ok - I'm mostly back as well
incoming 900 number?
a friend trying to get Windows XP to run for more than 14 seconds without getting a virus ;)
Were you thinking of writing the ini conversion program using c++
C or C++ probably
keep xp off line
actually, the problem started before he plugged in the network cable
So a line by line conversion
maybe it was a floppy or CD
ah. itunes, sony, or ms
I was thinking more that there should be a configuration file for the translator
a rule set.
with snippets of hal and/or ini code that get pasted into the output file(s) wen certain things are found in the input ini file
something like the file you built for kbuild demo
hmmm - I'm not sure about that
Paul is working toward xml for config stuff.
I thought he was opposed to that???
yup on the BDI fork
cool. we should sync up
Was till he built a demo for a LUG meet.
heh - probably was 100x easier to configure
* rayh looks for the parser he used.
I think it was libxml or something
I really like the KDE-style config management scheme
too bad the standalone library is unmaintained
short memory, switches with my kvm
yep expat is it.
He also favors using autotools nowadays.
that's the step above ./configure?
He calls it's advantage building "selective systems."
yep - better tuned configs for the specific build environment
I tried to talk us into selective builds for EMC
yep - that's one of hte things the KBuild stuff was aimed at
"don't build ppmc, because it doesn't compile, and I don't need it anyway"
the configure stuff is a big step forward.
it's amazing how you have to think "don't tick the spacebar, thumb."
eventually (maybe in the 2.4 or 2.6 timeframe), we may be able to make a new emcHALsh, plus an extended VCP-like thing, and get the last "feature" of mach off our bnacks - configurable screens
then there is vtcl with an EMC enabled lib
emc or emc2?
I think that's what I meant with emcHALsh
I made it originally for emc but it would work equally well with 2
ie, emcsh + HAL in tcl-usable form, with UI controls that are placed from a config file
One has to be careful about direct applications to hal.
witness the recent "weaving" discussion.
sure, but I see that not as a HAL problem, but that emc isn't meant for what he wants
many times better control through gcode
true, and the subroutine support makes the problem mostly moot
Yep. But because HAL is so damned accessable many view it as the first line of customization.
there is a slight breakdown between the "device driver" aspects and actual control software
btw if you put the weave generator between pid and motmod it works
only if you allow the max weave offset as ferror
no because ferror is computed from pid input
doesn't motion calculate its own ferror?
not that I see.
odd. I thought it did
You are without a doubt much closer to truth there than I am.
of course - I'm just telling you what I think ;)
I may be totally wrong about pid/motion though
it's been a while
yeah.. didn't see you much lately
busy with life & work
mostly with life
how are you doing?
I am rewriting this software again
hmm.. pretty ok
alex_joni: on my new machine I get bug 1205237
(mode not switching)
from the bug report it looks like you know the most about it; do you have any idea what we should do next to figure it out?
btw make install/make deb work now
also jepler got the same behavior on his laptop
seen the make install stuff, nice
ok, now regarding to that bug, I didn't touch it in a while, and I don't have a machine now that does it
15:05 < cradek> alex_joni: on my new machine I get bug 1205237
15:05 < cradek> (mode not switching)
15:06 < cradek> from the bug report it looks like you know the most about it;
do you have any idea what we should do next to figure it out?
15:06 < cradek> btw make install/make deb work now
15:09 < cradek> also jepler got the same behavior on his laptop
15:13 -!- rayh [email@example.com] has joined #emc-devel
cradek: how fast is your debugging process on that machine? Is that your home machine?
cradek: how often does the bug appear? (only after a lot of mode-switches, or just after a few?)
cradek: last time I've seen that ( a few months ago iirc), I had to change modes like crazy (about 50-60 times till it locked)
* alex_joni is reading the bug notes to refresh his memory
alex_joni: yes it's my home machine, and it shows up immediately/always
I can switch modes 1-2 times and then it always hangs
ok, that's GOOD
I think it's some racing condition, that ends up in having those bits set wrong
that sounds like it won't be very fun to hunt down
What gui is the bug 1205237 reported on?
