* jepler reads Andy Holcomb's last e-mail, about using touch-off in axis
I can't imagine how it does that
I can't either
I'm trying to build emc2-sim deb on a powerpc
(no idea whether that ought to be possible or not)
yep that's the idea
rtapi/rtapi.h: In function ‘void rtapi_mutex_give(long unsigned int*)’:
rtapi/rtapi.h:242: error: ‘test_and_clear_bit’ was not declared in this scope
we've seen this before haven't we?
sounds a bit like the one you get building 2.1 on edgy/x86_64 -- fixed on trunk
is this a favor, or for some machine you actually have?
I actually have it
don't know if it's a keeper or not - just playing mostly
dapper really works nicely - all the built in stuff works, including wireless
beats me if these macros will work in userspace like the x86 and x86-64 do -- http://www.gelato.unsw.edu.au/lxr/source/include/asm-powerpc/bitops.h
wonder why I don't get that #error
did you get defined(ASM_BITOPS_H_USABLE)
looks like configure tests that an <asm/bitops.h> is includable without any warnings or errors, not that it has the macros required
/usr/include/asm-ppc/bitops.h is empty
you could try changing configure to see if this test catches that <asm/bitops.h> isn't usable:
+ [test_and_set_bit(0, 0);],
ok i'm getting trunk to play with
the funny thing about this machine is that it has excellent engineering (strong/clever hinges, handle, ports) but it's the ugliest computer ever made
I wonder if I could sand it down so it has some square corners, then paint it black
and add a few more mouse buttons
hmm, I still get that define
oh, duh, user error
ok now there's no user error but I still get it
configure builds that with -c, which doesn't get any error
if I build it without -c, it does (of course)
argh, I can't get it to work, I'm clueless about this assembly stuff
cradek: I know how you feel about "this assembler stuff", even if I manage to muddle through from time to time.
lerman_ is now known as lerman
steves_logging is now known as steve_stallings
jepler: since powerpc support is only for simulator, is there any reason I can't just write these two functions in C?
cradek: yes, but they have to work properly across processes
jepler: when do you leave?
cradek: in about 5 minutes
and good luck with the keyboards with the Y in the wrong place
or whatever it is
cradek: it's not THAT hard to use that
and you can always switch to english, and not look at the keys :P
the problem will be all the windows machines I have to use
heaven help me if they're vista
back in a few weeks..
hope he has a nice trip
yeah, me too
cradek: do you have emergency access just in case?
don't expect trouble though.
me neither, but better safe than sorry
although judging by the last commits ...
strange! there's still inb/outb/iopl in iosh.cc - must be legacy stuff
rip it out!
so jeff is off on vacation...
jmkasunich: you said something about joomla the other day :P
good for him
alex_joni: yes, I did ;-)
what was wrong?
and start looking at the results
the first one is our homepage - thats good
but now other current linuxcnc.org content appears
google knows about old stuff (the 2nd link is an ancient manual)
yes, google has a nasty habit of not indexing .php
alex_joni: can you help with a configure thing?
probaby to avoid all that dynamically generated content
cradek: I can sure try
jmkasunich: there is an option for SEO friendly URLS in joomla
but I think I didn't get it to run.. there are some things needed.. I'll look at it
I'd like to change the BITOPS_H_USABLE test to actually try to compile and link a program calling test_and_set_bit(0,0)
jmkasunich: it might mean I need to refix some links
right now it only includes the file
(on powerpc the file exists but is empty)
and the stuff jepler posted doesn't work?
I would have suggested the same
no it doesn't
rayh_ is now known as rayh
it compiles it with -c
so it compiles to an object file just fine
alex_joni: I think we have to use AC_CHECK_FUNCS but I don't know how
* alex_joni reads through the autoconf docs
ld: objects/classicladder_rt.tmp(.text+0x9c): R_PPC_PLTREL24 reloc against local symbol
objects/classicladder_rt.tmp: could not read symbols: Bad value
I think you can only check funcs in libraries
alex_joni: i almost got it to work
but it doesn't print the right things to the user
I think there needs to be AC_MSG_RESULT or something
does that work?
I think so, but it prints some extra junk
I think the check headers does that
maybe try AC_CHECK_HEADERS(asm/bitops.h, , [ASM_BITOPS_H_USABLE=no])
jmkasunich: better now?
