cradek: Running your little func. here - No NaN to report.
jepler cradek: The NaN bug would have cropped up when any usr space app did FP calculations.
she didnt stay long
Time: 22:05:02 -0500 GMT, Windows XP Professional, Service Pack 2 (5.1 - 2600), AMD XP 2800+ running at (1-AMD , 2123MHz, 512KB (0% Load)), DDR400 RAM Usage: 555/1024MB (54.20%), C: 43.97gb of 75.42gb free, D: 0gb of 0gb free, N: 1.67gb of 372.62gb free, Current Uptime: 6hrs 32mins 9secs, Record Uptime: 3wks 12hrs 48mins 50secs, 3 Samsung 19" flat monitors (1 Trinitron, 2 CRT).
# lpg - axis (emc display) likes glx loaded and hates dri loaded
# This loads the GLX module
# This loads the DRI module
# Load "dri"
* LawrenceG going to bed... goodnight all
let there be speaking, at some point in time
using video ram as swap: http://hedera.linuxnews.pl/_news/2002/09/03/_long/1445.html
whats up anon
working on this drawing
a slide throttle body..
a car :)
like they use in racing..
but this design is a bit special
I have the cam generated in a bit
so I am going to machine a prototype of it later tonight
right on... ; )
I've yet to decide on some stuff related to it
what are you going to try it on.. your car??
I am building a new engine soon..
ahh i see
slide/roller throttle bodies are like 1000$.. ;)
[15:07:37] <anonimasu> http://motorsport.bdg.com.au/roller.html
going to modify the design later when I have access to cfd software :)
might be a bit overkill though
think it through.. there's possibly a good reason for them to be 1000$
pre machine of alloy steel components... heat treat.. grinding.. possibly lapping
fuel injection components are not easy parts to mfg
that's great news about fp on bdi being fixed
twas a compare with a stupid bloody float...
I didn't see a checkin go by ... are you saying it was a simple bug in emc?
I haven't got as far as doing a commit.
I'm anxious to see what it was. Does it need to be fixed in several trees?
but, yes. It was just a couple of lines that were probably the cause.
it was good to get some more data points from Oscar Dalem
I hadn't seen him around before
It won't affect the EMC1 tree as that has FP enabled in all RT threads.
ok, that's good.
and emc2/cvs works correctly
If something of a similar nature exists in the emc2 head, then it will be something for JMK to fix.
at least, not that we could find
we tried last night
there are other bugs in HEAD that JMK needs to fix.
oh I'm sure there are bugs in all the trees!
but I think this is exciting - people will finally be able to use axis without jumping through flaming hoops
now I hope *we* don't have any important bugs left
The next BDI build will use the cvs snapshot from the 16th
I will do some testing now
made yet another (minor) change to setup.py so that a deb-packaging warning disappeared.
someone reported a problem with offset handling
ok, just send one of us a patch if you want it committed
unfortunately I can't test emc-bdi4 here, only emc1 and emc2
Much of the user space & build system is identical to that of emc2
despite JMK's claims to the contrary.
do you have sim on bdi now?
(I have a new hard drive now and I may have room to install bdi soon)
(it sucks to not be able to effectively test)
well, I will try emc2 for now then
paul_c: which python version does axis use on bdi?
darn, I have 2.2 and 2.4
I will use 2.2
and the 20050916 snapshot
/home/chris/emc2/rtlib/motmod.o: unresolved symbol emcmotLogAdd
/home/chris/emc2/rtlib/motmod.o: unresolved symbol emcmotLogInit
is this a known problem (emc2 head)?
got your patch
03paul_c * 10emc2/src/emc/motion/command.c: halscope makes emcmotlog redundant - ..
that got rid of one of my errors
/home/chris/emc2/rtlib/motmod.o: unresolved symbol emcmotLogAdd
still have this one
03paul_c * 10emc2/src/emc/motion/command.c: Another ToDo for someone to take a look at - Are the emcmotlog functions still required ?
anytihng new or exciting happen today?
just the stuff on the mailing list
oh, I learnt how to spot-weld with a mig torch ...
