jepler: "save gcode as..." gives me a traceback
File "/usr/bin/axis", line 2911, in save_gcode
except (shutil.error, os.error), detail:
AttributeError: 'module' object has no attribute 'error'
oh ... in an unwritable directory
probably should be (shutil.error, IOError)
good catch; I'll test and fix here
there's also a little debug output when it works
I see that
maybe whether it works or not?
whee, optional stop and block delete work
testing is fun when stuff actually works
DTG units are right...
halshow doesn't work
Error in startup script: couldn't load file "/usr/share/emc/tcl/hal.so": /usr/share/emc/tcl/hal.so: cannot open shared object file: No such file or directory
probably missed by 'make install'?
that seems like a wrong path...?
yes the path seems fishy too
tcl's hal.so isn't in 'make install' or the debian package
but in addition to that we need to discriminate between a 'lib' and a 'share' tcl dir, since it now has an arch-specific file in it
any idea why comp.1 and motion.1 are listed in .cvsignore? I think I added them, but that makes no sense
what was I thinking?
umm sorry, no idea
are they generated from something else?
jmkasunich_: no, I am pretty sure they're not
cradek: I'm working on fixing the hal.so problem as well
I'm continuing to test
shift-home (touch off to zero) fails with a traceback
6,7,8 (u,v,w) are missing in the key reference
if I try to home with a shared switch closed, I get the error twice, but some things do move/home anyway
seems like it ought to not do anything
(I was using home all)
doubleclicking on the item to probe in halmeter does nothing
I thought I had fixed that at Fest
or maybe I was working on keyboard navigation only
the keyboard navigation sort-of works
I think I can select, but I can't see what I am typing because it doesn't scroll to that spot
I think the last I knew, we could have that bug, or some other focus-related one plus the ever-shrinking horizontal scrollbar-handle
so I guess I chose to have only one bug, annoying though it is, until I (or someone) could figure out how to get rid of it
should I put this stuff in TODO?
what - "stop using UI libraries with annoying bugs"? :)
I'm working on shift-home and uvw keyboard shortcuts now
SWPadnos: I also want keyboard search very much, but I think for now I'd trade the half-functional search for a doubleclick like halscope has (if that's what broke doubleclick)
I don't think it is, thinking about it more
I remembered that I was working on it at Fest, but didn't remember exxactly in what way
I have exactly the same (fuzzy) memory
I'm pretty sure now (without looking at the logs) that I didn't change double-click behavior
maybe it has always not worked...
that may be true
(too bad halmeter/halscope don't share those common bits)
what is double-click supposed to do?
yeah. that was a bit too much to do at Fest (for me at least)
you can single-click and hit ALT-A though - I made the accelerators work
select a signal in halmeter?
fenn: probably select that thing and exit the chooser
or, I'd be happy with 'apply' behavior instead
but it should do _something_
probably the "OK" behavior is the most stnadard
OK that's all I'll be fixing tonight
see you folks around
cradek: thanks for doing the unglamorous task of looking for the bugs
you could probably add "double-clicked" to the list of events used
oh hey, turning on the brake actually turns off the spindle
that's a nice feature
but what if you're going around a curve and you want to shift into low gear quickly
then you ought to get back on your medication
you brake before the curve, and accelerate out of the curve
hmmm. it appears there's no specific double-click event for a GTK_CLIST
so I think it would be a PITA to get double-click working
does halscope do something strange?
oh maybe it works on single click
(since there's only one operation)
it does single click
halmeter has ok vs apply
halscope selects on the single click
ok that's bogus too then
halmeter actually updates the meter when you apply, whereas halscope does it after the dialog is dismissed
what's wrong with single-click for halmeter?
halscope probably needs that since it has to tell the RT portion what to sample
fenn: ok vs apply
the idea behind not dismissing the dialog immediately is that you might want to look at a few signals for a second
click on the item and it applies?
so you click / apply on each one
it doesn't at the moment, but that may be a good idea
if you can live without "cancel"
could add a cancel too
and get rid of apply
halmeter has cancel
but you can't cancel if single-clicking applies and dismisses the dialog
i think the idea is so you can look at a bunch of things quickly without having to open the dialog each time
it would be possible to apply if the selected row is clicked again
why not just apply it the first time?
because that eliminates the possibility of canceling
why does it?
