EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/ (Master_Integrator_fr.lyx Submakefile): remove reference to deleted file in french manual
EMC: 03bigjohnt 07v2_2_branch * 10emc2/docs/src/ (Master_Integrator_fr.lyx Submakefile): remove reference to deleted file in french manual
good night everyone
see ya. thanks again
hey all, is there a component for getting the inverse?
for bits anyway
you'd scale with a -1 gain for floats
oh - 1/x?
there isn't a 1/x - it would blow up if is close to 0
if X is ...
t div by 0
Roguish: take a look at the bottom of this page http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?Simple_Remote_Pendant
sure, but what's the output?
as the input goes from +1 to -1, you'd go from 1 to some large positive number, then 0 inside some deadband, then -some large number, then on up to -1
it's a simple enough component to write, but I don't know that the output would always be useful (and there's no way of propagating computational errors in HAL)
simple for some people...
there is a float inverse example on the page I linked
* BigJohnT must go check the ribs he can smell them down here in the beer cave
invert and negate aren't the same thing :)
i don't see the inverse. i do see changing the sign.
SWPadnos, recall i was having a realtime delay problem?
yes - you thought it was CL
i have isolated it to my spindle control hal file.
does that include classicladder, or is it just other HAL stuff?
still not sure what the problem is, but if i exclude the hal (hal only, not cl) i do not get the error.
can you post that file?
i would rather not, it's really messy.
eh - we're all friends here
that's what i am working thru now.
and you can set the post to expire on pastebin ;)
tomorrow i have to start tearing out an old delta-tau control and replace it with emc2 on a vertical lathe.
x,z and c axes.
oooh - sounds fun
for some values of fun
SWPadnos: do I misunderstand the question?
he means 1/x, not -x
[00:36:31] <Roguish> http://pastebin.com/d30b8dca8
hope you enjoy my pain.
did you do a separate thread to try debugging the RT delay?
i want an encode on the spindle and to control it like an ordinary axis motor, just with the spindle speed.
ok - was it originally in the servo thread?
ribs have my mouth and eyes watering
no, country style what ever that is
looks like short ribs
& cold beer?
4 kegs on tap
kegerator only half full
sounds delicious and nutricious.
yummm it is
so, with the code as it is, except for using the servo thread, that HAL file causes problems?
(or even as is, with the extra thread)
well, i've been into it for a little while right now, so it's gotten a bit more confusing with the line comments and such.
ok, so the RT blip occurs basically as soon as you start EMC?
it's probably something to do with the pid loop.
no, takes a minute or 2.
and then you don't know if it happens again because the error is suppressed
there is a problem with cl and opengl, but that is a different issue.
i was asking earlier about the tool diameter and length just to get them into vismach more realistically, and dymanically.
i like vismach for visuallization. probably turn it off on a production machine.
well, I don't see why that file would cause a problem. I'd probably need to load up your whole config here to get anywhere
and I can't do that for a few weeks :)
if it happens again while i'm messing with the spindle, i will bring it back up. otherwise i'm not too concerned. it was just experimenting for now.
the cl problem is more of an issue. cl gui dies after a few minutes of working.
turn off all desktop effects, if you haven't already
never have any of that fluff running
it's on by default in 8.04
yeah, i shut it all down. hate that eye candy.
ok. that's the only solution I know of for display problems on EMC2/8.04 :)
(shouldn't matter, as CL isn't openGL)
i'll try to capture the error message and post it.
only matters for setting up the machine. i won't run the gui in production anyway.
I assume it's only the GUI that die - that the RT ladder code continues to work ...
will the emc2 install script work on the server version?
i want a really stripped down system.
yes, it should
it will bring in a bunch of stuff though
I think people have used xfce, but I haven't heard anything on the subject for a while
that's a gdm isn't it? i don't mind gnome, i just don't want the myriad of applications .
yeah - don't install ubuntu-desktop - I think that's the one that pulls in openoffice, the multimedia suite, etc.
an uninstall permanent?
until a reinstall (of the package) :)
I don't know what happens if you uninstall ubuntu-desktop though - it's a metapackage that just depends on stuff
I don't know if it will remove any other apps
I've tried emc2 on several of the light desktops. They all seem to work.
ok, cool. I just hadn't heard anything about that stuff for a while
ok - I write an invert component - do people think it should output 0 when the input is near 0, or should it output some large value (1/deadband, for example)?
I guess I vote for 0, since that gets rid of issues like flipping between +/- 1/deadband
how do you plan to handle 0 deadband?
it's set to 1e-8 minimum - that's checked
That is a difficult thing to get my head around.
small numbers ought to produce large and large produce small
but what are the limits of large and small.
yeah - I had to pick something. I could have chosen something like 1e-20 I guess
still can actually ;)
if I clamp to positive numbers, then outputting 1/deadband makes sense
it's the toggling between +/- "infinity" that gets me
No scaling just 1/xxx
I could stick a gain parameter in there as well. maybe offset too
but that gets a little weird (when is the offset applied?)
if 1/0 produces 0 and 1/0.00001 produces a hundred thousand there is a discontinuity somewhere in there.
there's always a discontinuity at (or in this case near) 0
and I can see the +- thing really being a problem for some emc variables that might be used.