jeff and I found (and pual fixed) an important bug in bdi
oh, I dont get the mailing list anymore, perhaps I should view the archive
cradek: That's with a 'P'
the NaN bug?
paul_c: according to my irc client, it's lowercase-p
paul_c is now known as Paul_C
* cradek has to research how to type those big letters
cradek: I presume this was related to how undefined values being hung on an NML message?
robin_sz: nope, some rt code was porking up the fp stack
had sodall to do with NML
that could have had alsorts of nasty consequences I guess
it did... all userland was affected
showed up worst when using axis (lots of fp)
but at least one other person reported fp-heavy software crashing when emc was started
so it may improve stability of the whole thing then
and jeff's test program was an excellent way to show this behavior
yes, it's a very important fix
kewl. thats two important fixes this year then
the long-pursued trajectory planner bug being the other I guess
[19:24:09] <cradek> http://cvs.sourceforge.net/viewcvs.py/emc/emc/src/rs274ngc_new/rs274ngc_pre.cc?r1=1.10&r2=184.108.40.206&only_with_tag=cradek_stable
the next can of worms will be Les and his toolchanger .. where the little used EMC1 toolchanger code gets a thrashing ... shold be interesting, as JMK will be thrashing out the EMC2 toolchanger code around the same time on his Mazak refit
Paul_C: you have this change, right? I am trying to figure out how to put it in emc2
cradek: Yes. That will move in to emc2 HEAD on the next merge.
oh so I don't have to put it in there now?
nah - I also have a couple of other minor bug fixes in the interp to merge in to head.
ok, I'm hacking it in for the sake of testing, but I won't commit it
No you need the altered file ?
toolchange will be an interesting bit of the code, theres quite a lot of "stubs" there waiting to hang real code on
if you want me to test it, I will, but I hacked that change into mine already
darn, emc2 still has a bug with the offsets in the stat buffer not being updated
I bet bdi has this bug too
err, I bet emc-bdi4 has this bug too
it's not necessarily a show-stopper - when you load a file, it works right
in other news .. I got my forst little servo amp this week
cradek: Tool offsets ?
g10 l2 ...
seems to have more adjustments than I expected ...
axis uses these
tool offsets are read in by the tool module...
sorry, I'm talking about coord system offsets
Did the cradek_stable branch get a fix for that ?
no, it has always worked in emc1
can you test it for me in emc-bdi4
start emc with axis
jog over a bit in X
push "zero offset"
you should see the axes move to the cone (relative origin), and a cyan origin symbol appears at the absolute origin
just need to apt-get axis.
Paul_C: jepler reports that he gets joint errors and following errors running 3D_Chips.ngc with the sherline ini on bdi
Sherline does not have "joints"
well, even with that in mind, jepler still reports he gets joint errors!
this is on the same bdi 4.25+updates machine by the way
unfortunately I can't try it here.
uh, how do I even load a file in mini?
(turning axis off in the config, even though that seems an unlikely cause of a following error)
try pushing those check boxes on the top
open 3d_chips and run
mini says "axis 0 following error"
is it just with the mm ini?
red bar points at line N80
I was using the inch setup before. I have axis in the inch .ini and mini in the mm .ini so I can easily switch between them
didn't get it that time on the inch config
just forget I said anything
wonder if this is a case of the scale values being too small in mm mode
Oh FFS... PID is getting trashed.
Paul_C: what do you mean? in what case?
cradek: you put the fast run-from-line support in emc1/cradek_stable yet?
I'm getting several parameters on AXIS_0 being reset.
I just did a 'run from line' on 3D_Chips in axis on bdi and it was fast
jepler: I don't remember - I don't think so
well you should
can you point me to the right diff?
"my" axis does start with a somewhat larger window than the bdi one
so does mine
it's probably just the kde window manager or something
no, that can't be it -- I've been doing all this testing doing a remote display to a nice sane icewm-managed desktop
well paul said he's using last night's snapshot of axis for this next release
I built that snapshot and it's fine
the only problem I see with emc2/head and last night's axis is this offset thing when no file is loaded.
