jmkasunich: still around? cradek was saying he wanted to talk to you.
re the FP thing - Linux is 100% right
when he says theres no place for FP in the kernel
but FP in kernel space is a different thing entirely ;-)
emc isn't a kernel, its an app that happens to run in kernel space
jmkasunich: obviously I wasn't changing anything about what we do in the kernel, just getting rid of the bogus inclusion of userspace files.
and thanks for doing that
I said what I said because I saw there was an old rtapi_math.h and it was removed with a note about how linus doesn't like FP in the kernel
removed by you-know-who
I wasn't going to say it
today I made my emc2 configuration website work again with 2.0 / HEAD
jmkasunich: I think I've got all the fest bugfixes in the release branch
cradek: I'd like you to generate a configuration file for max using it
but there's a but that I bet is a one-liner that you could fix, can you look at it?
jepler: I'll sure do that
I can look
jmkasunich: after you get a FE and you turn the machine back on, the position jumps, often causing another FE
I was just starting to try actual tests on the farm machines
I think there was a tracker but you marked it fixed - it's not
I bet free mode works ;-)
the entire mode switching code and the TP/controller interface are kinda messy
coord mode I'm sure
when it faults it goes into machine off mode (not even sure at the moment what thats called in the motion controller)
but in that mode the command is supposed to track the feedback
so that when you go back to running mode there is no problem
but I don't know how to make the TP do that
meanwhile when I started testing I found another "bug"
halcmd uses libnml
oops, I just realized what that is
the ini file reference stuff
I don't want hal stuff to have any dependency on emc unique stuff like libnml
but the ini file reader for some strange reason is part of libnml
well where else you gonna stick it
and now I remember that I even put a #define in there to let you disable ini file refs for that reason
it sure as heck ain't NML related
same as posemath - that has nothing to do with NML either
anyway, I digress, I realized my "but" isn't, so forget it
so I still don't know what to do about this FE bug
actually, although the dependency on libnml isn't a bug, the fact that the BDI-2.2 box can't find libnml.so might be
ok, one thing at a time, lets look at FE
what are you doing to make the FE happen?
(which config, etc)
well there was a bug but I fixed it
ok, that was easy
you could set your stepgen limit too low
I thought it was a stepper thing but I saw jon E's BP do it too
I'm going to go get my computer from the car... I'm on a borrowed one and pretty helpless
still haven't finished unloading? wow
you are so slow, i finished hours ago
ahh much better
hmm, I can't seem to reproduce the problem
I set the stepgen max vel to 0.4 instead of 1.4
a G1 move faster then F24 faults, but I can hit F2 and its happy
does the position change when you hit f2?
don't think so, doublechecking
that test must be invalid for some reason
what _exactly_ are you running to make it happen? which config, which ini changes, and what command/.ngc file to make it happen?
I'm doing MDI here
there was a bug in TP that caused FEs but I fixed it earlier today
now I'm getting confused
so I can't answer that unless you want to go back to yesterday.
I thought you wanted me to fix a bug
yes I do
btw, I'm running old code, I was away Wed pm and all day yesterday, haven't done a cvs up since Tuesday at the latest
ok update to yesterday
cvs up -rv2_0_branch -D '24 hours ago'
then you can see it exactly how I did
* jmkasunich makes fresh checkout
ok I'm working on a checkout too
sorry you want 12 hours ago
probably just a couple differences
is the branch tag sticky?
cause I got a crapload of differences (files with P)
cvs stat tp.c
it should be one version old
john@ke-main-ubuntu:~/emcdev/fetest$ cvs stat src/emc/kinematics/tp.c
File: tp.c Status: Up-to-date
Working revision: 1.54
Repository revision: 1.54 /cvs/emc2/src/emc/kinematics/tp.c,v
Sticky Tag: (none)
Sticky Date: 2006.05.26.12.34.08
Sticky Options: (none)
that's not the branch
don't think the branch stuck
cvs up -r v2_0_branch -D '12 hours ago'
thats what I'm doing
john@ke-main-ubuntu:~/emcdev/fetest$ cvs stat src/emc/kinematics/tp.c
File: tp.c Status: Needs Patch
Working revision: 184.108.40.206
Repository revision: 220.127.116.11 /cvs/emc2/src/emc/kinematics/tp.c,v
Sticky Tag: v2_0_branch (branch: 1.43.2)
Sticky Date: (none)
Sticky Options: (none)
thats more betterer
oh good I reproduced it
(I'm running, not reproducing)
I'm using stepper/stepper_inch.ini
only change I made was to switch it to axis
run it and hit escape when you get to about 5:00 on the circle
no change to the ini file accel or vel or anything?
