#emc-devel | Logs for 2008-11-25

[00:00:27] <jepler> seb_kuzminsky: I'm not sure what would set a limit lower than "unlimited" for scripts run from init..
[00:00:52] <seb_kuzminsky> buildbot runs from init, and it has a memlock limit of 32k
[00:02:01] <jepler> seb_kuzminsky: there's an /etc/init.d/buildbot that's run as root by init?
[00:02:07] <jepler> I'd try putting a ulimit-setting directive in that script
[00:02:07] <seb_kuzminsky> yes
[00:02:24] <jepler> I would expect this to work if running as root: ulimit -l unlimited
[00:02:49] <jepler> or 20480 or whatever
[00:03:08] <jepler> jepler@platte:~$ sudo -s
[00:03:08] <jepler> root@platte:~# ulimit -l
[00:03:08] <jepler> 32
[00:03:08] <jepler> root@platte:~# ulimit -l 20480; ulimit -l
[00:03:08] <jepler> 20480
[00:04:59] <seb_kuzminsky> yep
[00:06:22] <seb_kuzminsky> i put it in /etc/default/buildbot
[00:07:49] <seb_kuzminsky> building now, we'll see what it does
[00:11:10] <seb_kuzminsky> that looks good
[00:11:16] <seb_kuzminsky> yay it passed
[00:11:17] <seb_kuzminsky> :-)
[00:11:55] <jmkasunich> interesting "saga"
[00:12:23] <jmkasunich> my farm would not have that issue (not sure if that is good or bad)
[00:12:40] <jmkasunich> I log into the VM as a normal user, open a shell, and start the farm script
[00:13:01] <seb_kuzminsky> i just wanted it more automatic
[00:13:16] <seb_kuzminsky> it start the buildslave at boot time without human intervention
[00:13:39] <jmkasunich> I should probably do something like that here too, but I'm lazy
[00:14:16] <jmkasunich> and what I have is convenient - I start it with "./run_farm &", then do a "tail -f farm_log", so all I need to do is glance at the shell to see what is happening
[00:17:51] <seb_kuzminsky> ok i just updated the hardy and dapper buildslaves to run with "ulimit -l unlimited"
[00:18:07] <seb_kuzminsky> and kicked off a rebuild of the top-of-tree (TRUNK and v2_2_branch)
[00:18:13] <seb_kuzminsky> i think it will be all green
[00:18:23] <seb_kuzminsky> bbl folks
[01:21:21] <seb_kuzminsky> yay buildbot's all green :-)
[01:29:34] <jmkasunich> I bet I can fix that
[01:33:11] <SWPadnos> heh - not a problem on this end either :
[01:33:13] <SWPadnos> :)
[01:54:44] <seb_kuzminsky> tomorrow i will clean the corner of the garage that i use as a workshop
[01:54:46] <seb_kuzminsky> :-)
[01:55:57] <seb_kuzminsky> i'm gonna install mailman on emc2-buildbot and make a list for build status
[01:56:17] <seb_kuzminsky> like jmk suggested: start sending mails when it start failing, stop after it start working again
[01:57:06] <seb_kuzminsky> but now i'm gonna go read bedtime books for the kids
[01:57:45] <SWPadnos> see dick and jane sleep
[03:37:53] <jepler> seb_kuzminsky: thanks for committing that rtai_malloc fixaround
[03:38:18] <seb_kuzminsky> sure
[03:38:29] <seb_kuzminsky> i wasnt sure from your comment if you thought it was a good idea or not
[03:39:40] <cradek> hi guys
[03:39:52] <jepler> seb_kuzminsky: in an ideal world we'd build the fixed rtai packages instead of working around it, since we supply both to our users
[03:39:57] <jepler> but .. that's no fun
[03:40:22] <seb_kuzminsky> hi chris
[03:40:49] <seb_kuzminsky> i'll work with the rtai guys to get it fixed, then when the bug is dead & buried in like a decade we can remove our workaround
[03:41:33] <jepler> heh
[03:46:00] <jepler> 'night guys
[03:47:57] <cradek> bye
[03:52:20] <seb_kuzminsky> seeya jepler
[04:02:06] <CIA-38> EMC: 03seb 07TRUNK * 10emc2/docs/man/man9/hm2_7i43.9: add a note to load probe_parport before hm2_7i43
[04:04:18] <cradek> I would not have thought of that...