apply doesnt close the dialog
unless you start keeping track of what was being watched before the dialog was popped up
well, cancel doesnt do anything now anyway
"apply" changes the name of the signal that the main meter window monitors
sure it does. if you click on a new signal, then click cancel, it doesn't change the signal it's watching
cancel doesnt do anything
it just closes the select dialog
cancel was a way to dismiss the dialog without changing what you are viewing
if you've clicked on a line in the list, I see nothing wrong with immediately switching the display to that item
it allows you to not change what's being monitored. the way it does that is to do nothing, rather than changing the monitored item
single-click select would certainly simplify that dialog - you'd only need a "close" button
cancel is for "oh crap, I didn't mean to open that dialog", not "oh crap, I didn't mean to open that dialog and I also didn't mean to click on an item in it"
open the dialog, click, you have the new channel
click close to dismiss it
and for convenience, open the dialog, and double click on an item to select that item and dismiss the dialog
we cant figure out how to double-click (well, swp can't :))
except there's no CLIST double-click event :(
you can do it, it's just a PITA
oh well, forget the double click part
single click to select, with a close button
you can probably access the parent class double-click event, or just apply if the selected row is clicked
you need a timer for double-click
if (ev->type == GDK_2BUTTON_PRESS) looks like its supposed to work
I guess that would be the parent (or grandparent ...) class double-click event ;)
you do that in the click handler
there isn't a specific click handler in the seelction dialog code
jmkasunich_: if the XZ home switch is closed, UVW still home on 'home all'; is it a feature?
HOME_IS_SHARED isn't set (to 1 or 0) for axes 1, 4, 5
surely it defaults to 0
sequential homing was added by jeff, I haven't used it yet and would have to read docs to see what it is supposed to do
it is set to 0 for 6,7,8, so I noticed :)
I think it may be a feature
I think it's not obviously wrong, but it was surprising
I wish I knew what I did to get double errors though
UVW are independent of the XZ switch, so XZ being homed has no effect on the UVW homing sewuence
s/has no effect/should have no effect/
it seems likely the error message could be better
oh - I didn't see an error
"cannot home ..." while it's homing some stuff right behind that dialog looks wrong
"cannot home while shared home switch is closed [OK]"
and then it proceeds on to YABC doesn't it?
not sure what it's doing
or does it wait for the dialog to be dismissed?
sometimes I get 2 errors, but everything ends up homed anyway
no it doesn't wait
I'm starting to think we should have all or nothing
OT silly question: is there a way (in bash) to negate the sense of an if, other than sticking the desired action in an else clause?
(fwiw, this config is axis_9axis in very recent trunk)
(I didn't see a "not" command)
you mean negate some test in  ?
this is so I can start realtime if it's not running, so I'm just looking at the "if $REALTIME > /dev/null" clause in halrun
no, it's not a 
yes you're just supposed to know that [ is spelled "test"
I did know that, acutally
oh, testing an exit code
that's a bit harder, I don't know the answer
err - thanks :)
maybe I'll just do the else thing for now
I prefer $REALTIME || otherwise
but that only works for a simple one-command 'otherwise'
can you $REALTIME || do blah blah ; done
$R || (a; b)
but that's getting stupid
chris@rover:~$ false || (echo hello; echo bye)
it would be better to know how to do if !$REALTIME
I guess test should do that though
chris@rover:~$ if ! false; then echo hi; fi
when in doubt, try the obvious thing
if only the bash man pages were easily searchable
without finding 17000 extra bits
yeah manpages get less useful as they get bigger
it says ! is a reserved word, for whatever that's worth
If the reserved word ! precedes a pipeline, the exit status of
that pipeline is the logical negation of the exit status
neat, I didn't know this
it works everywhere, it's not special `if' behavior
chris@rover:~$ ! false; echo $?
I get "bash: !": event not found
I probably need to excape my strings better
put a space after the !
I was trying to echo real! or unreal!
with "" around them
I probably needed a \ also
use single quotes if you don't want the shell to mess with anything
double quotes allow several kinds of expansion
quoting rules could be a semester-long class, I think
(if sub-shells and remote logins ... are part of it)
yeah no kidding
and don't forget ````m4''''
I'm sorry, what was that? I forgot already
the first week would be getting all the whiners to use a font that lets them tell ` from '
I don't even know how to type the upside-down quotes
hey look at the time
I have to go to work tomorrow and relax. I better rest up.