when you hit escape you will get a FE more often than not
notice the position, hit f2, you'll often get a position change and sometimes another FE
if I say "maybe its an axis bug" will you yell at me ;-)
I wouldn't yell but you would be wrong
farg, gotta screw around making axis now
you don't need to use axis
ln -s <myaxischeckout> <thischeckout>/src/axis
will that work?
I think so
(followed by a configure and make)
just run make again
err right configure and make
ok, got the same thing
what exactly does "esc" do? abort?
I'm sure that also happens when servos go wrong like jon's
yes it sends an abort
but I think it has nothing to do with abort - we're just triggering the FE that way.
is that the bug you fixed?
yes abort->FE is what I fixed
how'd you fix it? enfore accel limits as it comes to a stop?
yes it was just a bug (aborting during a blend)
spiral triggers it easily because pretty much the whole program is blended
but it's FE->machine on->position jump->FE that I'm asking you about
is that why the first fe is intermittent - sometimes I hit exc when its not blending?
yes I think so
I'm gonna try setting maxvel to 0.4 instead of 1.4 and use MDI
(though I tried that already)
maybe thats a waste of time
I think it probably doesn't matter how you get the FE
but the worse it is, the better your chances of getting a big jump at machine-on
it has to matter a little
for instance an FE when jogging won't do it I bet
I think jon E got this behavior when he got FE during homing
I wouldn't bet money on it but I think I remember correctly
dang, I wish you were sure
if it happened during homing, that means the free mode planner was in use
let me try some other triggers
the current example is with the coord planner
if they both do the same, the bug probably isn't in either one
ok I get it while jogging
ok, what did you have to do?
[00:57:34] <cradek> http://timeguy.com/cradek-files/emc/stepper_inch.ini
just jog in +X for a while until it ferrors
about 5 inches
5" at ultraslow?
no turn it up all the way
turn what up ;-)
I guess I should update my axis?
the jog slider that would be there if you would update
or use another gui
stand by ;-)
ok this shows it nicely, I get 3-4 in a row
just got one while jogging
ok try to turn machine back on
this is a good thing, because I actually understand the free mode planner
I did, and saw the jump
not big, but apparently enough
I get 3-4 FEs in a row
it doesn't jump, it ramps
following error goes from 0 to 0.010" over 15mS
0.010" in 0.015 sec = 0.6 inches/sec
thats fine, but the accel is messed up
it goes to that speed pretty much instantly
why does it want to move at all?
the code at control.c line 1728 is supposed to handle that
that code DOES sometimes work, I've seen jepler draw on the axis screen using his etch-a-sketch with the drives turned off
(so just turning the encoders)
ok, thats freaky
the position _command_ is fine, the feedback is moving
jon was using his board, not stepgen?
hmm, something is really screwy here
I'm scopeing Xpos-cmd and Xpos-fb
cmd isn't moving (it shouldn't)
that is strange
even tho Xstep shows only two steps
the fb should change in steps aligned with the step pulses
unless Xpos-fb isn't what I think it is
gotta check the hal file
jmkasunich: Xpos-fb will only be updated every servo cycle, right?
duh, can't trust step pulses when you're sampling at the servo rate
ok, its actually making a lot of step pulses as soon as you restart
I think the stepgen gets disabled when the ferror happens
there might be "stored up" motion in there that gets release when its re-enabled
would that explain servos doing it too?
only if the PID had a similar issue
yeah, stepgen starts pumpin out the steps as soon as its enabled
starting at full speed?
probably starting at the speed it had when it stopped
I think I see the prob (at least for stepgen)
that would sure explain it
src/hal/components/stepgen.c line 787
if its disabled I just bail out
probably should be zeroing some internal var there ;-)
normally, it only gets disabled when you are already at zero speed
but not on ferror
that doesn't explain Jon's problem, I'm much more certain that pid.c doesn't do that
should it reset *(pid->error) if not enabled too?