[04:04:36] <cradek> not at first anyway.
[04:11:43] <seb_kuzminsky> well that might not be it
[04:11:50] <seb_kuzminsky> epp man, what a drag
[04:11:52] <seb_kuzminsky> gimme pci any day
[04:12:03] <cradek> yeah I wasn't going to say that, but I was thinking it
[04:12:42] <seb_kuzminsky> we all were...
[04:12:48] <cradek> parport makes me think of all the cheap consumer devices of the late 90s (pre-usb)
[04:13:21] <cradek> when scanners (and tape drives, and quick cams, and card readers) were too cheap to be scsi, but there was no usb
[04:13:31] <seb_kuzminsky> back in the bad old days
[04:13:38] <cradek> that stuff was always just a big pain
[04:13:51] <seb_kuzminsky> remember zip drives?
[04:14:00] <cradek> sure
[04:14:10] <cradek> I bet the external ones were parport
[04:14:29] <cradek> I never did use them. they were just 'big floppies' and I already had networking.
[04:15:24] <seb_kuzminsky> the network pretty much saved all our bacon
[04:15:43] <seb_kuzminsky> grr, eric still got problems
[04:16:44] <cradek> has he sent a full dmesg?
[04:16:51] <seb_kuzminsky> no
[04:16:53] <cradek> would be nice to see it with probe_parport etc
[04:17:07] <seb_kuzminsky> when will they learn we're not clairvoyant?
[04:17:19] <cradek> never
[04:17:28] <cradek> it's so irritating
[05:34:59] <CIA-38> EMC: 03cmorley 07TRUNK * 10emc2/src/hal/classicladder/ (5 files): Changes to allow the user to select what variables MODBUS is mapped to. read coils,inputs and write coils done. TODO registers
[07:02:22] <CIA-38> EMC: 03cmorley 07TRUNK * 10emc2/src/hal/classicladder/protocol_modbus_master.h: Fix transposed constant
[14:34:11] <seb_kuzminsky> the rtai guys say the funny return value from rtai_malloc is a feature, not a bug...
[14:38:46] <cradek> but that seems so unlikely to be true
[14:39:07] <cradek> (does he go on to say why he thinks that?)
[14:54:27] <Guest198> feature by design..
[14:54:34] <Guest198> Guest198 is now known as skunkworks_
[15:00:16] <skunkworks_> alex is on the move.
[15:00:50] <alex_mobile> Hi
[15:01:18] <alex_mobile> Actually waiting for someone in my car.
[15:02:47] <alex_mobile> What's up?
[15:03:25] <skunkworks_> heh - nothing. all I know so far is that the funny return value from rtai_malloc is a feature according to the rtai people.
[15:03:59] <alex_mobile> Yeah, I read that too.
[15:04:14] <skunkworks_> :)
[15:04:43] <alex_mobile> Paolo is strange sometimes...
[15:04:54] <skunkworks_> and there is a 4 day weekend coming up.
[15:05:10] <alex_mobile> How come?
[15:06:19] <alex_mobile> Thanksgiving?
[15:08:52] <skunkworks_> yes
[15:08:59] <skunkworks_> turkey day
[15:09:17] <skunkworks_> we actually are going to have wild turkey this time.
[15:09:50] <alex_mobile> Wild as in crazy?
[15:10:06] <alex_mobile> :P
[15:10:19] <skunkworks_> :)
[15:10:47] <skunkworks_> should be interesting.
[15:11:33] <cradek> wow, look at classicladder/arrays.c:204 (in a good unix editor)
[15:11:37] <cradek> I wonder how that's compiled
[15:12:05] <Lerman_____> Lerman_____ is now known as Lerman
[15:13:27] <alex_mobile> No unix editor on my cellphone :(
[15:15:05] <cradek> // comment ^M if (condition)
[15:15:06] <cradek> {
[15:15:07] <cradek> ...
