Back
[02:59:34] <micges> jepler: it's a bug, there is only one cards index in driver, should be one per type like in hm2_pci
[06:52:55] <skunkworks> zlog,
[08:09:33] <skunkworks> runing the bug424.ngc - arc_blend_kink_ratio set to .01
http://electronicsam.com/images/KandT/testing/kink.01.png
[08:09:52] <skunkworks> and set to .7
http://electronicsam.com/images/KandT/testing/kink.7.png
[08:10:03] <skunkworks> 22in/sec^2
[08:11:04] <skunkworks> I have a question as to why it doesn't just run at full speed - it isn't using all the acceleration. Rob may be limiting jerk a bit - I don't know.
[08:11:46] <skunkworks> If you raise the axis acceleration - it does run without slowing down.
[08:31:11] <jepler> running overnight with 3 cards at 2kHz I got a spew of these
[08:31:12] <jepler> hm2_eth: enqueue_read ERROR: reading packet: recv() -> -1 Resource temporarily unavailable (expected to read 120 bytes)
[08:31:27] <jepler> and the thread max-time is now 1895590945
[08:31:35] <skunkworks> yeck
[08:31:49] <jepler> ugh that's nearly 2 seconds!
[08:32:05] <skunkworks> lose power to the switch?
[08:32:41] <skunkworks> or the switch overheated and rebooted?
[08:32:47] <jepler> don't think so
[08:32:51] <skunkworks> darn
[08:32:57] <jepler> hard to be sure though
[08:33:01] <jepler> dmesg has grumpy stuff in it too
[08:33:01] <jepler> [ 1885.272010] INFO: rcu_preempt self-detected stall on CPU { 1} (t=5250 jiffies g=20036 c=20035 q=738)
[08:33:12] <archivist> packets piling up?
[08:34:02] <jepler> something like that
[08:34:26] <jepler> this is still with the 10us delay removed from hm2_eth, now replaced by a sched_yield call
[08:37:34] <jepler> http://paste.debian.net/282268/
[08:37:52] <jepler> it seems weird that recvmsg calls sendmsg
[08:37:53] <jepler> [ 1885.272088] [<ffffffff8148819e>] ? udp_sendmsg+0x3ce/0x890
[08:37:53] <jepler> [ 1885.272090] [<ffffffff81488c8a>] ? udp_recvmsg+0x9a/0x450
[08:38:48] <jepler> though I am told that "?" in a kernel traceback means it is a code address seen in the stack, but not proven to be a return address by unwinding the stack
[08:40:46] <lair82> Good Morning Guys, I am having problems on my machine, last week we started noticing that when we would try to load a program, there would be nothing in the nc_files folder, when there are over 100 programs. I would shut Linuxcnc down, and there would be zero icons on the desktop, not my shortcut link start linuxcnc, none of the folders that would normally be there, nothing. I would manually shut the pc down, with the pushbutto
[08:41:24] <lair82> about 10 seconds, and it wouldn't boot linux, it would just go to the bios for the MB.
[08:53:47] <lair82> I thought the HD took a poop, so I installed a fresh ubuntu build on a new HD, re built my build, and now today it is showing similar symptoms, any thoughts?
[08:59:56] <seb_kuzminsky> skunkworks: thanks for testing
[09:00:10] <seb_kuzminsky> those curves look nice, except for the accel blips
[09:00:19] <seb_kuzminsky> did you mail rob?
[09:06:12] <skunkworks> seb_kuzminsky, no violations
[09:06:26] <seb_kuzminsky> just those little wiggles
[09:09:41] <CaptHindsight> lair82: I'd ask in #linuxcnc but maybe a bad MB, are you using PASSWORD for password and accessing facebook with it?
[09:12:34] <skunkworks> those accel blips are not violations
[09:13:43] <seb_kuzminsky> ok good
[09:15:10] <seb_kuzminsky> lair82: try running memtest86 on it (it's available from the grub boot menu)
[09:15:35] <lair82> seb_kuzminsky ok
[09:53:32] <jepler> Subject: [ANNOUNCE] 4.1.2-rt1
[09:55:31] <skunkworks> oh - does that help your dragon board?