I'm not sure I like this way better, but I think google does
[20:30:38] <alex_joni> http://www.linuxcnc.org/content/blogcategory/15/12/lang,en/
<-new link style
[20:30:56] <alex_joni> http://linuxcnc.org/index.php?option=com_content&task=view&id=11&Itemid=10&lang=en
<- old link style
cradek: did that work?
I think so
alex_joni: much better, thank you
the extra junk (some blank lines then "sim") must come from somewhere else
* jmkasunich is old school - a URL is a path to a file
jmkasunich: I see it's not always consistent.. but it works with both
so .. I'm not that worried
tain't no ? in a path
[20:34:04] <alex_joni> http://www.google.com/custom?domains=linuxcnc.org&q=emc2&sitesearch=linuxcnc.org&sa=Search&client=pub-noaccount&forid=1&channel=pub-nochannel&ie=ISO-8859-1&oe=ISO-8859-1&cof=GALT%3A%23008000%3BGL%3A1%3BDIV%3A%23336699%3BVLC%3A663399%3BAH%3Acenter%3BBGC%3AFFFFFF%3BLBGC%3A336699%3BALC%3A0000FF%3BLC%3A0000FF%3BT%3A000000%3BGFNT%3A0000FF%3BGIMP%3A0000FF%3BFORID%3A1%3B&hl=en
jmkasunich: that shows even some ? in indexed files
retch puke hurl
(I'm retching at the 250 character long URL)
guess they pass the icon there :D
about 9000 pages indexed by google for linuxcnc.org
that's quite a bit
seems like more than really exist
no, lots of irc logs
and everything else there
speaking of handbook
it's quite a bit of information ;)
* alex_joni is not responsible for THAT
jmkasunich: go on..
[20:39:42] <jmkasunich> http://www.linuxcnc.org/docs/2.1/
shows an integrator manual
but its not linked on the documentation page
in fact, I have no idea how to get to the page I just posted from the homepage
then I guess we need to add them links
I should just be quiet, since I don't have anything positive to say....
no, I appreciate the pointers
I'm bookmarking http://www.linuxcnc.org/docs/
* Documentation for the latest EMC version 2.1, in HTML and PDF format, is available.
as my new "preferred way to get to the docs"
that's the last entry on the Documentation page
and it points to www.linuxcnc.org/docs/2.1
the www.linuxcnc.org/docs/ is more general, but more confusing for new users I think
the prior links on that page, to User Manual, Hal Documentation, go to the trunk versions I think?
might be so
if you have any oppinion I can easily change that
opinions are cheap - everybody has them
I feel like moving the line I just pasted to the top
and move the others down
I'd be more inclinced to make the others refer to the 2.1, and say "docs for the latest unreleased version are available"
95% of the people to whom we find ourselves wanting to say RTFM, need to RTF 2.1 M
that should be the one that is easy to find
I really should learn how to edit joomla, so I can fix instead of bitch
I think the pdfs in linuxcnc.org/docs/ are the same ones as in 2.1
if so, thats dumb
I changed it right now.. let me know what you think
you moved around the directories too?
why not just change the page?
nope.. just changed the page
I like the old dir structure
didn't touch the dirs
a new dir?
no, that's the TRUNK version
so in linuxcnc.org/docs/ we have 2.0/, 2.1/, and decel/, AND we have another copy of the manuals at the top level?
why the duplicate copy at the top level?
question about charge_pump when you get a chance.
jmkasunich: I think the top dir were there first
I want charge pump to reflect the state of emc's estop
then jeff added the 2.0, 2.1 and devel
rayh: you want it to stop pumping when emc is in estop?
you could make one with a not block, and turn it on and off with an and
the charge pump component is a simple oscillator - we have two choices - AND its output with a controlling signal, or add an enable pin to chargepump
the latter would only help people using TRUNK, but is cleaner IMO
right i see that.
I was hoping for a pin on charge-pump named enable.
I can add that
hmm, I wonder why jeff made it a singleton?
anyway... adding an enable is easy... making it default to enabled if you don't connect to the pin might be a bit harder - I'll have to read the comp manual
* jmkasunich tells /me to RTFM
hmm.. there are a lot of people visiting linuxcnc.org
right now 200+ visitors
I wonder if all the checking robots that are crawling it right now appear as multiple users?
