andypugh: I never heard of them before, thanks. I don't like them already (just from their web page) because they use those hidden links to their screenshots. Why not have it so you can right-click the screenshot "open in new tab"? Annoyed me enough to tease out one of their obfuscated links: http://www.avl.com/wo/webobsession.servlet.go/encoded/YXBwPWJjbXMmcGFnZT12aWV3Jm1hc2s9ZG93bmxvYWQmbm9kZWlkPTQwMDA2MzYwMSZub2VuY29k_0AZT1KUEcmY2FjaGVidXN0ZXI9.jpg127
[00:14:14] <KimK> http://www.avl.com/wo/webobsession.servlet.go/encoded/YXBwPWJjbXMmcGFnZT12aWV3Jm1hc2s9ZG93bmxvYWQmbm9kZWlkPTQwMDA2MzYwMSZub2VuY29k_0AZT1KUEcmY2FjaGVidXN0ZXI9.jpg1278115037469
I'm sure the DRM police will be stopping by any moment now. BTW, andy, what's "alpha" (gauge lower left of photo center)
Alpha is what we call the pedal position.
Ah, OK, thanks.
Basically there are only two controls, dyno current and pedal position, but you can use them to control to different things.
ie you can close the dyno loop on dyno speed or dyno torque and the pedal loop on alpha, indicated torque, fuel pressure, power......
ries_ is now known as ries
EMC: 03jmelson 07master * r06a279c50077 10/src/hal/drivers/hal_ppmc.c: make error message more informative
EMC: 03jmelson 07master * ref766804c070 10/configs/ppmc/ppmc.ini: use Axis GUI
EMC: 03jmelson 07master * r444a3e07bd0b 10/configs/univpwm/univpwm_io.hal: update for pendant
EMC: 03jmelson 07master * r31f9b8b8584c 10/configs/univstep/univstep.ini: update to Axis GUI
EMC: 03jmelson 07master * ra49f54ba2f93 10/configs/univpwm/univpwm.ini: update to Axis GUI and add some uptions
EMC: 03jmelson 07master * rb20474ff129d 10/configs/univpwm/pendant.hal: add pendant file
EMC: 03jmelson 07master * r180c5ce12e10 10/configs/univpwm/spindle.hal: add spindle speed hal file
EMC: 03jmelson 07master * rd5a61e7d2ed8 10/configs/univpwm/spindle.xml: add spindle display xml file
jepler- is now known as jepler
I think G76 might be broken in Master.
I am getting some good passes (2 or 3) then a few full-speed passes, then a couple of good ones...
Looking in Halscope the return to sensible passes coincides with an encoder reset / index pulse.
The same Halscope, same G-code, same material has just done it right first time in 2.4.1. In that Halscope you can see the index-enable followed by an encoder reset at the beginning of every pass. That behaviour was not seen in 2.5~pre
what is your encoder hooked to?
A hostmot counter
can you scope index-enable and encoder counts and encoder position
I've seen this behavior when ppmc encoders were broken and would reset index-enable without resetting their count/position
OK, need to switch computers.
I cut several threads on 2.4.1/hostmot2 at cnc workshop
I didn't see the problem - not sure what firmware I was using - I didn't check
How o I get a screen cap/
Please forgive typing ghastly metal eybord
or alt-printscreen for just the current window
any other way?
it's on the menu
thanks. ghastly swarf-proo kb has no prntsc
bbl after shower/breakfast (is it morning already?)
All I have done between the two is change emc version
it looks like index-enable is being driven differently. It could be an illusion
that 2.5 trace is AFU
please test at the tip of v2.4_branch; if the problem is there as well I will hold 2.4.2 until this is sorted out.
[15:36:20] <cradek> http://imagebin.ca/view/B9LChyu.html
looks right to me
[15:36:42] <cradek> http://imagebin.ca/view/oxXHjuf.html
what is gpio.004.in?
please verify which version is the Screenshot -1
screenshot 1 is a fresh git pull of the 2.4.1 branch
on Screenshot -1, I see above the 10 button a count reset without index, but above 11 an index without count reset
gpio-4 is the index pulse into the 7i43
are you using the same config and same firmwares?