I don't see where we set a larger-than-minimum size for that window
well I don't care too much - the toolbar does fit
ah, wait -- axis.nf has a 'wm minsize' line, and that seems to correspond to the geometry the window actually takes
ok, the offsets are written into the .var correctly, but the stat buffer isn't updated
right, that's what was happening with emc1+simulator
oh really? emc1?
yeah, what you're describing is what happens on my machine
when I load a file, the buffer is updated
we could make axis load an empty file when it starts. ugh.
yep, same here
but that'll mess up things like the state of the 'run' and 'reload' buttons...
yeah, we really don't want that
other than stat being wrong, it does work right (mdi g0 and g0g53 do the right things)
I wonder if I can figure out how to use tkemc well enough to check this
when I mdi g10 l2 p1 x... in tkemc, displaying relative coords, it changes
it is using emcStatus->taks.origin.tran just like we are (emcsh.cc:2702)
oh god. to set an offset, tkemc writes out the var file and then tells the interpreter to read it.
worse, it does it with TASK_PLAN_INIT which "unloads" your file
sometimes the more I work on emc, the worse it smells
Can you change the offsets by instead MDI'ing a #nnnn=oooo ?
(not that I think that would be a good way to do it)
yeah, but that makes me ill
but only a little
I'll try it
[20:41:01] <jepler> http://www.linuxcnc.org/handbook/RS274NGC_3/RS274NGC_33a.html#1008243
no, that still works the same way
well, until someone (jmk?) can help me find the problem in the interpreter, this problem isn't going to get fixed
I refuse to set offsets the tkemc way - I think it sucks.
damit... My earlier fix will piss off JE again - It breaks his damned drivers.
Paul_C: if we were closer I'd say we should go get a drink.
You're out in the mid-west ?
be there in, say... nine hours ?
you're in the US today?
Still in England.
bring some Newcastle with you then
you can get that stuff from Wall Mart
ok then, bring plenty
Fursty Ferret be OK ?
no idea about that
bring whatever you like.
are there any facilities in EMC to allow you to build external button panels easily?
I'm designing a stepper control board now that I'd like to use with EMC, and would like to know what facilities there are internally so I can naturally integrate against that
icee: nothing's easy in emc.
You can use Tk, Python or Qt
cradek: I kinda got that feeling ;)
icee: sorry, you probably caught me on a bad day.
icee: so I retract that statement
you know tho, i want like start and stop and jog buttons without having to use the UI
do you know of anyone who's done that before, or am I going to have to ghetto it up myself?
icee: the python emc interface module that axis uses is very clean and easy to use.
icee: what are the buttons? do they look like a keyboard to X?
icee: if so, it's a simple matter of key bindings
hmm. I suppose I could make them look like a keyboard to X
i could put a USB interface on my board
icee: jepler wrote a linux-joystick to python-emc-module program for jogging
icee: it all depends on your hardware
cradek: do you know where that's available? I'd prefer to put the button presses back over RS-232, all being equal
but I could make it look like a keyboard, if it's substantially easier to integrate
[21:06:04] <cradek> http://craie.unpy.net/aether/index.cgi/software/01119021973
if self.s.task_state == emc.STATE_ESTOP:
usb is much better for nearly any new hardware design than rs232.
neat! looks very easy
yeah the interface is very easy to use
the only hairy part in this project is the joystick interface
cradek: yah, but writing code to cause your USB thingy to enumerate as HID, and then marshall packets, etc, is substantially more of a pain than just shuffling characters over rs-232
but, imagine all the whining from your users about not having a serial port available
oh, this is purely for my own use
oh, forget it then
i may put the schematic, gerbers, and BOM up on the web, but otherwise..
I have a few old serial numpads around somewhere
it would be a simple matter of programming to use one of those.
yah, just an infinite loop and dispatching EMC events
i was hoping to find something like this
thanks :) it'll make things really easy
cradek: Want to try and work through the offset prob then ?