You can see the growth in popularity from the rash of new names posting to the lists.
Year 694335 87330
roughly for the last year
jmkasunich: see match8.comp
about 10% open the download page
cradek: thanks - I was reading the pre-2.1.5 version of the comp doc, which showed "= STARTVALUE" for params only
about 8% go for docs
current doc shows it for pins too
and 5.6% for About
heh.. bet those 200+ visitors were robin's webcrawler
not to self - don't use man page markup in a pin's doc string
so others may?
rayh: charge-pump with enable is now in TRUNK
sorry it took so long
I missed a }, and in comp its a bit harder to find that kind of error than in straight C, because the error messages are obscure
Hey thanks John.
I'll pick it up right now and try it.
anyone want anything else in 2.1.5?
clever that it starts enabled unless a sig is attached.
not sure about the mini fix I just sent
and maybe we want to tackle the tkemc G92 stuff?
we'd have to agree what to do before tackling it
I haven't heard yet what ray wants to do - and he's clearly busy with something else - so it's probably not the time to try to get something in there at the last second
yeah, I agree
I can talk for a few.
but I don't see the need for an urgent 2.1.5 though
the solution that axis uses is to change g54 when you right click.
I wonder if we shouldn't wait a bit for more stuff to pop up
... when you push the touch-off button
yeah maybe we should wait another week?
does axis try to deal with any of the other coord systems?
Okay by me.
no the gui doesn't do any special handling of the others
It's probably just the quirky way I use g54 that makes me worry.
axis's touch-off is useful for the usual program that runs at one default work offset
Right and I can see the value of that for many users.
This week in a couple of programs I was using three at a time. And g54 was my machine home.
that way I don't have to do something like
to get away from the work.
g28 z0 might do what you want in that case
when doing something like that as an advanced user, do you not use g92 or rightclick at all?
I don't use g92 cause it's difficult.
I use teach in either tkemc or mini
rayh: can you describe the exact steps?
but I leave g54 alone so when I need machine home I can get it.
what do you suggest we do to fix these common questions about rightclick in tkemc?
maybe only rightclick needs to go away, and the rest doesn't need touching?
I've always imagined tkemc giving the widest possible range of reach into the EMC part of EMC2.
I've never thought of it as a machine control
I imagine a machine control having a subset of emc abilities.
that's new imo
and that is what axis does.
It gives the user the ability to use only one of the coord systems.
that's not true
cradek: ray talks about GUI here..
You still have the ability to use others through g code.
you can use the systems however you want. You set them with mdi.
I think the mini teach can let you select which coord you use
if axis would let you set any system, would you be ok with making it the default for stepper/inch and stepper/mm?
I think axis is good for new users
I don't doubt that.
if that's all that you dislike about it, it can be fixed
No my dislike has little to do with the future of emc2
you, we need to consider wide audience.
rayh: your oppinion is one of the appreciated ones
so it's good to hear your concearns
right now I'm thinking only about the default settings for new users - advanced users can do all sorts of things
(in a productive way ;-)
And if we changed default interface for stepper and most sims to axis.
I'd probably bitch but...
I guess the only real coordinate issue I've got is how to get rid of trash when some unexpected offset happens.
In tkemc and mini I can look at all of the coordinate offsets and zero them out.
that's a SMOP
but it wouldn't be something bad to add to AXIS..
using the same touch-off scheme but with a specified coordinate system would be nice
If we had a menu item that would do that from mini, I think I'd be able to say, well push that it see if it goes away.
you could then use your probe/wobbler to teach your fixtures just like touch-off
tkemc and mini use a reset to clean out or change these so that both the var file and the global variables match up again.
long ago we had a bug that would nuke the var file (5220) but I'm pretty sure that is gone -- I have not had a screwed up file with emc2
rayh: so basicly you're talking about a dialog with a table on it which holds all coords offsets, and user can see them easily, change them, etc
I remember some of those, it would write 0 in 5220 and you couldn't get emc to run again.
rayh: right I saw that too, but not for a long time
Yes. And I don't think we need to even teach axis users about 55-59 just make it possible for them to clean out the garbage.
rayh: that's quite easy imho
let me think about a good design for touch-off for any system
cradek: how about having radio buttons for the coord system on the touch-off dialog
G54 selected by default
If you get one, how about it zeroing out g92 any time axis starts up.
rayh: shouldn't M2 do that on shutdown?
rayh: I think g92 is not applied by default at startup??