[15:15:33] <alex_mobile> Ouch
[15:15:39] <seb_kuzminsky> about a quarter of our coverity issues are in classicladder
[15:15:39] <cradek> yeah no kidding
[15:15:50] <seb_kuzminsky> (another quarter is in libnml)
[15:15:50] <cradek> seb_kuzminsky: hmm.
[15:16:30] <alex_mobile> So half in the actual emc2 code
[15:16:53] <seb_kuzminsky> lol @ arrays.c
[15:17:20] <seb_kuzminsky> see you guys later
[15:17:47] <alex_mobile> Later seb
[15:21:48] <alex_mobile> * alex_mobile needs a BT keyboard
[15:55:10] <jepler> oh god, I hope it turns out there's not any audio of RMS signing this tune:
[15:55:11] <jepler> http://web.cecs.pdx.edu/~trent/gnu/bull/04/bull4.html#SEC11
[15:55:53] <cradek> ack
[16:15:18] <cradek> AskFrame[ FrameSize++ ] = Value-((Value>>8)*256);
[16:38:15] <jepler> ??
[16:49:57] <skunkworks_> jepler: nice email
[16:50:39] <jepler> skunkworks_: thanks, I tried very hard to be civil and factual.
[16:50:41] <jepler> it becomes a struggle.
[16:50:51] <CIA-38> EMC: 03cradek 07TRUNK * 10emc2/src/hal/classicladder/arrays.c: remove ^M from this file to fix a very shady situation: // comment ^M if(condition) n { ...
[16:51:37] <CIA-38> EMC: 03cradek 07TRUNK * 10emc2/src/hal/classicladder/config_gtk.c: fix overruns due to undersized arrays
[16:52:13] <skunkworks_> jepler: you did good :)
[16:55:17] <CIA-38> EMC: 03cradek 07TRUNK * 10emc2/src/hal/classicladder/protocol_modbus_master.c: remove dos line endings
[17:02:01] <jepler> cradek: there's a lot more where that came from. http://emergent.unpy.net/files/sandbox/dos-contaminated.txt
[17:02:16] <jepler> (assuming my program's right, anyway)
[17:02:29] <cradek> I'm only fixing the ones I also want to make a substantive change to
[17:02:36] <jepler> ok
[17:02:48] <cradek> in fact, I'm done crapping around in classicladder
[17:02:54] <cradek> which makes me happy
[17:06:20] <CIA-38> EMC: 03cradek 07TRUNK * 10emc2/src/hal/utils/meter.c: this does not have to be global, because it's passed to refresh_value
[17:07:55] <CIA-38> EMC: 03cradek 07TRUNK * 10emc2/src/rtapi/sim_rtapi.c: this result covered the other result - no bug, just silly
[17:12:11] <cradek> interesting that CIA nuked the \ in my message
[17:12:30] <cradek> I figured it was my fault, but I don't think so, since it's in the email
[17:14:45] <BigJohnT> * BigJohnT kicks cia-38
[17:14:45] <CIA-38> ow
[17:14:53] <BigJohnT> maybe that fixed it
[17:15:06] <cradek> can't hurt.
[17:15:28] <jepler> sounds like it hurt
[17:26:07] <jepler> Did you mean: "mt dew" segmentation fault
[17:26:11] <jepler> gee thanks google
[17:26:13] <jepler> no, I didn't.
[17:27:24] <seb_kuzminsky> mt dew is implicated in 69.8% of my segfaults
[17:27:41] <jepler> in this case, I'm trying to figure out why 'mt rew' segfaults on this machine
[17:28:03] <cradek> it's still doing it?
[17:29:37] <jepler> cradek: yes
[17:29:45] <jepler> today I discovered that 'mt rewi' doesn't
[17:29:53] <cradek> uh
[17:30:11] <jepler> ikirst@nori:~$ ltrace mt rew
[17:30:15] <jepler> ...
