hm, esc-to-cancel-loading is broken in one obvious way, but after that it's still broken :(
maybe I'm half-mistaken
esc-to-cancel-loading doesn't work if you're starting emc on an infinite-loop program, but works if you load it after startup
that sounds odd
I *think* it starts loading too early, and when the stat buffer is first read or the X axis button is first mapped, it takes the focus away from the progressbar, which is the thing that has the Escape binding
yuck, that startup stuff sucks
I didn't intend to remove any pins, but I applied a patch that did it (removed the oddly-named motion.motion-inpos)
and by commenting out those lines, no less
I guess I didn't review the patch I was applying carefully enough
EMC: 03cmorley 07TRUNK * 10emc2/src/emc/usr_intf/pncconf/ (pncconf.py pncconf.glade): work to get the pintype editbox display the right name and dispaly the right signal names- next to get signal names to change when GPIO switches from input to output
fenn_ is now known as fenn
cradek: I built some emc2-2.3.0-beta2 packages for dapper
I copied them to the proper place, but you need to work your magic on them (some extra packages, bwidget & co, and repo changes)
EMC: 03jepler 07TRUNK * 10emc2/src/hal/user_comps/hal_input.py: revert to rev 1.16 to fix regression (SF#2569072)
EMC: 03jepler 07v2_3_branch * 10emc2/src/hal/user_comps/hal_input.py: revert to rev 1.16 to fix regression (SF#2569072)
EMC: 03jepler 07TRUNK * 10emc2/src/emc/usr_intf/stepconf/stepconf.py: SF#2712817: location of sample-configs with --prefix=/usr/local
jepler: did you try to contact Brad McLean about hal_input?
if you must just revert his change, you should revert the docs too - but it would be better to ask him to fix whatever is wrong.
Hi Everyone. I have seen emc2 controllong 5-axis machines on youtube.
[16:29:08] <cfrank> http://www.youtube.com/watch?v=7WCrqqoZkPg
Where do I find configs to do that?
the configs are all part of the CVS repo
the 5-axis stuff is here:
[16:30:16] <seb_kuzminsky> http://cvs.linuxcnc.org/cvs/emc2/configs/5axis/
if you have a CVS checkout it's in there too
Thanks seb_kuzminsky. I will check it out.
if you have installed the 2.3beta deb, the configs are in /usr/share/doc/emc2
Even better. I have.
oh hey, that's my machine - I forgot about that video.
this is a great demonstration: http://www.youtube.com/watch?v=DjPCEpZybXs&feature=related
argh. flash is dead on my laptop
I like this comment "NICE! but how do you control 5 motors at once? and how do you set up emc, i saw 9 axis config in the set up but i think it was only for simulation?"
jepler: does 2c3fc905e7646577188cecdb5248dd811b5bd662 also belong on the 2.3 branch?
* BJT-Work gets out his secret decoder ring to see what that says
strange - when I search for it in gitweb, it doesn't show me that change
odd that I can search by committer in gitweb but not gitk
jepler: how did you locate that in gitweb?
cradek: I used 'git log' to find it, then scanned the shortlog for the commit message
strange that decoded to "the butler did it in the conservatory with the candlestick" if I use my x-ray glasses
cradek: Thanks for the update. Nice! I was looking to build an XYZBC machine. I read that the BC tables result in less room than a multi axis head.
jepler: is there any smarter way to handle the situation where we have a branch that is stabilizing, and a bugfix change needs to go on it and the trunk both? I can see where maybe git could notice that it's the same change.
I was hoping to have B axis on the Z column and the C rotary table on the base. Taix mill is too small for a gomble
yeah, that's what I built too
it helps a lot if the center of B rotation is near where a typical tool tip will be
otherwise you use up huge amounts of X travel when B moves
cradek: this is what git-cherry is about
I see. Could you cut metal on it?
on mine? very marginally
cradek: or, rather, git-cherry-pick
depends on the metal I suppose
git-cherry to find things on another tree that you don't have; git-cherry-pick to apply one of them to your tree
I cut aluminum mostly. I was hoping to keep things very compact for stiffness.
cradek: so for instance, 'git-cherry -v v2_3_branch master | grep '^+' finds changes on master that are not in v2_3_branch, including 2c3fc. Apply it by checking out v2_3_branch and 'git-cherry-pick 2c3fc'
* seb_kuzminsky blinks
is cradek using git?!
maybe there is a god ;-)
seb_kuzminsky: did you get the config from yesterday?
did you send it to me?