I have to leave for the airport at 6am
jmkasunich_: the HGR guy called back and quoted me $250 shipping! haha
why didn't he just say "f__k you" while he was at it?
he could at least buy me dinner first
he was probably pissed off about the red sox
so, I'll look around some more :-)
not like there's any rush
I just would like to be rid of the noisemaker someday
rotary phase converter?
what kind of transformer(s) are in the machine now?
maybe some creative rewiring can be done to make it work on single phase?
the main one (DC bus) is a 3 phase deal, then there are a half dozen other single phase
so it will take one new transfomer AND creative rewiring
all the others are small 115v or 24v out, 240 single phase in
those are a piece of cake
I've already shuffled some around so all the electronics are on the good phase
but that main one needs to be replaced, and it's big
three core legs, and windings on all three?
6 diodes and some caps
you'll likely need more caps - a 3 phase bridge has much less ripple current than single phase
yeah I had figured on that too
why do you need such a huge transformer?
if they're original (probably) it wouldn't hurt to replace them anyway
I'm wondering if there isn't a way to use two legs of the three phase transformer on single phase
jmkasunich_: not enough iron is there?
gotta think about that a bit
there are toroidal transformers of all sizes on ebay, reasonably priced i think
fenn: the books say it wants about 3kva
if reasonable is $150/kva
toroids only seem to go up to 1kva unless there's a style I haven't seen
the transformer at HGR is only $40 for 3KVA
I think mine is 2 kVA
oh, well that makes a bit of difference
the boss control won't ever rapid all 3 axes together - it may even be a bit underpowered
there are two issues, droop and transformer heating
the latter is a non-issue, unless you are gonna write a program that does nothing but three axis rapids all day
for the former, I'd be astonished if an industrial transformer (like the HGR one) droops by more than 10% under rated load
I won't worry so much. I bet it could get down pretty low without faulting anyway.
tissf_ is now known as tissf
tissf: you should go ahead and "cvs commit" the axis_fr.lyx file with the improvement to the index.
Yes I will try, but I had to restore my machine and I lost my. ssh: (
the accent problem is resolved
All works well except that the compiler asks a xref.html
tissf: re: ssh, if you don't have the private key anymore, you need to generate a new keypair, and send the public key to cradek
[12:50:26] <alex_joni> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?DeveloperAccess
tissf: I can also get the 'xref.html' error now; I'll fix it
ah ok :)
before commit, I will await a little to know if I organize the translation under gutsy or under dapper
I think it would be best to work on dapper. otherwise you have to go through the extra "lyx2lyx" step before each commit.
yes but on this PC I have just feisty and gutsy, I must remake a partition for dapper
The PC with dapper is at the workshop and it does too cold:)
jepler It is impossible to have lyx2lyx -- à 221 automatic one with the compilation?
Guest998 is now known as skunkworks_
tissf: no, because on ubuntu dapper, lyx2lyx can't read the newer files
it has to be done on the system where the newer version of lyx is installed, not where the older version is installed
ah ok thanks
I will install dapper on a disc usb that I easily will be able to transport
I sent my ssh to Chris by mail
I have again the following probleme
unable to parse objects/xref.xml
make: *** [../docs/html/xref.html] Erreur 6
tissf: did you cvs up?
can you tell me how you fix ?
jepler>tissf: I can also get the 'xref.html' error now; I'll fix it
no at this time cvs lgin dont work ... I have to read the doc before !!!
or can you please tell me the syntax for commit ?
I think I saw a commit message from jepler fixing the xref problem
so you need to do 'cvs up -dP' in your dir to get the latest changes
if you want to commit something the syntax is : cvs commit -m "message"
ok for cvs update no problem but for commit ?
in order to commit you need to have checked out the sources with your username
cvs co -d:ext:tissf:...
if you are working on an anon-checked out repository you won't be able to commit, as it's readonly
so the best would be to check out the sources again (with your username), and move the changes over, then commit
(test, then commit :)
ah ok ok thank you !)
no problem :)
I probably did not use cvs commit for 15 years!
well.. I only started using it 2-3 years ago :)
cvs [checkout aborted]: must specify at least one module or directory
cvs co -d:ext:firstname.lastname@example.org/cvs co -d emc2.TRUNK emc2
It is necessary to indicate the branch?