I could be wrong
I saw some really quirky stuff but most of my work this week was with old BDI.
"Sometimes the values of a G92 offset get stuck in the VAR file. When this happens reset or a startup will cause them to become active again."
I'd have to take the time to see what a g92 offset in the var file does on startup.
that's from http://www.linuxcnc.org/docs/2.1/html/gcode/coordinates/index.html
The whole implementation of 92 is really difficult to use so probably it should changed.
I tried to figure out what the "bug" in it was but when I set it up like the spec said, it worked pretty much as advertised.
I've used it within a program (cleared at the end), but having the gui let people manipulate it seems to just lead to trouble
I *think* the main problem with G92/tkemc is that the user has no idea what happens when he right clicks the display
yeah I read your paper describing the tests - I also agree it works as it's defined
exactly/genau The problem IMO is that it is a very complex concept.
rayh: when a user right-clicks the display, he has no idea that the offset he enters will vanish after he stops/reloads the file
or that it might cause bigger problems on restart
yes it took a lot of readings of the spec to understand how the coord systems all work. a new user just isn't going to know all that (nor does he need to!)
unless he uses g92.x or
I think the proper solution for this would to pop up the coords dialog on right-click
I can't even keep em sorted out.
rayh: he has no way of knowing that
* alex_joni talks about new users now
The axis solution is a good one for new users.
rayh: so it's ok with you that I change the default gui for stepper?
well, how about datumtalbes?
* rayh bites his tongue.
and the tkemc/mini solution for multiple coord offsets is a good one for existing users
cradek: the _only_ problem I see is for people building emc2 without python & co installed
rayh: we could vote or something but that seems silly to me
As an old time emc user, I have my ways and they are not the axis ways.
the configure script will continue, and they'll get a working system..
not the same can be said about tcl
no need to vote
rayh: you also know how to easily configure it to use any gui
rayh: I don't think it's the experienced users we have to worry about at all
I think we have to worry a lot about shop folk.
because none of the existing gui's work like a shop machine tool.
Axis works pretty much like a PC program so folk who do other PC things can appreciate it.
It has a great tool path display.
if we are taking a more machine-tool-ish display touchscreen usage and stuff should be considered..
anonimasu: we're trying to decide what's best for new users who are just seeing emc for the first time
I wish I had the ability to write python and adjust it to my liking.
I'm really on anomimasu's way of thinking here.
cradek: yes, but I agree with rayh.. it's not nearly anything like a professional control..
we have a wide range of users with a wide range of expectations
we won't make them all happy at once
but I like the sim/xxxxx configs that show all the gui options
No and that is why we need to preserve the ability to use more than one kind.
cradek: axis is all great but if you look at it it's not reallu useable..
what I want to do is change the default gui for stepper only. no industrial users are using stepper anyway.
anonimasu: why is that?
The system I'm building now is stepper powered.
small buttong making touchscreen usage a pain..
I don't like it but price point is king.
all the guis require a bit of mouse, but are better used through keyboard
anonimasu: the typical new user with a stepper machine will not have a touchscreen. he will have a mouse.
you're one of the other audiences I was talking about.
This is a good topic for discussion.
cradek: I like axis very very much..
I would like to see us produce a great system for guys who use fanuc or siemens.
cradek: it's easy to use and works like most pc applications..
I'm sure that the newbie would be reluctant.
cradek: but if I were to implement a gui at a machine tool I sure wouldnt run axis..
And Axis can fill that need.
I'd be happy to see someone write a fanuc-a-like gui. it would have to be someone with knowledge I don't have
[23:23:33] <alex_joni> http://www.foremostmachinery.com/Images/Products/JFYRTFanuccontrol.jpg
<- would that be a Fanuc GUI ?
they dont look too nice right?
