Guest638 is now known as skunkworks
woo whoo - let the bidding begin :)
or woo hoo
how's that going?
one bid :)
at least you started it at $999, so you're sure to not get piddly squat
goes out sunday night - so this should be fun. Next best thing to gambling.
except you come out ahead instead of behind!
it's like gambling ... from the casinos point of view :)
[12:35:37] <skunkworks_> http://download.gna.org/tuxcnc/design_notes.pdf
He has an interesting take on the mesa card..
morning ray - how is it going?
doesn't anyone want to talk to me? ;)
heh - hi :)
rayh: have you seen this? http://download.gna.org/tuxcnc/design_notes.pdf
it is an interesting read
Now that I got it, do I want to read it.
I was talking to a guy that has been looking at tuxcnc.
He said there were some good ideas but the implementation wasn't there.
plus it's hard to read through the bullshit when you actually know about EMC2/HAL
I said earlier - (to my self) it seems inflexible.
I agree that there are some interesting ideas there though
it's super-flexible. you have the source code
or, you would if there is ever a release ...
cough - seem inflexible to someone like me..
I really like HAL at the hardware level.
For either rt or user paced stuff.
I just wish that we could build some of that flexibility into the rest of the EMC
rayh: did you jlmjv hooked up his encoders on his stepping machine and now is able to detect following error. That is flexible. :)
Without turning all of EMC into a giant patch panel.
Yes. Nifty. Now invert that signal and use it to control dynamic feedrate.
exactly - measure the following error and as it encreases - decrease the feedrate.
it would be a cool experiment anyways..
Surface finish goes to hell if it happens during a cut but for rapids and such it would be great.
I've thought about the "modularity without complexity" thing, and that's a hard problem :)
one way to achieve part of it would be to have some concept of "bundles" of signals - several related signals that get connected as a group
so motor velocity, enable and feedback all get connected as an "axis"
then that is different 'plug' if it is steppers vs servos.
Right. That was the idea of lego blocks.
well, acutally the connection between motion and the next level is the same
"Nothing should be committed until potential sources of memory leaks, segmentation faults, and other run time errors have been eliminated."
maybe this is why it's slow going...
35 instances of the word "will.'
unfortunately, source code is a potential source of memory leaks, segemntation faults, and other run-time errors
'tool tables should be declared in the preamble of a gcode file'
interesting that we were recently talking about this
(I still don't know if it's a good idea)
it's certainly more flexible that way, so it isn't strictly a bad idea
especially if you can both define a tool (as you suggested) and load a tool table (as I suggested)
I do that now with comments in the gcode.
build a table, save and load it.
It's a custom post processor for synergy.
so I assume you feel doing it in the gcode interp would be a good change
I'm all for flexible.
(I think I'm on that side of the fence)
The mega machines with tool hives only use a single tool file.
work boring for you too?
err - yeah, maybe
rayh: the ones I've seen had a robot to handle the tool hives
SWPadnos: not you, people doing actual work
oh, ok then
rayh: seen one with 4 hives & 4 robots on one DNC
they were building machines for rolling steel
Most of the shops I know have a tool file with the same name as the gcode file.
Wah! That would be something to watch.
Mazak in Kentucky runs 24/7 with only a single shift of humans.
I was amazed at how they test the software & panels & all without having the actual hardware
they set up simulations of all devices (motors, switches, etc), and hook those up to the real panels
the rest are monkeys?
the final machine gets only mounted at the final customer
don't tell anyone - the monkey union gets very annoyed
it's about 5-600m long
You know some of the original NIST stuff allowed for very similar testing.
ok, a bit bigger than a large printing press
SWPadnos: and waaaay heavier
one roll of finished sheet is about 10-20t
that's only about 2x the size of a paper roll for a large press :)
then I estimated it wrong :P
2x the weight, probably 1/4 the size ;)
it's human size in diameter
1.5m I think
ok, then you may be right. paper rolls can be ~20 feet wide
The paper mill I worked with built a 40' wide roll.
what's funny is that I only know about the printing industry because a photographer I used to work with shot the catalog stuff for Heidelberg-Harris
3000 feet a minute was max speed.
it's very bad to shoot 300 foot long printing presses with your flash on - the web break sensors don't like it :)
s/flash/bank of high power strobes placed all around/ :)
as one person said, "magazine printing is like painting on toilet paper with a squirt gun at 130 miles an hour"
That seems like a quote I should remember
heh. it's worked for me for about 10 years :)
I have sand http://www.electronicsam.com/images/house/garagesand.JPG
I have coffee
I am starting to not like black walnut trees.
nice wood from those (if you're allowed to chop them down)
yeh I actually got money for the one that used to be in the middle of where the garage is going to be.
suprisingly - they don't normally like then from in town
but it was bit.