Hold on. I want to be sure what version of emc we're talking about. There's no such thing as "the 2.4.1 branch". Is it v2.4.1, or is it v2.4_branch? What does "git describe" print in the source directory for the thing you're saying is "2.4"?
yes, all else (supposedly) the exact same
let me check
not working is v2.4.1-358-g6d6b94a
working is v2.4.1
commit 6d6b94a is not one from linuxcnc.org. Do you deliberately have some of your own modifications there?
including a revert of the last G76 patch
can you find a linuxcnc commit that shows the problem
I haven't tried.
if you have no uncommitted changes, then 'git checkout origin/master' to get source that exactly matches linuxcnc. 'git checkout master' to get back to the master branch which is presumably where you have your committed changes.
Can I finish this component first? switching versions mid-part seems rash
I am puzzled by: "screenshot 1 is a fresh git pull of the 2.4.1 branch" vs. "working is v2.4.1"
I have no idea what versions you are comparing and what code you are running that the shows bad behavior
That is because I am puzzled by alost everything to do with version control
sure, when you have time to nail it down, let us know
the first screenshot was a freshly downloaded and compiled 2.4.1
that worked fine made a nice thread
first is this one? http://imagebin.ca/view/B9LChyu.html
I agree it looks right
then the confusion is that the page >>
[15:52:17] <jepler> http://imagebin.ca/view/oxXHjuf.html
says "Screenshot-HAL Oscilloscope-1.png"
right we need to figure out what code causes http://imagebin.ca/view/oxXHjuf.html
The second screenshot was whatever state I left my dev version in last time I got bored
meanwhile I'll test sim/lathe on the tip of origin/master and origin/v2.4_branch to see if I can observe a problem (but I'm sort of suspecting that whatever's going on is in the mesa driver)
Ah. I hadn't even looked at the screenshot file names
andypugh_garage: you said you reverted a g76 patch. I also don't know what that means.
You made left-hand threads work since the last time I did any threading
So I reverted that to see if it was that
so you reverted 57c8a38 Fix threading and tapping with spindle turning backward (M4)
did you revert it with git or manually?
I am going to have to stop beore I go literallyn msad with this keyboard
git revert <long stringofhex>
talk later. get on withb your day
is your servo thread really 10ms?
I get an OK result with the tip of master: http://emergent.unpy.net/files/sandbox/2.5pre-simlathe-100hz-threading.png
so all I know is that my setup is different enough from andy's
EMC: 03jmelson 07master * rcd2aaa9ac182 10/src/hal/drivers/hal_ppmc.c: remove unneccessary buf declaration
wonder if he intends that these changes go into 2.4
At least some of them, I think
06a279c for instance
They all look fairly safe
I'll figure out what makes sense and cherry-pick them back
the android emulator emulates a physical keyboard
Unfortunately the key legends don't match what the keys do
E.g., it shows that less than is on H with the ALT modifier, but that actually produces [
Which is shown as being on the F key with ALT modifier
git checkout master says "you are already on master"
git describe says "v2.4.1-358-g6d6b94a"
'git checkout origin/master'
different than 'git checkout master'
Ah, I thought that was an either-or !
And perhaps this is the problem?
fatal: Entry 'src/hal/drivers/mesa-hostmot2/hm2_7i43.c' not uptodate. Cannot merge.
(Though I doubt it)
that indicates you have non-commited edits to that fi
commit them with 'git commit' or get rid of them with 'git checkout src/hal/drivers/mesa-hostmot2/hm2_7i43.c'
I think I tried that checkout the other day.
look like a proper version, written by grown-ups?
OK, I guess this is something I can't check through SSH
sweet, probing when rotated works great
I was boring a hole that was split along X, so it was better to probe it while rotated 45
jepler: in other news, I think I contributed to the failure of the spindle bearing by mounting it in a too-small hole
the reason was the spindle is not remotely round - it is 7-sided
I measured it with a micrometer and then bored a round hole to match that measurement!
nice for probing not nice for bearing failures
assuming it's round, that works great! if it's 7-sided, not so much
jthornton: fixed already - I got it apart without destroying it and pressed in a new bearing
I really wonder how he turned it 7-sided
yep, evenly spaced, like it was on purpose
I don't have a good way of measuring the deviation, but I'm pretty sure I can even feel it
now I have to fix the motor or find a new one
cradek: Got a DTI ?