Yep and that one is even getting cluttered.
still they do what they should, run a machine
[23:24:47] <alex_joni> http://www.shodausa.com/images/fanuc2.jpg
you'll never hear me say I don't want people to write more gui options for emc
that's the only way, because no one gui will please everyone
One chinese factory I visited said they liked one "local" control best because it was easy to run.
before axis, I happily used xemc
So I asked to stand beside a lathe operator while she worked.
and this is siemens? http://www.knuth.de/prod_img/steuerungen/Siemens802d.jpg
The major pages she used were tool offsets and program edit.
Sometimes she would mdi for some specific purpose.
Then back to auto and cycle start.
It is possible to accomplish some of the motions she used with the keyboard.
but dedicated panels reduce error.
I think reducing the possibiltiy of error is the key thing for a commercial gui.
ok, so to summarize it before I go to bed ;)
* anonimasu nods
we need to extend AXIS to allow touch-off for other coord offsets
we need to add a dialog to AXIS to edit/clear out coords offsets (g54,g55, etc)
we need to make tkemc right click bring up something else than the current G92 offset applier
we need a new GUI for shop floor setups (not something for this night :)
alex_joni: perferably with different modes for auto/mdi/manual..
and rayh will grumble but not object to changing the stepper sample config gui to axis
hitting cycle start in the wrong place shouldnt cut your hand off ;)
sounds good to me.
great conversation guys, thanks
There will not be a lot of agreement among commercial users about the look and feel of that gui.
anonimasu: there are different modes for manual/mdi/auto
yes it's good we have other options
some like heidenhein, other fanuc, others siemens, others haas.
and it sounds like maybe they need an option we don't even have yet
we dont need to copy a commercial interface, though but it's sane to have one that works similiarly to one..
also, the safety stuff..
One that separates out functions and minimizes the possibility of bad mouse clicks.
auto isnt a separate mode..
alex_joni: if I hit start in manual or mdi the machine will start right?
or is that onlu in manuall..
you can't go to manual or mdi while the machine is stopped
err.. I mean paused
or you mean starting a new program?
In the original tkemc start or program load would not switch mode.
but switching mode from auto could raise all kinds of havoc.
I do allow program edits while mini is running auto
alex_joni: Well, auto shouldnt be easy to go to from manual, you may be having your hand close to the spindle with a piece of paper..
for setting zero as a example..
but if the interpreter is not idle it asks if you really want to kill of the current run.
But I agree with alex's list and with chris's thought that we need an easier than tkemc for newbies.
Well listen guys, I'm sitting in a warehouse that is rapidly getting down to 40f.
rayh: thanks a LOT for the talk
You bet. I've missed these "talks."
I'd hesistate to complexify axis a lot for the newbie.
my goal would be foremost keeping it as clear and easy to use as possible
I didn't even like the idea of allowing all the pyvcp stuff crowding the image.
rayh: newbies surely won't use that
I feel that way too, but that's an advanced user thing only
flexibility + simple defaults = winning combination
cradek: what kind of toolkit do you use for axis?
I found some movies here: http://www.festo-didactic.com/de-de/lernsysteme/cnc-technik/emco-cnc-technik/software.htm
would those be consistent with shop-typed machines?
let me see
I'm lined up for the Bejing tool show again this year.
I promise to take and put up pictures of controls.
I only saw one PC based control at IMTS last year that I liked.
and it was running an edm
alex what are we talking about
touch was good cause it took hours to run a program.
alex_joni: well, _remotely_
anonimasu: they emulate siemens and fanuc controls :P
they dont look like dos programs in the real world :D
how are you doing
we got started talking about the look and feel of some commercial tool controls.
in what respect
Mostly the differences between PC look and feel and a Fanuc like.
there's been aloot of changes over the last ten years
I'm with Chris that we need a PC look for the first visit by newbies.
alot of them went to touch but know there changing back
* alex_joni changes back to sleep mode
good night all
Swarf and coolant will kill touch in a big hurry.
Catch you later alex. Sleep well.
true and scratched sreens
thanks.. night all
Gotta get some food and warmth. I'll probably not be back till morning.
morning is so close ...
closer there than here.
alex_joni: that looks like some controls, but it's hard to get a idea how they actually work..
without keyboards/keys and stuff
they don't use as many hard wired buttons any more
* rayh really away!
we didn't use many on the old k and t' either
i personly like hard eired button's but they cost more
alex_joni: you should probably find a haas userguide/fanuc/heidenhain..
alex_joni: for better idea about it :)