it's not a gui bug
the problem is I've changed two things at once: I got a much faster computer and I switched to rtai
I've switched a couple thousand times here in the last 10 minutes and don't see anything with tkemc
so either of those might have been the cause
it happens for me with either gui after just a couple switches
it's a totally fatal bug
jepler gets the same thing on his (fast) laptop
Okay. This is bdi-4.30 with the 2.95.4 compiler
do you know which rtai version that is?
my current guess is it's related somehow to machine speed
it might be hard to find.
was away for a while welding some pipes, back now
cradek: it is related to machine speed, but so far I've only seen it on very slow machines
on a PII-233
I think our machines are around 2GHz... also maybe it's worse with rtai3.3
don't think so
you at the machine now?
or only at home?
the machines I tried are 1.4GHz (one Athlon, one Centrino), but also a 2.66GHz P4
and never seen it there
* alex_joni heads home..
the machine is at home but I have ssh to it
maybe we should compare latencies instead of processor speed
like I said, maybe rtai3.3 is faster/slower and therefore makes the problem worse
I think we can nail the problem
for any speeds/latencies
it's a flag set when it shouldn't
or a check too early
maybe the mode switch shouldn't get done while not INPOS set
I see some strange things at first glance...
but I need to study the stuff in control.c a bit more thoroughly
but, dinner first ;)
actually late lunch :D
hey cradek - what's the best place to put in AXIS feature requests?
email, or here
we don't have a feature request tracker except the wishlist (TODO) in cvs
ok - I considered a SF tracker item in the axis category
I considered that, but it's overkill
have a look at the TODO in cvs
I looked at the list of possible additions for 1.2 (I think)
so all this aside, what's the request?
I finally fired it up on the 800x600 touchscreen (though I wasn't using the touch driver)
first, a full-screen mode would be great
that's a function of your window manager.
second, the tooltip text is unreadable because it's so small (about 5 pixels high)
that's an X misconfiguration.
true - or you could make it draw on the root window
hey these are easy
pass the buck
no, that would be an abomination.
the problem is that X can't read your monitor's DDC information, so it calculates a wrong DPI.
set your monitor's size in xorg.conf
also, add several controls that allow you to change the left-drag function on the preview - zoom, pan, and rotate
I think it's set, but I can check
basically that last one makes it so you can still use all the preview features with only one button and dragging
DisplaySize 123 456
replace with your display's size in mm
what font size is that supposed to be?
let me see if I can find it
heh - it is certainly small
SWPadnos: run xdpyinfo and check for these two lines
dimensions: 1600x1200 pixels (411x311 millimeters)
resolution: 99x98 dots per inch
actually, the size was set wrong
does that fix it?
not sure yet - I'm creating a desktop shortcut :)
to save time, dontcha know
btw, in icewm (and also gnome I think) you can make a window fullscreen with Alt-F11
I see that KDE has a fullscreen option as well
we should make a good icon for emc
* alex_joni plans to do that
what size is an icon usually?
32x32 or bigger?
I remember someone drew chips2 and he was pretty good
do you remember who that was?
I expect 256 colors, or 24bit
yeah he was like chips but drawn better
he was holding an end mill in one ... flipper
huh, that might have been me
hang on, I'll give you an URL soon
fenn did one
he was wearing safety goggles
oh, then probably it was fenn's
cradek: what size icon?
I don't remember what it was
how about svg, targeting 48x48
I think it was a medium-sized image, not an icon
no, I mean what size should the icon have?
fenn had made a desktop / bootup background, I think
48x48 is the standard, I think
I only made an chips2 ages ago, but I don't like how it looks..
at least it was for NextStep ;)
[17:17:29] <alex_joni> http://www.robcon.ro/emc/chips2.jpg
maybe a chisel with a motor on it
[17:18:45] <SWPadnos> http://fenn.dyndns.org/pub/emc/chips2.png
I think that's the one I remember
I'm not sure I love either of the chips2, but this is the kind of thing I stay out of because I have no artistic skills.
my problem as well
is your font problem fixed?