I guess I could mount it in the lathe and rotate it, duh
cradek: that's what I was thinking
I thought I'd need a 3 point measuring system, but 1 works too
* Jymmm strongly believes in the KISS method =)
yep I did that - I bored the hole oversize until it fit :-)
heh, if it doesn't fit, you didn't use enough explosives!
cradek: one of those curve heptagons like the british 50p piece? (Constant diameter, non-constant radius)
yes I suspect that's the shape
centerless grinding can easily cause that
[19:38:53] <andypugh> http://www.imagebin.ca/view/CtealI.html
Same problem with the latest master.
Might well be Hostmot, the index-enable lines are not going high for long enough to register in the servo thread.
But I am surprised that G76 carries on without an encoder reset.
I might add a base thread to get a better halscope.
if you know a good and bad revision, you can use git bisect to narrow it down to one change
(whoo-hoo 1/2" sch 80 pipe has a near-as-makes-no-odds 14mm bore)
I don't know when I last threaded.
we do know a good revision -- v2.4.1. git bisect start; git bisect bad origin/master; git bisect good v2.4.1
Just like that?
then git will check out a commit that is between those two in history. build, test, and depending on the result: git bisect good OR git bisect bad
I might be some time.
yes, it's not at all instant
but it will probably be less than 10 steps to test
atal: ref HEAD is not a symbolic ref
on which step?
I only typed the "git bisect start"
try doing a 'git checkout master' first
looks like it might be a bug in git circa 2008 that they've fixed since then: http://lists-archives.org/git/645686-git-bisect-start-on-a-tag-revision-fails.html
git checkout master; git bisect start; ...
is "won't bisect on seeked tree" bad? It is saying 176 revisions left to test.
git bisect reset; git checkout master; git bisect start; ...
stop at the first one that gives you an error
and tell me the error
it looks like mmmmaybe the first 'git bisect start' that complained about HEAD left you in a funky state but I'm not sure
It looks like it was working before, actually. (Same node, same number to sort through)
make clean, or just make?
it won't hurt to 'make clean', except for the waiting
Yeah, it was the waiting prompted the question
I've got a short little g code file a few lines of G1 moves and when I add a G0 Z move Axis complains that the program exceeds the machine maximum on axis X, if I take that one line out it quits complaining
you can skip the "make clean" unless you get a message about 'unable to rebuild' .. then clean and make again
JT-Hardinge: I guess you have to leave that G0 move out then
didn't know if that might be a clue
sorry, my answer was a bit flip
JT-Hardinge: I ask only because I did it today, but are you sure the sign is right on the G0 move?
andypugh: the G0 is a Z move and the warning is about X axis
if I take the G0 off the line and leave it a G1 it is happy and does not complain
I guess you could join the git bisect club..
if I add another line with G0 after some G1 moves it complains "Program exceedes machine maximum on axis X"
the limits exceeded warning is a pile of crap and somebody smarter than me should replace it with something that works.
I'm not smarter than you but I'd be happy to try if you could point me to where the existing exceeded warning part of the code is located
a G0 by itself after a G1 move triggers the error
I ought to move the home flag, it would be an efficency gain during this process. (can't home with the guard down, can't run with the guard up)
is the early warning system part of Axis?
OK, I have ended up with an answer from git bisect, but I might have marked-as-good one that was bad (I am hopeless at keeping my concentration at repettive tasks)
I think I reverted the implicated commit, and the problem remained
git revert seems to always open a nano editor window, is that normal?
(It's a pain in the neck as my lathe keyboard has no Ctrl key, so I can't get out)
andypugh: I wish I knew what you were talking about
So do I, it would be a huge help
andypugh: you can tell git what editor to use, not that i have it memorized how, but it's easily changed
andypugh: git config --global core.editor "mate -w"
<--- google says you do it like that
copied invisible linebreaks :(
Oh no! it's homing in on one of _my_ changes!
andypugh: you messed up on the "good to have someone else to blame" concept
Right, I have found which commit broke lathe threading. It is going to be Mesa-hardware only.
