#emc-devel | Logs for 2009-10-02

[07:14:19] <micges_work> hello
[09:57:03] <alex_joni> morning
[13:14:36] <BJT-Work> morning
[13:14:46] <SWPadnos> bummer
[13:47:48] <micges_work> today I saw two strange things: once Axis did backplot only when moves was in X+ direction and not in X- direction
[13:48:13] <micges_work> and few times after executing M2 flood and mist was enabled
[14:09:51] <cradek> should G53 do the full unwind on a stuart-style rotary, or should it just disregard g5x,g92 offsets and go to the position?
[14:13:00] <SWPadnos> there ought to be a setting
[14:13:19] <SWPadnos> similar to the modulo axes thing that you/Jeff(?) had worked on before
[14:14:12] <cradek> a setting for g53 behavior? or a setting for stuart-style versus old-emc-style?
[14:15:04] <cradek> or a third style? ss vs oes vs pjobis (previous jepler oops bad idea style)?
[14:15:08] <SWPadnos> stuart-style vs old style
[14:15:12] <SWPadnos> heh
[14:15:21] <cradek> yes of course ss vs oes is a setting
[14:16:15] <SWPadnos> or just add a G/M code to reset the axis to whatever it would be in modulo-land
[14:16:23] <jepler> I worked on modulo axes but never put it in because it didn't make anybody happy
[14:17:16] <SWPadnos> these things happen
[14:17:39] <jepler> and with git it's even easier to not put in things that don't make people happy.
[14:17:43] <SWPadnos> heh
[14:17:53] <SWPadnos> local unhappy-making changes are easier to maintain
[14:19:29] <SWPadnos> gah. is there a program that should be used to remove fingerprints from ./ssh/known_hosts?
[14:19:43] <cradek> vi
[14:19:43] <SWPadnos> other than some sed/awk-foo
[14:19:55] <cradek> it's just a text file...
[14:19:56] <SWPadnos> :P
[14:19:59] <SWPadnos> yes, I know
[14:20:51] <SWPadnos> I could swear there was some utility that could remove them by some reference, but I don't remember it
[14:21:25] <jepler> ssh-keygen
[14:21:29] <jepler> -R
[14:21:32] <cradek> a moduloing gcode is tempting but as far as I know that scheme doesn't exist outside our fevered imaginations
[14:21:58] <cradek> I actually like that the DRO stays in [0,360)
[14:22:24] <cradek> the +/- bit feels a little goofy but it's ok
[14:22:39] <SWPadnos> jepler, thanks
[14:23:13] <SWPadnos> yeah - and the CNC bible that Steve B quoted doesn't actually have all the information we need regarding modulo/not
[14:24:13] <cradek> as far as I saw, it didn't talk about modulo at all
[14:24:26] <cradek> it described exactly the oes
[14:24:34] <jepler> absent a reason to unwind, I think that g53 should do the least motion to get to the place the user wants
[14:24:50] <cradek> jepler: yay, thanks for answering my question
[14:24:52] <jepler> once we become aware of a reason, we could look at changing it or making it an option
[14:25:03] <cradek> I think that's probably right, too
[14:25:30] <cradek> if there is one, the reason to wind would be purely an "internal" reason (internal to the software)
[14:25:39] <cradek> I think it doesn't represent anything in the real world
[14:31:01] <cradek> I will do that.
[14:48:03] <cradek> ok I have new-g53 working. It uses the sign to determine which way to turn. It goes to the nearest given position in that direction, disregarding the g5x and g92 offsets.
[14:49:42] <SWPadnos> are -0 and +0 distinct?
[14:49:47] <cradek> yes
[14:49:49] <SWPadnos> cool
[14:50:34] <CIA-7> EMC: 03cradek 07ss-wrapped-rotary * ra8aa3ddced24 10/src/emc/rs274ngc/ (interp_find.cc rs274ngc_interp.hh): fix G53, ss-style
[14:52:40] <cradek> but did you know that -0.0 is not less than 0?
[14:52:47] <cradek> but did you know that -0.0 is not less than 0.0?
[14:53:13] <jepler> of corse; they're equal
[14:53:55] <cradek> I was smart enough to know I had to read lots of man pages before trying to program inside that tar-pit
[14:54:08] <SWPadnos> heh
[14:54:09] <cradek> * cradek exercises his metaphors today
[14:54:29] <cradek> is a bunch of pies in the sky a nebula of pies?