22:53 < alex_joni> hmm.. now I have a config for seb, but he's not here :)
22:53 < alex_joni> http://pastebin.com/m2573b842
seb_kuzminsky: I'm pretty sure there's not...
right, if there were, it'd be telling you to use bzr ;-)
I think we need to switch to something - I'm trying to figure out which one I think is least intolerable
so far I think git has shown the best import of our old data
what did you not like about the bzr import?
the only one I've seen is the one on launchpad which discards all the existing forks
alex_joni: is it strange that the default acceleration (line 138) is 100, but the axis 0 max accel (line 158) is only 80?
cradek: the lp folks deliberately convert just one branch... there are other tools that purport to convert them all
ok thanks, I might dig into that
seb_kuzminsky: strange, yes.. but it shouldn't be an error
it should clamp to max accel
cradek: i had intended to try cvsps-import, which claims to convert a whole CVS repo to BZR, but it needs local read access to the repo
[19:45:01] <seb_kuzminsky> https://launchpad.net/bzr-cvsps-import
alex_joni: i'll give that config a whirl and see if i can reproduce the problem
seb_kuzminsky: strange, because cvsps doesn't need that
seb_kuzminsky: (but of course we have it if we need it)
I tried bzr cvsps-import on a small project of my own (400 changesets by cvsps's count) and it does seem to have given a tree structure with every branch reflected
but each branch has its own .bzr directory
[20:23:23] <jepler> http://hackage.haskell.org/trac/ghc/wiki/DarcsEvaluation
jepler: each branch *should* have its own .bzr directory, but not every .bzr directory needs to contain a repo
i usually use "shared repos", where a top-level dir has a .bzr with all the repo info in it, then one dir per checked-out branch, each of which contains its own (tiny) .bzr dir with just the checkout info
jepler: "Darcs makes the simple things tough, the difficult things near impossible, and the impossible merely really slow."
seb_kuzminsky: I've got a bzr branch called BRANCH_STABLE and one called HEAD. How do I get information about the difference between them. For instance, how do I find the last common version they shared in history?
the easiest is probably "bzr vis"
cradek: yes, one common complaint against git (at least last time i checked) was that it doesnt treat rename as a first-order operation, it tries to do post-facto rename detection and then gets itself confused...
bzr vis --help
bzr: ERROR: unknown command "vis"
jepler: it's in the bzr-gtk plugin
Oh, I was looking for something that worked on the terminal
jepler: say you want to find what's new in HEAD since last time there was a merge between HEAD and BRANCH_STABLE
bzr diff -r ancestor:../HEAD
i think that's it
that's from the "bzr help revisionspec" info
or log instead of diff, etc
bzr log -r ancestor:../HEAD..
ok I think that's it but it's a bit odd
it's just not like git
particularly having ".." mean "parent directory" and "since" in the same word was a bit odd
cradek: I think renames should be like abortions. legal, safe, and rare.
god help us from the well-meaning developer who shows up one day and renames the whole source tree because he can
jepler: here's another way to find the most recent ancestor between two branches, if that's what you're after:
in fact I'm sure we've had a few who would have done that and were only stopped because our system was CVS
bzr find-merge-base HEAD BRANCH_STABLE
(how come nobody has implemented a dvcs that also lets each developer run 'indent' with his favorite arguments and still gets merges right? Now *THAT* would be awesome)
that spits out a revid, which is the globally-unique revision identifier, mapping that to a revno is branch-dependent and requires (i think) "bzr log -r revid:$REVID"
revids are almost as gross as git revision identifiers :-(
good thing you almost never have to look at them in bzr :-)
in no system could I commit a globally unique tree identifier to memory, or give it over the telephone. but I don't do those tasks very often either
in bzr it's enough to say something like "trunk revno 1234"
that's unambiguous, human-readable, and globally unique
but when I branch, it's not trunk anymore
it becomes "my branch revno 1"
and "your branch revno maybe some other number"
i see what you're saying, and i agree
but i think that's fine
we all need to agree on some master line, like our current trunk
we'll all have clones of that tree on our local development machines
we make local branches by forking from the local copy of upstream
hack in the local branch
commit straight to the upstream trunk
everyone can agree on revnos in the trunk mainline, and can easily diff against it, etc
that brings me to another question
the bzr oracle is: ONLINE
git has git-shell which is a restricted shell that only allows git to send and receive "packs"
it lets you give developers permission to push to a central system without giving them full shell access
is there a bzr equivalent?