Nice pic, skunk.
this is the front of the house http://www.electronicsam.com/images/house/newhehhouse.JPG
Hey it took 10 minutes to get the last one!
ouch - sorry
it is only 640X480 - come on now..
skunkworks_: the first time I looked at that URL I thought it said "new hen house"
woohoo! I finally have analog boards to test!
I thought, "I didn't know sam raised livestock"
new to me I guess. It was built sometime in the early 1900.. We are not sure yet.
SWPadnos: that was quick'
err - yeah :/
seems like you had just drawn the circuit up.. or is this a different thing.
it's been closer to 2-week turn than the 3-day turn I asked for
Our's was rebuilt after a fire. 1939. They used newspapers to stuff the space between framing and door frames.
or have a lost track of some time..
lost ttrack of time, I'm afraid
gotta go.. out of battery
the house needs a little insulation yet.. and new fangled 'double pane' windows
see you tomorrow guys
see you alex.
see you Alex
I sure wish emc would spit out nothing but the hal error in cases like ljm's
those are buried in so much crap that they're invisible
I guess the real question is what to do with the other messages
I wonder if they're always noise, or sometimes useful
heh - ideally, the "useful" mesasge would be printed and the others suppressed :)
I wonder about using color (like dircolors)
or bold - that should even be available on a monochrome terminal
hm I'm still waiting for you to say "just kidding"
oh. just kidding
Bold is good!
*bold* is good :)
all that is needed is for someone to identify the useless messages and remove the lines that print them
I'm not sure the useless messages are always useless though
I'd say that most of the ones from the run script aren't needed any more
Guest539 is now known as skunkworks_
cradek: first thing I notice is that there are at least 5 static msg_levels
I got it to act more like I expect... but yeah, there's a bunch of stuff I don't quite get
I think some of my tree wasn't building, it shaped up after I cleaned
err, no. I still don't understand.
are you on sim or rtai?
I figured I'd try to tackle it first
extra messages from sim aren't as much of a prblem.
so you changed rtai_rtapi.c and rtai_ulapi.c
-static int msg_level = RTAPI_MSG_INFO;/* message printing level */
+static int msg_level = RTAPI_MSG_WARN;/* message printing level */
something like this
oh fascinating, rtapi_proc.h has a bunch of code in it
it gets included
some of it messes with msg_level
[14194.584326] new msg_level from proc: 3
scripts/realtime: if [ "$DEBUG" != "" ] && [ -w /proc/rtapi/debug ] ; then
scripts/realtime: echo "$DEBUG" > /proc/rtapi/debug
# Sets the default DEBUG level
* cradek grumbles
yuck, I can't change the HAL:123:ERROR: except by changing every damn line
maybe that could have been in a function or something.
"HAL:%d: ERROR: pin '%s' not found\n", linenumber, first_pin_name);
there are lots of variations
I think I'd like something like ERROR IN HAL FILE \"%s\" ON LINE %d:
*nobody* recognizes HAL:123: as a line number, and you still don't know which file to check if you do recognize that
I think one wrinkle is that a filename isn't available there
I don't remember exactly *why* that was the case..
strange I get the error on stdout, I thought it went to dmesg
HAL: ERROR: pin 'hypot.0.in9' not found
HAL:65: link failed
HAL config file /home/chris/emc2.trunk/configs/sim/core_sim.hal failed.
Shutting down and cleaning up EMC2...
userspace stuff goes to stdout or stderr, kernel stuff goes to dmesg
this is actually pretty good
ok this is better now...
HAL: ERROR: function 'ddt.9' not found
core_sim.hal:20: addf failed
oh did you do something better after I checked in my crap?
that's great if so.
well .. during
hopefully I didn't pork anything up too bad..
core_sim:9: addf failed <- 9 is the line number right?
no way to say that?
filename:number:message is a standard unix way of reporting errors
yes all programmers will understand that format
and programmers are the target audience of emc after all
:) I guess normal people will figure it out after a bit..
jmkasunich__ is now known as jmkasunich