Paul_C: I tried to find it, but failed
Paul_C: I would like to know if you have the problem on emc-bdi4 I guess
Paul_C: I bet you do
Let's start from the top - What var are you reading ?
none, I'm reading the stat buffer
and I change the offset with g10 l2 p1 or whatever it is
one sec while I build axis
in emc1, the stat buffer updates - in emc2, it updates but only if you have previously loaded a file
G10 L2 P1 X1 Y1 Z0
yeah that should do it
but easier is to use shift-home (or push the "zero offset" button)
Origin jumps, and the display shows an offset applied
before you load a file?
do it with me: start emc, machine on, jog in some axis, hit shift-home
no file loaded.
origin should jump
maybe this is only an emc2 problem then
one sec - shutting down & reloading the var
right arrow for a while
does the origin move to the cone?
now load a file
only then will it jump to the cone
that's the bug
zero offset - also zip.
the offset IS set
try g0x0 and g0g53x0 for instance
load a file, and offset is there...
it's just not reflected in the stat buffer, so the origin doesn't move
so you have it too in emc-bdi4
now that a file is loaded, jog and shift-home
you will see the origin (and program) jump
this is how it's supposed to work
So status is only being updated with offsets once a file is loaded.
HERE is another interesting point
emc1 sim DOES have this bug
emc1 non-sim does NOT (only version that works right)
what NML var are you reading for the offset ?
what NML command do you have the zero button bound to ?
it sends that G10 L2 P1 command via MDI
the interp isn't processing the MDI line...
Hrmmm.... Back in five.
the offset DOES get set
after you set it, run g0x0 and g0g53x0 and you will see the offset works
I have to go for a while - be back in a few hours
I see what the prob is, and where it needs fixing...
emc hasn't given me any serious problems since i got it working for me; i mention this as I have been able to trivially segfault the latest version of Varkon
(i.e. kudos to the EMC developers and maintainers ;) )
Don't worry - When EMC segfaults, boy, you know about it....
The system crashes out BIG time !
heh, that's reassuring :)
* zwelch just can't wait to get a mill to hook up to EMC after hearing that ;)
You ever seen a null pointer in kernel space ?
* zwelch has contributed kernel code
written a couple o' device drivers, etc.
It's not so bad with 2.6 kernels...
it's still ugly ;)
i.e. the effects are usually highly undesirable
with realtime, there's a bunch of other gotchas waiting to kick your butt
rt_task_create with zero period...
* zwelch didn't want to install RT just to try EMC for that very reason
i'll use dedicated boxen for RT systems
There is the BDI-Live disk if you just want to take EMC for a spin.
is that the same as the BDI iso i've downloaded and burned?
i was going to pop that in and look at it, but the only system i can afford for that at the moment is PPC ;)
EMC hasn't been compiled to run on ppc yet.
i could port it, if there's any interest
as for my needs, i'll set up one of my spare x86 sometime this weekend
If there is an RT patch, it should be fairly trivial - There are no x86 specific code.
wonders if emc would complie for this http://www.embeddedx86.com/epc/ts7200-spec-p.php
AchiestDragon: I think it's a bit slow
Paul_C: is EMC being run on other big endian platforms than PPC?
200MHz is faster than the original platforms EMC as compile for.
zwelch: I don't know of any non-x86 platforms running EMC
the only concern there would be size constraints, which are irrelevant if you are willing to use something like a 1GB CF card :)
(and of course, the issues of getting the ARM RT kernel bits working)
but the NML library is used on a number of other architectures.
AchiestDragon: You get me one of those boards, and I'll have a little play.
does "the NML library" == "the NIST librcs"?
was just thinking that i could build on of those into the controls then just feed it the gcode file by ethernet
I'd want a 16bit PC104 connector on there.
it says it has a PC104 bus
8bit standard, not 16bit
therea a onboard i/o port on there 20bits so should be able to wire the controls directly from that
Paul_C: actually, the hardware says 8/16 bit
erm... hardware page
this x86 (geode) board has a dozen TTL I/Os. http://www.soekris.com/net4801.htm
It's $213 qty 1 though
I've looked at those ARM boards in the past...
there are a ton of options :)
cradek: Bug FIXED !
* Paul_C tests again just to be sure....
almost finished my cnc http://www.whipy.demon.co.uk/