I'm just thinking about a signal that may have "noise" - it pops between +something small and -something small
I guess the check can be for in==0 - if it's 0 then don't update the output
so it would leave the last value until a nonzero input came along
What are you thinking of attacking with this module?
I don't know :)
Roguish asked about it
actually, it should be a divide component, not invert
that way it could be used to get ratios (so you could see things as percentages, for instance)
could you make the near zero value a parameter so folk can play with noise canceling?
it is, but it's made so the parameter can't be closer to 0 than a magic number I picked (1e-8)
you can make it bigger if you want :)
maybe I'll go with 1e-12 and be done with it
I still worry about the uncertain meaning of zero because both very big and very small divisors will produce it.
But why not. If you write it they will come.
I could just decide that anything within the deadband gets you +1/deadband
that should be a bit clearer.
actually, it's not a big deal to make it -1/deadband from -deadband to just under 0, then plus 1/deadband from 0 to +deadband
That sounds like it would work for something like deadband.
deadband is a constant?
it's a parameter, which must be >= 1e-12
(I changed it)
Okay. And why do we then need a module to invert it. Or are we watch fe in relation to deadband?
trying to watch...
this is a component that takes a float input, and outputs 1/input on its output
I put in a parameter that I called deadband, which tells the component how close to 0 it can go before deciding to stop doing 1/x (since 1/0 the easiest way to get NaN results)
I wasn't sure what the output should do when the input is close to 0, which is what I thought we were discussing :)
Oh not axis deadband. Variable deadband.
out = 1/in unless |in| is too close to 0
with "too close" user definable
as joe Pisci might say ok ok ok
he might also say "they fuck you at the drive through" :)
but that's not important right now
never go to the drive through
"because they know you'll be miles away before you know you've been ..."
Good luck with that component. I'll leave you alone to work.
I'm working on a 2-gear gearchange thing now
catch you later.
the logger text file is goofy.
(just on devel)
Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emcdevel/2008-07-07.txt
it is missing usernames and CR's
it looks fine to me
oh - ok
this is what it looks like for me http://imagebin.org/22054
well, that's not right
looks good to me using Mozilla on Windows
that's the link above -07-07.txt?
heh - first time using IE7?
to look at the logs I mean
I bet it thinks it's a marked-upu file, since there are all those tags :)
no - normally they are just fine
like <SWPadnos> ;)
any recent updates?
this is emc yesterday - only cronic replied today.
[02:28:47] <skunkworks> http://imagebin.org/22055
I don't understand
yesterdays devel log looks fine. oh well - probably just a fluke.
I bet one has a non-ascii character in it and the other doesn't
yeah, could be
I doubt the pages are being served with the right (mumble mumble)
oh, I see
mumble headers mumble apache
oops - I should have said I'm using Firefox on Windows, not Mozilla
heh. I hate to hit and run.. but I need to go to bed now. :) ttyt
EMC: 03swpadnos 07TRUNK * 10emc2/src/hal/components/ (invert.comp gearchange.comp): Two new components: 2-gear spindle gearchanger/limiter, floating point inversion (1/x)
EMC: 03swpadnos 07v2_2_branch * 10emc2/src/hal/components/ (gearchange.comp invert.comp):
EMC: From TRUNK: two new components, 2-gear spindle gearchanger/limiter, floating point
EMC: inversion (1/x)
hmmm. any naming suggestion for a component to do the "deviation from average" thing?
(I still don't understand how it's different from ddt)
it's not ddt, since it's deviation from the long term average, not deviation from the last sample
can that be constructed with ddt + some other block?
I don't think so
you can do two filters with different time constants, then divide one by the other
(scale by 1/x :) )
EMC: 03swpadnos 07TRUNK * 10emc2/src/hal/components/gearchange.comp: Oops - make negative speeds work correctly (M4 now works)
EMC: 03swpadnos 07v2_2_branch * 10emc2/src/hal/components/gearchange.comp: backport from TRUNK: fix negative speeds
SWPadnos, thanks for the inversion comp, way to go.
make: *** No rule to make target `../docs/src/examples/misc.lyx', needed by `depends/examples/misc.d'. Stop.
BigJohnT: did you miss adding a file?