(-d emc2.TRUNK means it will create a directory called emc2.TRUNK)
I usually create more than one dir (emc2.0, emc2.1, emc2.TRUNK, etc)
cvs checkout: No CVSROOT specified! Please use the `-d' option
can you tell me the command you tried?
maybe try : export CVS_RSH=ssh;
and in what directory
cvs co -d:ext:email@example.com/cvs co -d emc2.TRUNK emc2
"-d:ext..." must go before "co"
oops.. my bad
cvs -d .. co -d emc2.TRUNK emc2
confusingly, the two -d mean different things
that load :)
After loadin I place my files translate and I commit?
you need to 'cvs add' if the file didn't exist
then 'cvs commit' to finalize the commit
Ok, fr_axis.po commited, now for cvs add I have to add the path ?
no, you go to the path where the file will be
and cvs add filename
(you need to create/copy the file first)
Thank for the help
tissf: no problem
can you tell me if *_fr.lyx compile under dapper ?
hang on ?
wait a moment :)
Time to eat here, soon
(I have some packages missing on this machine.. dvipng for now)
there's at least one problem, but it's one I need to fix: UnicodeEncodeError: 'ascii' codec can't encode character u'\xf4' in position 3857: ordinal not in range(128)
I think it's because there are non-ascii characters in an index entry, and I'm not handling that properly
after that fix, the html documentation builds .. let me see if it looks OK
yes looks good including the index
I can't check the PDF documentation right now, however
jepler: did you commit the fix?
(nm.. seen the P ..)
tissf: the pdf seems ok over here
alex_joni: yes, I did - <cia> must be lazy today
How you doing?
jepler: a few things on this page http://www.linuxcnc.org/docs/devel/html/gui_axis.html
are incorrectly center justified
cradek: yeah I noticed that too
I probably don't know HTML well enough
it's because I want the figure itself centered, but that's spilled over to center the text inside the figure as well
every word of our documentation should be read by one of us ... but man what a task
maybe we could find a way to split it up.
whenever I start, I get through ten pages and then quit
well.. I think I read it twice so far
but then again, that doesn't mean much :)
rayh: pretty good, a bit busy
I can see that.
alex_joni, Roland was wondering if you were going to make it to the cnc workshop this year?
yeah, I was wondering too :D
June 16-21 is the date.
ok, that's good to know
* alex_joni puts it on the TODO list
you better come
We planning for another EMC get together there?
* skunkworks_ hopes so..
it's always hard for me to see so far in the future, but I'll probably be there if anyone else is
I sure look forward to spending a couple days with you guys
rayh: do you think you would be able to come this time?
I sure hope so.
Doing pretty well these days.
Good to hear. :)
Roland wants me to take a couple days and work with him on the Mazak before the new year.
I'll need to find out where we left it after the last workshop.
My ISP has had a real problem with email for the last three days.
I think we were contemplating putting the head lifter under EMC control
Were you thinking of using the stop switches or what?
dunno, actually ;)
I think jmk had a couple of things he wanted to do, but I don't recall what they are
I think there are contactors already in place and working that move the head - but they're not hooked to that knob on the front panel
one thing I know he wanted, but which we didn't do, was make the "tool prepare" happen before the "tool load", rather than having both happen when the tool finally could be loaded
not sure what emc should do with the head actually
(iocontrol needs some fairly extensive modifications for this to work right in a number of strange circumstances, so I put it off)
rayh: the only trick to the tool changer is to unload the tool before exiting emc (T0 M6). it works great then (puts the tools back where it got them)
oh yeah that's a bit of an annoyance
one night I think chris, steve, and john spent an hour trying to recover from a mistake where emc's idea of the tool-in-spindle didn't match reality
I wish it had a sensor there, but it doesn't
not sure what else we can do without a sensor
rayh: you might want to take some american ER40 collets with you when you go... (at least I think they're ER40)
we had virtually no tooling that worked
I think that Roland has some now.
maybe a procedure needs to be devised and written down, or maybe something can be done on the vcp to make the needed operations easy
I'll confirm that
(for fixing up after a tool change problem or mistake)
EMC was waking up thinking tool 0 was in the spindle?
jepler: I think the procedure is yank out whatever is in pocket 0
Where would the "sensor" be.
yes, emc always assumes no tool is in the spindle at start -- that's the same as "tool zero"
yes when EMC starts it thinks "tool 0" is in the spindle
then to load something real, it "puts tool 0 away" into pocket 0 which must be empty
Most of the control's that I've worked with did have a vcp like manual tool operations page.