[17:30:15] <jepler> strncmp("rewoffl", "rew", 3) = 0
[17:30:16] <jepler> strlen("rewoffl") = 7
[17:30:23] <jepler> ikirst@nori:~$ ltrace mt rewind
[17:30:35] <cradek> ohh
[17:30:37] <jepler> er, I guess I spelled it out all the way
[17:30:39] <cradek> different command
[17:30:48] <jepler> yeah, looks that way
[17:31:10] <jepler> actually, I *think* that it may be trying to produce the message to tell me I typed an ambiguous command..
[17:31:56] <cradek> how could mt have broken? is there a lot of development going on in there?
[17:32:11] <jepler> beats me
[17:32:11] <cradek> someone probably decided to i18nize it
[17:32:22] <jepler> setlocale(6, "") = "en_US.UTF-8"
[17:32:25] <jepler> bindtextdomain("cpio", "/usr/share/locale") = "/usr/share/locale"
[17:32:28] <jepler> textdomain("cpio") = "cpio"
[17:32:30] <jepler> of course
[17:33:01] <cradek> if only there was a way to tell what changed between dapster and harny
[17:33:42] <SWPadnos> cvs diff -rdapper -rhardy > bigfile
[17:33:46] <SWPadnos> :)
[19:48:04] <alex_joni> hi guys
[19:48:17] <jepler> afternoon alex_joni
[19:48:29] <alex_joni> what's up?
[19:54:08] <jepler> oh, nothing worth talking about
[19:54:40] <alex_joni> haha, read that
[19:55:13] <SWPadnos> that's never stopped us before
[19:55:53] <alex_joni> crap.. think I caught a cold
[19:56:38] <jepler> :(
[19:56:40] <jepler> don't do that
[19:56:54] <SWPadnos> yeah, toss it back
[20:06:49] <SWPadnos> jepler, do you remember why stepconf only outputs 0.95* (vel based on ideal_period) for max_vel?
[20:07:24] <jepler> SWPadnos: I think it was the easiest way to add stepgen headroom
[20:08:44] <SWPadnos> ok, looking at it, it seems that the ideal_period() function doesn't use the user-supplied velocity value
[20:08:54] <SWPadnos> oh wait, of course it does
[20:09:09] <SWPadnos> ok, so that's the confusion
[20:09:38] <SWPadnos> the guy on CNCZone who hasn't been able to make his G540 work has mentioned that the velocity is only 95% of what he entered
[20:09:54] <SWPadnos> which makes sense, since I see that calculation right there in stepconf :)
[20:11:17] <SWPadnos> but it's a bit confusing that his machine moves at 21.8IPM instead pf the 24 he entered. it seems that the hz value should be calculated using (1/0.95)*maxvel, so that the actual velocity limit will match user input if the machine can do it
[20:31:44] <jepler> like I said, I think it was the easiest way I saw to add stepgen headroom
[20:31:48] <jepler> I agree it's not ideal
[20:32:08] <SWPadnos> oh, it works but it's suprising for some users :)
[20:33:07] <SWPadnos> I could swear that I added code to account for the needed encoder counting rate when calculating min_period
[20:33:12] <SWPadnos> but I can't find it
[20:33:16] <jepler> hah!
[20:35:56] <SWPadnos> maybe I just thought about it, but someone convinced me it was too hard :)
[23:39:51] <seb_kuzminsky> http://www.linuxdevices.com/news/NS7876661673.html
[23:40:21] <seb_kuzminsky> commercial robot arm controlled by Linxu & Xenomai & open-source software
[23:41:27] <SWPadnos> cool. looks like 6 joints plus the grip (unless that's also a positionable servo joint)
[23:41:44] <SWPadnos> but it does mean that there's a working RT kernel on PPC
[23:41:47] <SWPadnos> which is nice
[23:58:26] <CIA-42> EMC: 03jepler 07TRUNK * 10emc2/src/Makefile: fix additional -O2 instances