my wife is a graphic designer though, so we may be able to get something done
they're readable now - thanks
still a bit small
they're supposed to be small but legible
I can't even find where the font is set
in KDE, there's a fullscreen mode, but if the application doesn't provide a way of getting out of it, then you have to use the keyboard to get through the application menus to restore to non-FS mode
it's legible if I'm within 2" of the screen
SWPadnos: what DPI does X think it is now?
alightly over 1 pixel/point
how's this? http://www.robcon.ro/emc/chips2icon.png
not bad, actually :)
ok, that was a quickie ;)
it can get better if I do it slower :)
but I need to rush out now
* alex_joni will be back later
cradek: if you can use it, be my guest
we need to package it along with emc2
so users will get the icon on the desktop after apt-get'ing
toss it in cvs and we'll figure out what make install should do with it
ok, along with the splash?
make a desktop link ro emc, probably ;)
we'll have to figure out how ubuntu handles putting things on the menu etc.
ok, will do that later ;)
is the "live" distro going to be ubuntu now?
none of us can predict the future, but I'm working on some ubuntu stuff.
"this document contains forward-looking statements, which cannot be guaranteed ... "
THIS PAGE INTENTIONALLY LEFT BLANK
this brain intentionally left blank
laters everyone.. I'm off
OK - so I guess the feature requests boil down to (1) possibly adding a button to toggle fullscreen mode (WM dependent, I suspect), and (2) adding drag mode selection for the preview window
oh - and if there's a context menu added to the G-code display, then have a button to toggle that as well (maybe click a line, then click a" menu" button)
can you explain what you mean by drag mode selection?
right now, dragging with the middle button rotates
left button pans
no option for zoom (right?)
put buttons on the toolbar or somewhere so that a left drag can perform any of those functions
I think the right button or wheel zooms
shift does something too
how about keyboard modifiers? shift+drag, control+drag?
those would let you use just the left button, but still not muddy up the user interface
that would work
I see that shift+left = middle (rotate)
so add ctrl+left=zoom, and you're all set
* control + left drag = zoom, which is the final thing needed for left-button-only navigation (SWP's touchscreen).
(assuming I can even drag with it :) )
but a touchpad would be the same
yeah, that part's up to you!
maybe I'll install the driver again
my touchpads all have three (two plus three-button emulation) buttons
for a very long time, X applications have always been able to assume three mouse buttons
yep - that's what we had on the Irises, way back when
with amazingly speedy 25 and 33MHz processors O_O
today we're near everyone assuming three buttons + wheel, but that's not the rule yet
what about those poor souls who are stuck with only a light pen?
I ran autocad on a sun 386i
not sure what the clock speed was, probably 25
oooh - the only 386 based computer that actually booted in 32-bit mode
I ran OS/2 on a 386 too
I've still got a copy of Warp somewhere
but anyway, the point is we had three buttons
os2 rocked :)
I had v3 as well, on floppies
actually I had 16
ok - a tablet with a 16-button puck
os2 only rocked when compared to windows 3.1
<os2> oh something's wrong? go wade through the 300 line config.sys </os2>
actually, it's still good compared to Windows 2000 or XP, but not necessarily in the eye candy department
lots of banks still use it, because it doesn't crash
I'm at some friends ;)
hrmm.. do you guys have any idea why hdparm is failing?
hdparm -c3 -d1 /dev/hda
setting 32-bit IO_support flag to 3
setting using_dma to 1 (on)
HDIO_SET_DMA failed: Operation not permitted
IO_support = 3 (32-bit w/sync)
using_dma = 0 (off)
the 32bit works
only DMA can't get set
is this a BDI?
no, stock debian
does hdparm work on scsi too?
not that I know of
ok.. but on IDE it should
maybe the specific chipset support wasn't compiled into the kernel?