(although if you are willing to take a look at it while I fix this maybe you can spot something)
well I only sort-of understand it
I wouldn't reset pid->error
the error calc can be done even when you're not enabled, you just don't want to generate output
I wonder if we can reproduce it with servo-sim
give it a shot
it will take me a few minutes to fix stepgen, test, and commit
I can't seem to get it in servo-sim
hmm, I really fscked up the disable in stepgen
it also bails out of the feedback code
it should probably always run that, so position is always up to date
I'm pretty sure servo-sim does NOT have the problem
could it also be somewhere in ppmc?
I think :-/
maybe, don't recall how its enable/disable stuff works
does axis display commanded position or feedback position?
I think commanded
(tkemc has button to choose)
well maybe you can choose
I no longer get FE when I hit F2 after an FE
the cone is at commanded, but you can choose for the readouts
I do get FE if I stop a fast jog
I don't have to wait for 5" of travel, I can just jog a half inch and then let go if the jog speed is at 71.6 or 72 upm
that's with my broken ini right?
broken on purpose, I mean
I was getting worried
oh I was confused about what you were saying
maybe I should still be worried
the problem you asked me to fix (FE when re-enableing) is gone
but I think its odd that I can get an FE just by letting go of the jog button
dunno if thats because of the ini file, or a headroom problem somewhere
I think I decreased the stepgen accel max
nope, just the vel
1.1 instead of 1.4
regular maxvel is 1.2
so it gets behind during the jog
did you scope it yet? I can't picture what it must be doing
just figured it out
ferror is 0.5, minferror is 0.010
so when you're moving fast, you need to get very behind to trip
when you stop, the error tolarnce gets tiny and it trips
I remember doing that on purpose now
I wanted the little jump to easily trigger it
so I wanted it to get way behind on the jog
that 0.5 is not in the standard config?
well, I'm inclined to call the stepgen problem fixed
gonna copy it over to my head checkout and commit it
cool, thanks for fixing it
then look at the ppmc driver a little
are you confident about putting it in the branch?
I want to release this weekend and I think everything is good in there
I'll be more confident if a couple people can test it, or look at the changes, or both
this should fix http://sourceforge.net/tracker/index.php?func=detail&aid=1390123&group_id=6744&atid=106744
I was never able to properly replicate that one before
SkunkWorks: did you have steppers?
I should have posted my ini file at the time.
thanks again jmk
* jmkasunich takes a quick look at freqgen, wonder if it has similar issues.
heh, guess not since it has no enable pin
wonder if it should?
something to worry about later
SkunkWorks, are you running CVS head?
I was hoping that _if_ you were running head, and _if_ the problem was still happening, you could update and see if it gets fixed
(after I commit, which will be soon)
Maybe tomorrow - but no promices - I will it I get the time.
don't get me wrong - Im excited :)
I'm sorry it took so long to get around to fixing it
I need to try harder to replicate things like that
It didn't happen often - normally I use abort. not estop
cradek: the ppmc handles enable in a completely different (and robust) way
it calculates everything regardless of the state of enable
so thats not the problem that Jon had
there were actually 2 trackers reporting the FE problem
both stepper you think?
hard to tell, they didn't say
but I bet 90% of the user base is stepper
so chances are good
I marked both pending and asked them to reopen if the problem doesn't go away after 2.0.1
on the subject of 2.0.1....
I want to investigate why it won't run on BDI-2.2 (thats the only slot I tried)
it appears to compile libnml.so, but it ldd can't find it at runtime
that may only be the first of dozens of problems, we haven't actually tested anything other than compile on those platforms for ages
if there are too many, we can punt
if you don't build --rip it won't find that lib
but I want to at least try
I'm pretty sure I do build rip, but....
duh, I don
I don't do rip
good, that probably explains it
its time for bed now, tomorrow I'll do a rip build and test
thanks for fixing the FE thing
I'm going to put it in 2.0.1
thanks for prodding me, I should have tried harder to replicate it when the tracker(s) were posted
just doing my job
being the release guy is a surprising amount of work
I'm actually kind of surprised
yes, surprised even
maybe I should go to bed too
do you want me to apply the fix to the 2.0 branch?
if you want
or I will do it if you want to get to bed
doesn't matter, my only reluctance is because it has very little testing
wish we could have automated regression tests
2.0.1 will need to be tested for other reasons
there's no reason to leave this out during that testing
ok, I'll do it right now
its cvs up -j rev1 -j rev2 filename, right?