(this is TRUNK)
I might have jepler
it is in my latest submakefile
jepler: thought you were travelling?
alex_joni: I am
just checkingi n
heh, have fun
I had to do a make clean to get rid of an error on make where the french intergrator manual was looking for a file I replaced with another one
jepler: yes I did :)
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/examples/misc.lyx: misc hardware examples
EMC: 03bigjohnt 07v2_2_branch * 10emc2/docs/src/examples/misc.lyx: misc hardware examples
alex_joni: does the french manual get updated automagicly?
time to go
I followed these directions here http://www.ryancloke.com/ubuntu-804-hardy-heron-live-usb-how-to/
seems to boot - but I am assuming what I changed in syslinux.cfg borked the realtime.
Good morning ray
how is the weather up there?
was a beautiful weekend here.
can you pastebin what you did to syslinux.cfg? (and maybe the original from the EMC2 liveCD?)
A bit cool but looks like a nice one today.
let me look
I saw a message during make of the docs last night that said the "forward" file was an article type rather than book type.
This is the one from ryans site
[13:16:06] <Skunkemc> http://pastebin.ca/1064601
I was trying to keep up with your work using 8.04 but that doesn't work. Wrong version of LyX.
yep - I'm looking at his page :)
ok rayh I can fix that when I get home
Okay if I don't get there first.
I'm setting up a 6.06 again here.
rayh: unfortunately that's true
This is http://pastebin.ca/1064603
on the live cd (isolinux.cfg)
the 6.06 Lyx works ok with 8.04?
I tried the LyX export as the older version and it makes a mess.
rayh: do you know where to change it from article to book?
hmmm. I don't see things like syslinux.cfg in cvs/infrastructure/livecd
or anywhere else in CVS, for that matter
coffee time bbiab
mv isolinux.cfg syslinux.cfg
That change is ... hang on a sec.
this is the command - mv is move I assume
the menu is Layout->Document
and it's the first spinbox there.
if I understand it right - the process moves all of the isolinux directory to root.. then renames the isolinux.cfg to syslinux.cfg. Then i pasted what was on ryans site into the syslinux.cfg file
*root directory of the keychain drive.. (in effect)
yep that sounds like the place that I found it :)
My 6.06 LyX is giving me fits so I may have to leave that one to you.
ok rayh I'll get it this evening when I get home
How did you install LyX 1.3 in Ubuntu 8,.04?
I don't have 8.04 installed
Oh. Okay. I'll play around a bit here.
I hit the upgrade button to 8.04 the other day but got the chance to chicken out and I did :)
* alex_joni has terrible response times lately
skunkworks, I saw what the instructions were - I wanted to see if there were any differences between the stock liveCD isolinux.cfg file and the one on the EMC2 liveCD
if there are any changes for the EMC2 disc, they would be overwritten when you copy the one from Ryan's site
the only other file that is touched is the casper file.. (I don't think that should effect anything though)
would it make sense to (a) put all the files from the iso on the website and/or (b) put things like the kernel config used on the liveCD in CVS?
also, unrelated, I shuold be checking in a PCNC-1100 config - what do people think about config directory structure?
I'm thinking we should have subdirs for each manufacturer, with either subdirs under there, or just multiple ini files
the livecd uses the same kernel as all the installs doesn't it?
as far as I know
but if you're going to tell me to look in /boot, that's not possible unless I've already downloaded and booted the ISO
I think the chooser doesn't support arbitrary directory nesting, so I guess something like configs/tormach/pcnc-100.ini
ah, ok - I'll look at that
yep - you're right. those pluses I've seen are for different inis - thanks
Hmm - I am wondering if it is an issue where the home directory is on the persistant usb install... Let me play with it for a bit.
SWPadnos: I started adding things regarding building a LiveCD into CVS
latency test ran just fine.
cradek, How much trouble would we have to fix the chooser to allow nested directories
I saw your notes
but I'm a bit worried about adding them all
then I will probably escape on a route of "it's already documented there.. anybody can do it... I don't have to"
I think it may be beneficial to track config changes actually
rayh: do you think there's a need?
well, there are notes on how to build a CD, but not (for example) a starting kernel/RTAI .config
SWPadnos: you can get that from every CD/install
it's in /boot/
to get that, you have to download and boot the ISO
or mount it on DH
mount -o loop ..
I don't see the config on the ISO
I did that on my laptop
then you probably need to extract the squashfs
you have to boot because it's in the [compressed] filesystem
Nested ini directories would clean up the configuration tree view a lot.
or is it supplied by the running kernel, like /proc entries ??