[21:15:47] <jepler> http://ww2.samhart.com/node/47
there's got to be something like it, that's how launchpad works
but i dont know exactly how they do it
is it one of the secret things they keep, er, secret while saying how open source they are?
they sure are interested in making you use their service, what with the integration of lp:
(maybe you can use sftp and "scponly" (http://www.sublimation.org/scponly/wiki/index.php/Main_Page)
to do this
jepler: that looks promising
i think my least favourite thing about bzr is the confusion of storage formats (repo dbs)
is there a way to find the direction of the path? and rotate a "cutter", so that it faces it?
it may be possible to look at the differences (axis.#.motor-pos-cmd-axis.#.motor-pos-fb) to see the instantaneous direction
you'll need an atan2 or something in there as well
but that will be very noisy
or use ddt on .motor-pos-fb
cmd - you want to know before the motion takes place I Think
if you do the subtraction I suggested just after motion.do-motion-calcs (or whatever it's called), you'll get the difference between the new command and the current position, which is hopefully closer to the direction it will move
when this has been discussed before, we've noticed the best solution is to write C directions in the gcode
ddt(cmd) didn't work or too noisy?
what about corners? or alignment when starting a cut?
ddt will always give you an answer after the fact, since it's looking at (now - before)
versus (soon - now)
might want to search the mailing list archive - I forget all the different things that were discussed
yeah, there are a lot of complications, and using C or some other "unused" axis is much more likely to work well
I've already tried, but can't find anything related...
jepler: ick, fedora think they need a chroot for bzr
with C axis -- then you need a custom converter from whatever input you use...
yes you need to change whatever creates the gcode
fortunately the math is very easy
yes, if you have source :)
and I guess it is not possible to use G02/G03 any longer?
sure, why not?
in EMC2 you can interpolate C over an arc
so you will get an exact circle with an exact tangent cutter
oh, so _any_ axis will be interpolated, not only Z?
[22:10:31] <cradek> http://www.linuxcnc.org/docs/devel/html/gcode_main.html#sub:G2,-G3:-Arc
XYZ are used to determine the feed rate, C will just be matched to the time needed to do the XYZ move
"If a line of code makes an arc and includes rotational axis motion, the rotational axes turn at a constant rate so that the rotational motion starts and finishes when the XYZ motion starts and finishes."
so using C is better in that the XY feed rates will not be affected
thanks all :)
welcome, hope you get it going
still need to convince mechanics about it :)
the bzr guys recommend bzr+https for "authenticated commit without shell access"
so I was talking to this guy about a nice little embedded PC, and I asked him if LinuxBIOS worked on it
of course, he had never heard of it :)
they changed the name
yeah, OpenBIOS or OpenBoot or something
it's coreboot now
[22:17:16] <alex_joni> http://www.coreboot.org/Welcome_to_coreboot
well, he'll find it if he searcheds for LinuxBIOS too :)
well.. I'm off to bed
cradek: did you see my message about dapper packages?
(2.3.0-beta2 packages to be exact)
no pressure, just wanted to let you know I uploaded them
ok, did you create some new directory for 2.3 then?
but you need to add some symlinks (I saw for 2.2 ..)
was afraid I'd get those wrong (especially since you rsync it)
I added 2.3 and 2.3-sim
anyways.. off to bed now
good night all
seb_kuzminsky: let me know if you discover anything with that config
alex_joni: I'll see if I can figure it out
well.. you did for 2.2 ;)
so I'm sure you'll figure it out again..
alex_joni: done, I think
here's the somewhat sparse documentation on how to set up authenticated push over bzr+https, while allowing anonymous read-only access: http://doc.bazaar-vcs.org/bzr.dev/en/user-guide/index.html#serving-bazaar-with-fastcgi
I have to say, git-shell is pretty appealing compared to that