[14:51:38] <mozmck> or does it need an older version (or has it changed)?
[14:55:56] <cliffb> Alex: I'll look into that, thanks.. Any source code which does this already i can browse?
[15:11:43] <CIA-2> EMC: 03jepler 07master * r920b18a1e3c2 10/ (5 files in 3 dirs): hostmot2: add support for onboard diagnostic LEDs
[15:11:50] <CIA-2> EMC: 03jepler 07master * rab151d6dae81 10/ (7 files in 3 dirs): Merge branch 'mesa-3pwm-led'
[15:11:51] <CIA-2> EMC: 03jepler 07master * rc9add70ba1d5 10/ (6 files in 3 dirs): hostmot2: add support for three-phase pwm
[15:13:14] <mozmck> cradek: I downgraded grub and it booted right up. I'm going to have to test this more but that may help after all!
[15:19:47] <cradek> that is good news (I think)!
[15:33:48] <mozmck2> logger_dev: bookmark
[15:33:49] <logger_dev> Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emcdevel/2010-05-15.txt
[15:45:04] <JT-Dev> where dies pncconf get installed with 2.4? I don't see it in usr/bin
[15:45:37] <jepler> $ dpkg-query -S /usr/bin/pncconf
[15:45:37] <jepler> emc2: /usr/bin/pncconf
[15:45:40] <jepler> it should be
[15:45:50] <JT-Dev> let me look again
[15:46:46] <JT-Dev> gee wiz I'm blind today... thanks jepler
[15:47:07] <JT-Dev> must be time for a nap
[16:26:35] <alSMT> I just came in here to "rally" for a single block pin it would make any line like having a M1 on it "finish line and wait"
[16:31:41] <alSMT> anyway I think it would be a great feature addition as it is now you must pause and goto step causing the tool to stop for a moment
[16:31:58] <alex_joni> mozmck: I don't know.. never tried
[16:32:06] <alex_joni> I used rcslib from emc CVS
[16:32:31] <alex_joni> got it to build on windows ages ago, and used a TkEmc to connect to an emc2
[16:32:38] <alex_joni> it was in the 2.0.0 days though
[16:56:25] <mozmck> alex_joni: just wondered.
[16:57:57] <mozmck> cradek: yeah, I don't know how good news. I'm not sure right now if it would work well on the livecd. I don't know if just replacing it on the livecd would do the trick, or if there are some scripts that would need to be changed.
[16:58:06] <mozmck> Need to do more testing for sure though.
[17:49:21] <SWPadnos> mozmck, was the system you had to downgrade grub on an upgrade to 10.04?
[17:49:34] <SWPadnos> I had trouble with an upgrade, but a clean install worked fine
[18:53:23] <mozmck> SWPadnos: no, clean install on 2 different computers.
[18:53:43] <SWPadnos> oh well
[18:53:49] <mozmck> This is with the rtai patched kernel though - the standard kernel boot fine every time.
[18:54:00] <SWPadnos> oh, ok
[18:54:38] <mozmck> patched kernel boots about half the time, rest of the time it stops immediately after selecting the kernel with a blinking cursor in a black screen
[18:55:16] <mozmck> so far on this machine looks like it's booting every time with legacy grub
[18:56:14] <SWPadnos> I wonder what the difference is
[18:56:23] <SWPadnos> (not between grubs, but between what they do)
[18:58:31] <mozmck> probably. grub2 has some bug grub1 doesn't :) I've looked some at the bug reports for grub2 and it's hard to say if any are relevant.
[18:59:01] <SWPadnos> well, there may be some initialization that grub1 does that grub2 doesn't, wouldn't be a bug but would be a difference
[18:59:48] <mozmck> yeah. I don't know. I don't know enough about what grub2 does yet. Need to dig into it some more I guess.
[19:21:47] <cradek> wonder what the harm is in just downgrading grub and forgetting about it until the rtai folks figure it out
[22:46:05] <KimK> mozmck_work: Hi Moses, welcome back. Thanks in advance for all your hard work on the new version. And it's not like you don't have enough to do already, but when you're done, would you be kind enough to update BuildingUbuntuPackages at http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?BuildingUbuntuPackages as I think you will then be the person best qualified to update it.
[22:46:49] <KimK> And then maybe the rest of us can read it and learn something.
[22:48:13] <KimK> * KimK suddenly remembers he's in the developer's channel
[22:48:31] <KimK> I mean maybe us mere mortals can read it and learn something
[23:23:27] <mozmck> cradek: if I can get it to work on the livecd I can't see why we couldn't do that.
[23:23:57] <mozmck> Thanks KimK, I'll take a look at that page.