[09:55:49] <pcw_home> Wow
[10:01:16] <pcw_home> for reference, with hm2_eth running at 3 KHz on my Core Duo (E8500 3.16 GHz)
[10:01:18] <pcw_home> the Ethernet interrupt uses about 3.5 % of a CPU and there's no noticeable slowdown
[10:01:19] <pcw_home> of anything else
[10:03:09] <jepler> skunkworks: not right away. I still haven't successfully built and booted a non-rt kernel and I haven't had an "aha" moment yet about why not
[10:03:48] <jepler> pcw_home: 18.3% CPU here though that's with sched_yield instead of rtapi_delay when waiting for read packets
[10:07:24] <jepler> 13% with the delay put back in, plus ~3% for rx and tx irqs
[10:09:04] <jepler> coreboot: ... It became clear that a new release is overdue. With our new release process only slowly getting in shape, I decided to take a random commit and call it ‘4.1’. ...
[10:09:08] <jepler> well that's one way to do ti
[10:09:10] <jepler> it
[10:57:55] <skunkworks> there is an issue with robs branch. Gcode is running about half as fast as master in some situations. He is looking at it
[11:03:51] <lair82___> seb_kuzminsky , how do I get memtest going if the grub menu does not come up during boot?
[11:08:13] <skunkworks> boot off the live image?
[11:12:23] <skunkworks> tapping the shift key should send you into grub if you hit it just right (right shift key)
[11:13:37] <mozmck> I haven't tried that, but in ubuntu the default timeout is 0, so you don't much time if that's going to work!
[11:16:36] <lair82___> Ok, its running, fixed grub, running memtest
[11:28:39] <seb_kuzminsky> you can hold down shift, you don't have to tap & time it
[11:30:29] <skunkworks> On a lot of my systems - holding didn't seem to work.
[11:33:41] <seb_kuzminsky> hmm, weird
[11:36:08] <micges> here works only tapping left shift key
[11:41:42] <seb_kuzminsky> maybe it depends on the bios? or maybe i'm misremembering
[11:43:10] <kwallace> Holding left shift starting with the BIOS screen works for me. I think shift is used because it doesn't repeat, but that's a guess.
[11:44:29] <seb_kuzminsky> my brother in madness!
[11:50:25] <skunkworks> you guys are crazy!! ;)
[11:50:51] <skunkworks> maybe usb vs ps2?
[11:51:39] <seb_kuzminsky> mmmmaybe
[13:07:13] <mozmck> anyone know offhand how halshow shows pin states without connecting with signals?
[13:09:54] <mozmck> I'd like to do something similar for a diagnostics panel in my GUI, just for the 7i92 inputs/outputs. In gladevcp I have to connect the widget hal pin to the signal the 7i92 pin is connected to, but I want to show the pin state even if it is not connected to anything.
[13:11:54] <micges> it access hal data directly
[13:12:11] <mozmck> Hmm, so I wonder if I can do that in python?
[13:18:18] <micges> atm no
[13:19:06] <lair82___> Pass 1, no errors
[13:19:16] <micges> mozmck: look at src/hal/meter.c function popup_probe_window()
[13:19:38] <mozmck> micges: thanks, will do.
[13:20:54] <micges> it's code that enumerates all pin/signals/params, you need to add this to halmodule.cc in some way, then you can use it from python
[13:20:57] <bjmorel_work> Does anyone have a link to a getting started guide for a Mesa 7i92 / 7i76 setup ?
[13:21:43] <micges> bjmorel_work: first: man hm2_eth
[13:22:08] <skunkworks> mozmck,
http://www.machsupport.com/forum/index.php/topic,30471.msg211591.html#msg211591
[13:24:04] <micges> bjmorel_work: besides loadrt hm2_eth line rest of hal is the same as 5i25+7i76 setup
[13:24:36] <bjmorel_work> micges: Perfect, thanks. I do need 2.7 correct?