[14:54:41] <seb_kuzminsky> good morning
[14:54:53] <SWPadnos> pie-constellation
[14:54:55] <cradek> if so, I can say things like "that sounds like very nebulous sky-pies to me"
[14:55:01] <cradek> hi seb
[14:55:30] <SWPadnos> oh. I should pay my bills
[14:55:44] <SWPadnos> so easy to forget when you only spend 3 or 4 days at home
[14:56:37] <cradek> jepler: seems that there is fmod but not fdiv. what is the right way to do fdiv?
[14:57:10] <jepler> cradek: I don
[14:57:14] <cradek> casting to an integer type has a range problem
[14:57:30] <jepler> 't know of a function that always reduces to a range [0,n)
[14:57:35] <jepler> despite how useful that is
[14:58:11] <jepler> er..
[14:58:21] <jepler> that's the opposite of what yo said
[14:58:22] <cradek> no I want the opposite of that
[14:58:30] <jepler> floor(a/b) ?
[14:58:51] <cradek> you are right I bet
[14:59:04] <jepler> but that's not hte one that corresponds with fmod() which says it rounds to 0 to find n
[15:01:09] <cradek> that's ok, I think - I'm not using fmod anywhere
[15:01:14] <CIA-7> EMC: 03cradek 07ss-wrapped-rotary * ree1472b4491f 10/src/emc/rs274ngc/interp_find.cc: fix eventual range problem caused by going through an int
[15:01:53] <cradek> I'd sure appreciate it if you'd check this for floating point related bloopers (or others)
[15:26:40] <micges> hi seb
[15:28:48] <seb_kuzminsky> hi micges
[15:55:43] <micges> did you saw my screens about Nan on hm2 encoder velocity
[15:55:43] <micges> ?
[16:11:03] <CIA-7> EMC: 03cradek 07ss-wrapped-rotary * r6504c7e89e8e 10/src/emc/rs274ngc/ (interp_convert.cc interp_find.cc interp_internal.cc): fix probe result, g28/g30, g28.1/g30.1
[16:20:23] <seb_kuzminsky> micges: i saw it, but i havent looked into the bug yet...
[16:20:32] <seb_kuzminsky> is that with top-of-tree?
[16:20:41] <seb_kuzminsky> master branch?
[16:20:46] <micges> yes latest
[16:21:02] <seb_kuzminsky> ok
[16:21:16] <seb_kuzminsky> did you get a good zoom in on what the encoder counts & position were doing at the time?
[16:21:40] <micges> I can do that
[16:21:50] <micges> on monday
[16:22:20] <seb_kuzminsky> do you still have it up, or is it easy to repeat?
[16:22:33] <micges> cradek also mentioned to check how long NAn was on velocity pin
[16:23:03] <micges> I don't know if its easy to repeat
[16:23:20] <micges> but I have config to do this
[16:23:39] <seb_kuzminsky> you have a config that reliably produces NaN on hm2.encoder.vel?
[16:25:44] <micges> I have machine that every time produces nan, very easy to repeat, and I can send you config from that machine, I don't have any rt machine handy to test in sim config
[16:26:01] <seb_kuzminsky> ok cool
[16:26:12] <seb_kuzminsky> sure, please send me the config from the problematic machine
[16:26:18] <seb_kuzminsky> seb at highlab.com
[16:26:25] <micges> ok
[16:26:29] <seb_kuzminsky> or pastebin
[16:31:12] <micges> seb_kuzminsky: thanks, I'll send you later, gotta run
[16:34:17] <CIA-7> EMC: 03cradek 07master * rff095f6b3d16 10/src/emc/task/emccanon.cc: yuck: fix home / g0a180 / g28 going nowhere
[16:35:57] <CIA-7> EMC: 03cradek 07ss-wrapped-rotary * r94c478f07639 10/src/emc/task/emccanon.cc: Merge branch 'master' into ss-wrapped-rotary
[20:38:55] <jepler> cradek: wtf ? yuck: fix home / g0a180 / g28 going nowhere
[21:16:44] <cradek> jepler: stupid tiny XYZ moves again. I was getting XYZ values of 1e-8. I thought it was from rotate() but when I printed the values going in and out they were all 0.
[21:20:54] <cradek> I did not find floats anywhere I looked. I don't understand it.
[21:22:40] <alex_joni> hi