I can see how it would be good to be able to change pocket number and such.
as well as manually rotate the change arm and intermediate arm.
hmmm. I thought we did get tool-prep working before load
maybe I'm thinking of prepping the "previous tool" pocket while the change is taking place
yeah that works
it all works right, except some prep work could be done ahead of time, and it isn't
Were you able to get it to rotate the carousel to the next tool?
oh sure, toolchanges work fine
rayh: yes Tx M6 does load tool x and put the previous tool away in the right place
they may not be time optimized though
the video on linuxcnc.org (and youtube, I think) shows a full drill, countersink, and rigid tap cycle, including tool changes
If I remember, the tool prepare NML is issued as soon as the last tool change has completed.
yeah and you can laugh at our tooling too!
I saw that video.
I also saw the tapped material.
heh - so you did :)
oh right - I think I played it for you on the camera
did you get plastic or aluminum? I forget
I think I have the tapped piece though
I have the first aluminum block with one tapped hole in the center :-)
(I have a few EMC trophies...)
Perhaps we should send that to the Smithsonian!
I'm sure they get many stranger things.
I just imagine.
interesting - looks like we had 2 collet chucks
(I'm watching the video)
lucky guy. That would take a week of download here.
and I can't tell what that jacobs chuck is held in...
it sure is long though
I'm seeing only one update to the demo_mazak config from that date.
I wonder if the final config was in the machine only?
[21:17:13] <cradek> http://cvs.linuxcnc.org/cvs/emc2/configs/demo_mazak/demo_mazak.clp?graph=1
the 2007-06-16 must be the latest
I'm positive we would have checked in the final config.
I'll try a cvs up here and see what I get.
+# Yes really, this .0024 backlash is right, workshop 2007
a little bit of a non-sequitur: does anyone know if EMC2/RTAPI/RTAI ... will run if the root filesystem is read-only?
Good. Thal should help.
SWPadnos: that depends on whether everything it needs to be writable is writable...
sure - I'm talking about mounting / ro
sure I understand
... it depends what's on /
I'm trying to kindly say your question is useless
... and I just failed :-)
I could have said "if all filesystems are mounted ro"
a ro root does not imply that there are no writable areas
surely some things need to be writable
I guess /tmp and /var would need to be writable
the idea is to allow someone to power off without shutting down, and not have an fsck the next boot
some systems start with /tmp and /var in RAM disk.
sure seems like you could remount "read only" after the application is started ..
yeah. I think that's where I'll end up
a setup like that is possible but will be a lot of work
loadusr mount -o remount,ro /
you want some stuff written: var file, tool table, gcode...
I tried that (at a shell prompt), but got the dreaded "/ is busy" message
cradek: his app is not emc
true - it's not EMC, it's just HAL/RTAPI
but ... he says EMC2 above in his question
you're right, he did
though the question is also applicable to EMC2, as an appliance
* cradek waves his hands and talks to himself
so he did
you could try figuring out how to activate the kernel "emergency remount read-only" which you can do from the console with alt-sysrq-u
when I've wanted this kind of thing in the past for an appliance, I ran my own replacement for init that did only what I wanted
jepler, interesting idea
that's very low level hackery that you probably don't want in this case
cradek, well, I'd still like to be able to log in to make changes, so some init is needed
I'll make a wiki page with the changes I've done to rc2.d to strip it down
(once I'm happy with them)
the bridgeport detects power failure and shuts itself down nicely in the brief moment it has before the caps run out
I did make a "halapp" script in /etc/init.d, but it needs to be pretty specific - I should make it read the hal start and stop script names from a file, or require them to be named something specific
I considered that today as well
but even a basic "sync" can take seconds and seconds
I think this will work anyway - there's only one application running, and it doesn't write (except for logs_
the disk is an SSD, so there won't be any head crashes
it's just an annoyance that it takes an extra minute to start up since it fscks then reboots
static void sysrq_handle_mountro(int key, struct pt_regs *pt_regs,
struct tty_struct *tty)
just call this from a custom kernel module
so there's no journald running
that's going to bite your ass one of these times
hence the ro question ;)
I think I'll get intimately acquainted with lsof options in the next day or two
my appliance app puts the (very few) things that have to be on the fs in /dev/shm
on a writable fs, you mean?
it's some kind of magic ramdisk
it has a few of its own state files that don't need to be saved across runs
good night all
night alex. say hello to the family.