Hi, Alex. First login from my new residence. Moved servers, workstation, spent night first time last night. Whew!!!
steve_stallings is now known as steves_logging
steves_logging: glad to hear everything is ok..
I'm really getting into python
I'm interested in the cam side of things
and wrote a small bolt circle calculator with a tcl/tk front end for it
I would like to find a place to upload it
to inspire more beginners like me to play with it
I can put it on my blog if you like..
but it's not very visited. :D
ok I will email it to you. It's pretty small.
you could even mail it to emc-devel or emc-users
it won't show up in the arhives, but people subscribed will get it..
maybe they are interested
As long as I don't distract people from the actual EMC development work
I'm just having fun with it
nah.. most aren't very active either way
and there's also the cad_cam_edm_dro group at yahoo
I know it has a files section, where you can upload stuff
not sure if you're allowed or not, never studied the rules ;)
I bet it's fine
ok. Thanks for the advice. I need to go and wake up my kids. My dad is in town.
dan_falck: I'm interested in python for gcode generation too.
dan_falck: you can also use the emc wiki (wiki.linuxcnc.org) to upload files
oh, right.. forgot about the wiki..
jepler: any idea how to run keystick over a ssh?
I get some weird errors:
juve@ubuntu:~/emc2$ scripts/emc configs/sim/keystick.ini
EMC2 - Prerelease CVS HEAD
Machine configuration directory is '/home/juve/emc2/configs/sim/'
Machine configuration file is 'keystick.ini'
xterm Xt error: Can't open display:
xterm: DISPLAY is not set
Shutting down and cleaning up EMC...
ok.. got it, xterm gets run around keystick for some reason..
how goes it?
I guess I should document G50?
nah.. that's for 2.1.x :)
lots of chores and stuff to do
but its a 3-day weekend, so woot!
if chris comes around.. let him know I tested 2.0.1 a bit and it seems ok to me
sat, sun, mon
monday is memorial day, holiday here
I tested cross-references between various files just now, and it seems to work rather well
between lyx files you mean?
cross-reference to another section in another lyx file
also, we should start using Index tags
the Index at the end of the document is rather usefull
I wonder if we should have a lyx equivalent to "coding_style"
just a text file in the documents tree, recommending the things that people should do in lyx
e.g. a dustyu document which no-one reads, and respects?
yeah, I hope that too
maybe more like a faq - all the stuff a new doc writer needs to know
and an old doc writer needs to check every once in a while
how about we write that when we'll have a new doc-writer?
I was thinking about your suggestion to use index tags
* alex_joni is good at postponing stuff
I have no clue what they are
they are like labels
sure you could explain them to me, but then you'd have to explain them to the next guy too
and the Index table at the end of the documents, lists them alphabetically along with the page numbers they appear on (and links of course)
ok, so you'd stick them where inportant terms are first used, etc
but it's rather difficult to say this is worth an index, and this isn't ...
for instance you have G50 as an index.. so at the end you see all pages where G50 appears
you can have the same for HAL (might be a rather long list...)
would it be all pages where the term appears, or just all pages where there is an index tag?
seems like the latter is much more usefull
only pages where you add an tag
sorry.. was shortly away to make some soup :P
I'll try to work my way through the document and add tags here & there
but not tonight ;)
so jmk would it be hard to do a script that "compiles" the lyx to .pdf and uploads it when there is a change in cvs?
if the conversion from lyx to pdf can be done from the command line, then it isn't hard at all
there is lyxport which should do that.. yet I haven't installed it yet
I still think we should have the docs in the source tree
that way we could have 2.0.0 and 2.1 and HEAD versions of the doc that agree with the code
the source: the lyx files
the lyx too
and the drawings and all the rest
they need to be _somewhere_ web accessible
like, not buried in cvs somewhere
yes, the output (pdf and/or html) should be on the web
fenn: it's unlikely the docs change that often.. at least for now (with very few people working on them)
alex: fenn is planning ahead
I have no problem with that
no really i was just planning backwards
I agree with him, I think auto-generation of pdf would be great
but from what I've seen so far, 3rd party lyx converters suck big time
because i didnt know how to upload stuff to linuxcnc.org and the pdf that was up there was missing about 50 pages
lyx can do commandline easily.. lyx -e pdf blah.lyx
then how come alex is fighting with third party tools?