-d Disable/enable the "using_dma" flag for this drive. This option
now works with most combinations of drives and PCI interfaces
which support DMA and which are known to the IDE driver.
that doesn't sound like a sure thing
[19:35:52] <SWPadnos> http://seclists.org/lists/linux-kernel/2002/Oct/0162.html
SWPadnos: it's a stock debian SMP kernel
I'll check the config :(
yeah - it should have support for most of the common chipsets, but who knows
it's not SATA is it?
hrmmm.. # CONFIG_IDEDISK_MULTI_MODE is not set
that's multiblock transfer, not DMA (I think)
I think that just sets DMA mode automatically, if the kernel knows it's usable
ok.. looking further
what's the chipset?
# cat /proc/ide/ide0/model
I think it's not discovered properly
but I have no idea what chipset there is :D
dmesg | grep -i chipset ?
I'm not sure how they're reported, actually
Jan 20 17:12:58 localhost kernel: ICH: IDE controller at PCI slot 00:1f.1
Jan 20 17:12:58 localhost kernel: ICH: chipset revision 2
Jan 20 17:12:58 localhost kernel: ICH: not 100%% native mode: will probe irqs later
Jan 20 17:12:58 localhost kernel: ide0: BM-DMA at 0xfff0-0xfff7, BIOS settings: hda:DMA, hdb:pio
Jan 20 17:12:58 localhost kernel: ide1: BM-DMA at 0xfff8-0xffff, BIOS settings: hdc:DMA, hdd:pio
eeeww - an Intel 865 or similar
recent P4, right?
no, dual P3 XEON
ok - the ICH threw me
# dmesg | grep -i chipset
ICH: chipset revision 2
agpgart: Detected an Intel i840 Chipset.
ah - OK
[19:48:34] <SWPadnos> http://www.ussg.iu.edu/hypermail/linux/kernel/0305.3/0055.html
I googled for i840 IDE DMA
that may not help though ;)
alex_joni: usually you just need the generic stuff there and it'll work..
alex_joni: that was the case on my laptop..
well.. it didn't
anyways, don't care right now, too tired :D
four hal variables in watch takes 956 microseconds per iteration
* alex_joni is back
rayh, is that with separate execs?
no that's with the open followed by separation of the reply and changing of the displays in a canvas.
If a person is conservative about how many variables they watch at a time it should be okay.
so that's a single HAL command, and then TCL parsing the result?
or separate commands?
No there is a halcmd for each through the open channel
that may be worse than using the TCL string functions
We talked once about sending a matrix of pins and getting back the values but that's for later I suspect
remember - halcmd internally does a string compare for every pin (or whatever) each command
and it's not likely to be better than TCL at string manipulation ;)
and then I've got a pile of parsing, ifing and setting displays.
I've a notion to commit this and let you guys have a go at it.
heh - that's not fair
all's fair in emc
syntax error in expression "": premature end of expression
never seen that before.
well - that is premature
I suspect that I'm reading a global while another is changing it.
where did it occur?
In the watch loop.
is there a way to remove watches?
while I was adding another pin.
I've seen some odd stuff with watch mode, (didn't mention it because I know it's a WIP)
I suppose that I should have the user select the pins to watch.
then kick off the loop to watch.
I'm not using the unix watch command.
and allow deletions
Have not gotten that far yet.
I ended up with 3 copies of the same pin (due to my own stupidity(
no problem - like I said, WIP
one other useful thing would be to disallow duplicates
Hey that is quite possible with the pattern.
I could build some watch control buttons and place them in the control frame.
sounds good to me
got the disallow duplicates.
cool - that was fast
I had noticed that sometimes the vars disappear fora couple of seconds
the blank line will still be there, just no text on it
the value variables or the varname?
both, I think
this is in watch mode
I think so (not running it right now)
okay. I'll watch for it.
I'll try it again
I've got a more complete error message.
syntax error in expression "": premature end of expression
(procedure "watchLoop" line 15)
invoked from within
ok - so $ret is empty
sounds like it.
you can check for that before the expr
'suppose i could just catch it.
That is where the value of a non bit var is converted from scientific notation to decimal
I'm getting a few oddities in the selection between bit and other also.
ok. I just cvs upped, and I'm taking a quick gander at it
Must try buttons to stop the loop while adding.