[13:25:00] <mozmck> skunkworks: thanks. I don't keep up with all the forums and lists - takes too much time!
[13:25:00] <micges> yes
[13:26:40] <bjmorel_work> Also, does it matter if I use an RT-Preempt kernel or the default RT kernel that ships on the live cd ?
[13:26:41] <mozmck> skunkworks: It actually works pretty well, but the M10/M11 issue he mentioned is a bit of a problem because of time delays.
[13:26:59] <mozmck> bjmorel_work: you have to use RT-Preempt.
[13:29:56] <skunkworks> mozmck, is it nice being a bit more realtimey?
[13:30:02] <bjmorel_work> Ok, i can get that all set up tonight, cards should be in tomorrow.
[13:30:09] <bjmorel_work> Thanks again.
[13:30:45] <mozmck> skunkworks: quite nice.
[13:30:53] <skunkworks> :)
[16:21:53] <KGB-linuxcnc> 03Sebastian Kuzminsky 052.7 5f9480d 06linuxcnc 10docs/man/man3/rtapi_print.3rtapi docs: ficks a typoe in rtapi_print(3) manpage * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=5f9480d
[18:19:37] <skunkworks> seb_kuzminsky: funny :)
[18:57:36] <Gurwinder_> Hi, I am Gurwinder. Now for GSoC Post mid term, I am going work on linuxcnc.
[18:59:34] <Gurwinder_> In past I with Chris morley have discussed on many required work on linuxcnc but due to lack of knowledge I was not able to work on those projects.
[19:00:10] <Gurwinder_> But Morley and I was working on stepconf help button.
[19:01:25] <Gurwinder_> For it I have to learn python. So, learnt all the basics of pyhthon and now able to understand whats going on in python file.
[19:02:43] <Gurwinder_> So, want to discuss about it for GSoC
[20:11:14] <Gurwinder_> anyone please?
[20:12:49] <cradek> Gurwinder_: I am probably not the right person to help you, but either way, I don't understand what your question is. If you have a question, please ask it clearly, and if someone can help you after reading it, they will.
[20:15:07] <Gurwinder_> +cradek: I want to discuss on Google Summer of Code. So that I can work on it.
[21:14:33] <jepler> > [100219.905145] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
[21:14:48] <jepler> [100219.905165] IP: [<ffffffffa04808f6>] nouveau_fence_signal+0x66/0xd0 [nouveau]
[21:15:01] <cradek> oof
[21:15:09] <jepler> and a RT delay to go with it
[21:15:33] <jepler> but all 3 mesa boards attached had their watchdog correctly bite, so that's good
[21:15:46] <cradek> yay
[21:15:53] <cradek> was that the three seconds?
[21:16:21] <cradek> or was it 2 - that you reported earlier
[21:17:31] <jepler> no the earlier problems were "self-detected stall"s, something else
[21:18:43] <jepler> I wish I had a different video choice in this box; seems like the intel chipset video is decent for rt these days
[21:19:08] <cradek> does it have a slot?
[21:19:16] <jepler> yes, I could take the nvidia out
[21:19:50] <jepler> cradek: this is the one you saw the console of when you were over recently, it has the trashed login screen, that's how well nouveau works..
[21:19:59] <jepler> I didn't even try with the proprietary drivers
[21:20:04] <cradek> yeah, that was lovely
[21:20:19] <cradek> it's a strange kind of screwup
[21:20:33] <jepler> almost of those ex-office nvidia cards have that behavior on debian wheezy with nouveau
[21:20:51] <jepler> very similar trashed login screens, and generally flaky until you install the proprietary drivers
[21:21:12] <cradek> blurgh
[21:21:25] <cradek> doctor doctor, it hurts when I buy nvidia hardware
[21:21:34] <jepler> somebody gave me the box for free, who am I to complain?
[21:22:04] <cradek> acetone and whiskey smell too similar for comfort, having them both right here
[21:22:24] <jepler> "eking out dramatic performance gains"