bash: lyx: command not found
how do you invoke lyx then?
juve@ubuntu:~/documents/lyx/emc2$ lyx-qt <- works
i dont even have lyx-qt
fenn: there's another lyx-xt
at least for Ubuntu that is..
i have /usr/bin/lyx /usr/bin/noweb2lyx and /usr/bin/reLyX
I have only the last 2
is lyx installed at all on myr machine?
jmkasunich: only if you added it
addressed at fenn
I know its on my machine, I did apt-get install
lyx - High Level Word Processor
lyx-common - High Level Word Processor - common files
lyx-qt - High Level Word Processor - Qt frontend
lyx-xforms - High Level Word Processor - XForms frontend
if i type just "lyx" i get the gui
fenn: what distro/lyx ?
but it can do "batch mode" stuff too
its an rpm, version 1.3.5 for fc3
packaged by dag wieers
mine (and alex's I think) is a apt package
my lyx manpage is an alias for lyx-qt
ii lyx 1.3.6-1ubuntu4 High Level Word Processor
it doesn't seem possible with this package to run the cmd line version
what happens if you type lyx-qt -e :)
seems to work
juve@ubuntu:~/documents/lyx/emc2$ lyx-qt -e
Missing file type [eg latex, ps...] after --export switch
ok so quit yer bitchin :)
Document exported as PDF to file `~/emcdev/documents/lyx/emc2/Master_HAL_Handbook.pdf'
* jmkasunich stops bitching (about that part)
now I'm gonna bitch about where the lyx files
files are stored
yeah cvs needs cleaned up a bit maybe
I'd like to move the entire documents/lyx/emc2 directory to emc2/src/docs or something along those lines
modify configure to see if lyx is present
and modify the makefile to build the docs if it is
that sounds reasonable
I think at least some folks will disagree
I should either make a motion on the board list or a suggestion on the developers list, and see what the reaction is
whats the major objection? that "closed' documentation is better?
no, that you'll have more than one version to maintain
the existing arrangement isn't closed
e.g. one for each branch
alex: not true
presuming a file got changed in all the branches
its worse now
then you want to add something that's common
I know.. I'm just saying what I think the majour objection will be
don't try to convince me.. I have no problem with moving the docs
I think I'll post on the dev list and see what people think
mmmm... noodle-soup :D
jmkasunich: btw, closed your bug-tracker about limit switch override
I thought about doing that last night
wait, which tracker?
I had no idea you opened one..
homing on limit switches
the one about removing the redundant limit checks in task?
thats good then
not the one about homing/offsets
I like chris'es idea on unhoming an axis though...
sent my msg to the list
* alex_joni goes to bed
good night all
jmkasunich: one downside: how much size is the documentation going to add to the source tarball?
jmkasunich: if it's all about branches, we can create branches in the "doc" repository just as well as in the emc2 one.
manual branch management
drat... bdi-2.xx still doesn't work
[John@cubix2 emc2head]$ scripts/realtime start
scripts/realtime: [: =: unary operator expected
/home/John/farm/emc2head/bin/emc_module_helper: Invalid usage with args: insert /lib/modules/2.2.18-rtl3.0/misc/rtl_time.o
but rtl_time _is_ on the list
dunno what the unary operator expected thing is
samco is now known as SkunkWorks
jmkasunich: and that /lib/modules is in the directory list?
could it be .o vs .ko?
(but I thought that came from a -Define)
hm, and everything in /lib/modules is supposed to be on the whitelist
cc module_helper/module_helper.c -Wall -Werror -DMODULE_EXT=\".o\" -DEMC2_MODULE_DIRECTORY=\"/lib/modules/2.2.18-rtl3.0/misc/\"
$ fakeroot ./a.out insert /lib/modules/2.2.18-rtl3.0/misc/rtl_time.o
insmod: can't read '/lib/modules/2.2.18-rtl3.0/misc/rtl_time.o': No such file or directory
looks like to me it should work, though admittedly this is not a very good test