I've got the blinky lines again, btw
what all is blinky
ok - maybe it's just some extra blank lines getting added or something
in watch mode
let me start over, and tell you exactly what I do to cause it
yes, in watch mode
ok, ran halconfig, waited for tree to refresh
this is with the univstep config
so lots of hal stuff around
expand the tree as follows: first pins, then ppmc, then 0, then encoder
click 00 (which should give me a list, I think, but it adds encoder.00.count)
watch screen says "ppmc.0.encoder.00.counthal_ppmc -1"
click index, delta, and position - no change
(clicking in the tree)
click count, and a blank line is inserted above the existing line in the watch pane
this toggles back and forth a couple of times (top line, then second line ...)
ok - simpler method. click 00 a couple of times - you'll see the watch pane change in odd and interesting ways (unless the dupe elimination fixed this)
I don't seem to be able to duplicate your display here.
I wonder if my wife's finished snowblowing the driveway
and you're sitting inside where it's warm
at least I started it for her
I'm sick, you see ;)
server her right - she gave me the cold in the first place
ah there appears to be a problem with the tree encoder.00 is a node rather than a leaf.
ok - no blinky here, plus I have nice little fake LEDs
that happens in a couple of places, I think
so it returns at least three or four lines.
ok - that could be
hmmm - I should change the index output to a bit
ok - it is a bit - you need to fix halconfig ;)
I'll try using the channel flag to prevent these partial reads.
is it a partial read during the tree refresh?
No it looks like some sort of tree building error. I see the raw reply split out okay.
It just does not build the tree right.
ah I do have five levels of nodes but a wrong index to the last.
is that a fixed 5 levels, or can it be more?
aha 'dat's better 'eh.
that's a fixed five levels.
* SWPadnos waits for cvs message ;)
would CL want more?
hmmm. probably not, since there's only one instance of CL
we could look at the stuff in demo_step_cl
is there an initial delay before the first tree refresh?
ah there is an issue with demo_step_cl. No stepgen max vel.
The tree should be there when the display comes up.
I did eliminate the open all nodes thing.
ok - it takes around 5 seconds before the pluses appear
on my celeron 500
it looked like it might be a 5-second delay plus a little, but it could just be that slow
* rayh makes a note to try on his slowest box.
heh - fast computers are the bane of developers
"it seems fine here"
Try adding a sig and see how long it takes to display the new.
adding a sig to the watch?
No to the show.
newsig MySig1 bit in the lower entry widget
as fast as I can click
oh - hold on
about 5 seconds
Okay. I'll have to change the way refresh works. Just add or subtract rather than a total rebuild.
But that will be later.
but the signal doesn't show up in the tree
hey - when did we get 8 axes?
bummer - the Tapmatic 70TC/DC I'm watching on eBay is up to $145
still a good deal though
we always get 8.
odd - I thought it was 6
It would be nice to limit to the ini variable's number
that would clean up stuff a lot.
like - read the ini to figure out how many axes to export?
out of curiosity, do you have any idea where to get nichrome wire?
No I don't.
I'd think that would pop balloons pretty well - don't you?
should. I know they use it round here for cutting foam and bending plastic countertop.
yep - I found some places that sell foamcutters, but they have no specs on the wire (like ohms / ft, or current capacity)
I built a veeneer trimmer with some.
how much current id it take? ;)
Seems like about 3 amps
Nice red glow.
what gauge, if you recall?
I don't remember. I do remember doing some testing.
But the guy I was working for got the wire.
heh, OK. maybe I'll order a small spool
Don't see it in McMaster.
there are places that sell Inconel wire as well
well whadda ya know: http://www.wiretron.com/nicrdat.html
lots of data on nichrome
You guys and those damn fast connections.
maybe we can all chip in for a satellite dish and a 600' mast for you
That should do it.
will that get you above the horizon for the geoxync satellites?
Sure. I'm a bit above 46, i think.
ok - about a degree north of me (44 75-ish)
that made no sense
all the variables are coming back 1.
need to do a bit of checking
oops - got to get to a meeting. see you later
SWPadnos is now known as SWP_Away
rayh is now known as rayh_away