# bad: [c9add70ba1d5ac28c21821639a3f41d51cca6f35] hostmot2: add support for three-phase pwm
git-bisect bad c9add70ba1d5ac28c21821639a3f41d51cca6f35
# good: [1a3e9e71de9a090ec79a993014c76d7011348d47] Revert "hostmot2: add support for three-phase pwm"
git-bisect good 1a3e9e71de9a090ec79a993014c76d7011348d47
I fear it is going to be difficult to find, I didn't touch encoders.
Ok, I'm working on the livecd again finally. I need to remove some things for emc2 to fit. I did away with gnome-games, and I'm thinking maybe rhythmbox can go?
mozmck: did you remove the win stuff?
wubi & co?
I don't see wubi installed...
it's an exe on the CD
outside of the linux stuff
otoh, how much do you need?
hmm, I'm not completely sure. I packed the iso and it is about 735MB vs 700 for the original
looks like wubi is only 1.4 meg
rhythmbox is pretty large, and there is still totem for playing audio and video...
oh no I use rhythmbox to play music while I machine!
tough! you'll have to install it manually :)
Oh No... Mr. Bill
I need to get something off the livecd though to make sure it fits on a CD.
anyone use bluetooth to talk to their machine?
I thought about replacing openoffice with abiword and gnumeric. openoffice is one of the largest programs on there.
the only thing I normally us is the openoffice spreadsheet
have you tried gnumeric for that? it's a lot smaller but seems to work fine for me.
I believe abiword will open any document openoffice will...
I've not heard of it
both of those are (or were at least) part of the gnome office suite.
[22:45:08] <mozmck> http://live.gnome.org/GnomeOffice
I'm installing gnumeric now with the synaptic package manager
gnucash is good for personal finance too in case you haven't heard of it. We switched to it from Quicken somewhere around 1999 or 2000.
I've used that one before
mozmck: last I checked abiword and gnumeric were crap compared to openoffice
but I must admit it was a looong time ago
Yeah, but, this is for a workshop machine.
andypugh: I have no problem with removing openoffice
alex_joni: that could be, but I've found them decent for most things I've needed, and way smaller and faster.
I, err, "fixed" the threading issue.
alex_joni: I think they are much better now that they were, but I don't use either very extensively.
Though a spreadsheet of some sort is probably useful, i spent much of last night using one to generate G-code
openoffice calc and gnumeric are both terrible compared to excel when it comes to large datasets (100000 or so records).
there are actually some folks that would love a completely stripped down liveCD with nothing but EMC on it.
Live credit-card CD for stealth installations
hmm, I notice a package called example-content. that takes up around 8 meg
Hi Micges. I nearly unfairly accused you of breaking G76. But then realised that the git bisect procedure breaks if you "good" a build due to testing it wrongly. Then I tried again and found it was all my fault.
andypugh: yay, you found it
cradek: how about removing compiz from the livecd?
Yes and no.
I found a fix, but the fix worries me.
I removed example-content and gnome-games which freed up some, but not quite enough.
(Unless I get to say "race condition" and walk away
mozmck: still having space problems?
yeah, I'm just getting back to this again. I don't want to take too much off.
how much more space do you need?
All they need is terminal, Pidgin, EMC2 and Synaptic.
Not sure. removing example-content should have removed about 7 meg, but I think I needed about 35 meg
did you take out ttf- packages for khmer/lao/punjabi/japanese/thai/korean? I have all those installed by default
nope. lao is tiny, but looks like thai and takao-pgothic would free up about 10 meg
Do we not have users of those languages?
andypugh: they could always install them again
I bought some little servos from Thailand, I think CNC homebuilds are quite big there.
andypugh: it may or may not be relevant that emc isn't currently internationalized for any of those languages, so they'll have to deal with english anyway to use it.
We can dump chinese too then, the only potential user I have seen seems determined that writing a new gui is the way to go. (I don;t think he has managed a base compile yet, though)