SWPadnos: nope.. it's a file
pico->univpwm, univstep, ppmc
mesa ->5i20, hostmot, +++
rayh: yes, but that makes ~/emc2/configs a bit more of a trouble
if the user pushes the button: create a config copy..
yea it does.
only copy the final dir, and hope that no two are named alike ;)
not the whole path
should that go to ~/emc2/configs/configname or keep the nested tree
SWPadnos: there is another way
download the linux-image deb
and the file should be in there
it seems it would be easier if some of those files were in CVS or on the website ;)
so far no-one needed them
I'm rapidly approaching the opinion that config is getting a case of bloat.
stepconf's config file may point to a direction here if we make a runtime setup.
steve@laptop-m570:/mnt$ sudo mount -t squashfs -o loop test/casper/filesystem.squashfs squash
mount: wrong fs type, bad option, bad superblock on /dev/loop1,
missing codepage or helper program, or other error
rayh: seems like we could get rid of all the basic stepper configs that over time become more and more out of date
I think we had considered putting configs into a separate package or area on the website/CVS
SWPadnos: can't mount the squashfs like that
only the iso
ok - that's what the web seemed to say :)
Certainly as stepconf matures we can switch users from hard coded configs toward using it.
What worries me just a bit is the "hard coded" data that stepconf uses to build it's configuration.
SWPadnos: hang on, I'm commiting the config soon
it's meant for simple stepper configs, not everything anyone can dream up
I wonder why booting from this usbdrive chunks the cdrom drive.. chunk chunk chunk chunk
Sure but a small change in the nature of EMC configuration will require changes to the data in stepconf.
SWPadnos: need to figure out pageant first
it's been a while :D
alex_joni, don't worry about it now - I forgot why I'm looking for the info :)
well.. now I got the config
and found my private key for putty/pageant
and I'm checking out infrastructure
rayh, what kind of changes are you thinking of? things like changing the name of a HAL pin on some component?
* rayh better retreat from the keyboard and read a bit of code.
EMC: 03alex_joni 07TRUNK * 10infrastructure/livecd/kernels/hardy/config-2.6.24-16-rtai: config file for linux kernel 2.6.24-16-rtai
now it's in CVS
hey - it might help if I had squashfs-tools installed ;)
according to a couple of websites, you are supposed to be able to mount a squash partition like I tried
maybe if you have squashfs as a known fs type
huh - stuff run from the home directory doesn't work.. :)
sorry - configs.
let me see if I can get some info
emc will run configs from etc... (or wherever the default configs come frim)
[15:21:36] <Skunkemc> http://pastebin.ca/1064695
(there are 3)
[15:22:29] <Skunkemc> http://pastebin.ca/1064696
[15:23:41] <Skunkemc> http://pastebin.ca/1064697
your var file is missing
when you start a config from the usb home directory - it never starts.
IOError: [Errno 2] No such file or directory: 'stepper.var'
odd - I wonder why that doesn't get coppied..
are your home directory and /tmp writable?
yes - I saved stuff to it.
let me play with it..
it is pretty cool with the usb drive being persistant - it keeps configs and such that I put into my home directory
That fixed it..
putting a stepper.var file in directory
there was a stepper.var but it was locked
since they were copied from a CD
is that it?
but - the other files hal, ini seem fine..
those don't need to be writable
vars get saved
I just did it agian.. and it coppied everything ok.. odd
[16:17:20] <skunkusb> http://imagebin.org/22079
neat. that's booted from a usb thingy?
plus it is persistant.. so everything I do gets saved.
hmm - doesn't boot on this vosto
Skunkemc: can you append boot options?
I don't get anything.. - it seems to be a bios error
I updated the bios - same thing.
I know I get the same thing at home with different hardware.
ah.. well.. that's it then :)
this is stupid.. I did this to make it work on newer hardware.. but really new hardware doesn't seem to boot off of this usb drive.. I wonder if it is the usbdrive or the motherboard.
It boots on pentium 4 class computers so far.
wait - older pentium 4 class computers - but new enough to have usb boot..
[20:36:53] <skunkworks> http://www.dellcommunity.com/supportforums/board/message?board.id=dim_bios&thread.id=60118
so - back to my old permissions question.
does anyone know how to make a directory that's accessible to a user via FTP, but which that user (a) can't get out of and (b) can't delete/rename anything in
I thought I gave you permission.. ;)
without me having root access :)
the ftp server would have to support chroot. also read about sticky directories (man chmod)
you cannot chroot without root permissions (or an already-existing setuid root ftp server)
ok - I looked at sticky directories, but didn't really understand
I'll read it again
/tmp is often sticky for the same reasons you want a sticky directory
I bet anon ftp is often done with chroot. I would not be surprised if ftp servers commonly support that but I don't know any details.
ok, so I guess the problem is that the way ftp is set up, you can't access any files outside your own home dir and below (a good thing)
so to get this to work, I had to create a directory owned by the "master" in the user home dir
below that are the files we want to keep safe, but also the directory name itself has to be left alone, or all the others who use this program to get their updates will fail
ok - made it work. probably a stupid way, but it works
logged in as the user, did chmod u-w on the home dir, and now the user can't even change his own stuff :)
once shell access is removed, it should be safe (since FTP can't get above the home dir)
thanks for the pointer cradek