#linuxcnc-devel | Logs for 2013-10-20

[16:24:01] <alex_joni> can one of you check if a remote connection (even from the same PC) works with a current master?
[16:24:13] <alex_joni> I mean a NML remote connection..
[16:24:40] <alex_joni> here's how I tried: copy sim/axis.ini to axis_server.ini and axis_remote.ini
[16:24:57] <alex_joni> then copy configs/common/linuxcnc.nml to server.nml and client.nml
[16:25:18] <alex_joni> hmm.. maybe it's easiest if I just put the configs somewhere :)
[16:30:04] <alex_joni> http://juve.ro/~juve/lcnc/remote/
[16:30:20] <alex_joni> from terminal 1 run linuxcnc server.ini
[16:30:38] <alex_joni> from terminal 2 run tklinuxcnc.tcl -ini client.ini
[16:34:14] <alex_joni> for me that fails even on the same host with master !?
[16:40:20] <alex_joni> same with 2.5 ... wth
[17:09:53] <cradek> sigh, guy on the forum says our tool orientation documentation is wrong, and so he sets it the way he thinks is right, and then he says his compensation is off by a tool diameter
[17:11:04] <alex_joni> hmm.. the remote stuff works in 2.4
[17:11:32] <alex_joni> but it's been broken since 2.5 :/
[17:11:51] <cradek> hmm do you know why yet?
[17:13:01] <alex_joni> nope.. just finished building 2.4 and it works there
[17:13:08] <alex_joni> was planning on running git bisect next
[17:13:20] <alex_joni> if I can figure out the incantation
[17:16:46] <alex_joni> at least a compile doesn't take that long on this laptop
[17:44:28] <alex_joni> 24 revs to go
[17:57:59] <alex_joni> origin/v2.5_branch
[17:58:09] <alex_joni> grr.. bad paste
[17:58:46] <alex_joni> [d214ba755156a6a5a9526c75ada0962a9a447b9c] libnml: avoid strict aliasing warnings
[18:03:51] <alex_joni> http://git.linuxcnc.org/gitweb?p=linuxcnc.git;a=commitdiff;h=d214ba755156a6a5a9526c75ada0962a9a447b9c
[18:11:41] <alex_joni> jepler: ^ any idea why that makes it fail, I skimmed it over and it looks ok..
[18:17:07] <alex_joni> ha.. found it
[18:21:02] <alex_joni> cradek, or someone else with push access: http://juve.ro/~juve/lcnc/0001-fix-typo-preventing-remote-NML-clients-talking-to-lc.patch
[18:21:26] <alex_joni> that should go in 2.5
[18:41:03] <andypugh> cradek: Ah! So that is is his problem?
[18:45:54] <alex_joni> yay, got axis to connect to a lcnc on a BBB :)
[18:48:34] <alex_joni> http://imagebin.org/274227
[19:31:04] <KGB-linuxcnc> 03alex_joni 05v2.5_branch 59adbb1 06linuxcnc 10src/libnml/buffer/tcpmem.cc * fix typo, preventing remote NML clients talking to lcncserver
[19:32:28] <KimK_1> alex_joni: OK, there you go. (Hope I did that right.)
[19:46:30] <KimK_1> Thanks very much for finding and fixing that problem, BTW.
[19:59:09] <seb_kuzminsky> looks good alex_joni, thanks!
[19:59:12] <seb_kuzminsky> and thanks KimK_1 for pushing it
[20:00:44] <cradek> andypugh: yes he should trust that image in the docs. I can't tell if he's saying he would like the image flipped to match the way his lathe moves (in which case it doesn't matter) or if he's saying the tool orientations don't match his expectations for some reason (like it's different from some other control for instance)
[20:01:38] <cradek> if he sets his orientation according to the docs he'll get correct motion
[20:02:12] <cradek> the preview will look mirrored if his lathe has the tool on top - different issue
[20:02:17] <andypugh> Orientation relative to X?
[20:02:35] <cradek> he may or may not be mixing all these things up in his head...?
[20:02:59] <andypugh> i can't recall if the picture has a nice big "X is bigger this way" arow
[20:03:22] <cradek> yes it does - it has axis arrows
[20:03:37] <cradek> +X is this way, +Z is this way
[20:04:08] <andypugh> I guess your forum-allergy means that you would prefer me to point this out?
[20:04:45] <cradek> you MUST not try to use the tool orientation to make the preview flip upside-down or mirror or whatever it is you want with a tool on top
[20:05:32] <cradek> ah hell I should really try to get over that shouldn't I
[20:06:01] <andypugh> I have. (I hate forums too)
[20:12:43] <cradek> alex_joni: good find, wow, that's an old one
[20:13:01] <cradek> we now know approximately how many people use this feature (0)
[20:15:09] <andypugh> I have a big to fix in lcd.c. The only way it can have been missed is is nobody uses it.
[20:16:55] <KimK_1> I'm not using it yet, but I've got my eye on a couple of Mesa's pendant/sw/LCD cards. So be patient.
[20:17:49] <pcw_home> The matrix_kb comp seems to work fine...
[20:18:30] <pcw_home> I didn't check all 64 output bits though
[20:21:04] <seb_kuzminsky> cradek: 2.5 seems to be working fine with linux 3.4 + rtai, too
[20:21:46] <cradek> andypugh: oh god someone has told him to use coordinate system rotation to flip the direction of X in the preview
[20:21:55] <cradek> no wonder he's confused by everything
[20:22:08] <andypugh> That was me
[20:22:22] <cradek> that can't possibly work can it?
[20:22:55] <andypugh> The thread I linked to suggested that it worked best of the alternatives.
[20:23:22] <cradek> this is a gui preview problem
[20:23:22] <linuxcnc-build> build #1077 of deb-precise-sim-binary-i386 is complete: Failure [4failed apt-get-update shell_1] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/deb-precise-sim-binary-i386/builds/1077 blamelist: Alex Joni <alex_joni@users.sourceforge.net>
[20:23:32] <cradek> somewhere there's an .axisrc hack to fix it
[20:23:50] <cradek> someone should just fix axis the right way and make it a gui option
[20:23:53] <andypugh> All the options are hacks
[20:24:05] <cradek> haha, that doesn't mean they're equally bad
[20:24:38] <andypugh> Yes, but there are rather few "someones" who speak Tcl
[20:25:41] <andypugh> GEOMETRY -Z (if I recall correctly) is more confusing than the rotated axis.
[20:25:45] <cradek> but no lack of people who use that as an excuse - any programmer can copypasta a language he doesn't know
[20:25:56] <cradek> that was an inside thought, sorry
[20:26:50] <andypugh> Having said that, the axis-rotation option was for someone with gang-tooling where he wanted to switch viewpoint from tool to tool.
[20:26:56] <seb_kuzminsky> oh jeebus, *more* network problems in the buildbot
[20:27:56] <cradek> I can't picture what coordinate system rotation will do to cutter comp on a lathe
[20:28:05] <cradek> it'll surely make it wrong because it rotates "Y" as well
[20:28:36] <andypugh> Yeah, that's an interesting wrinkle.
[20:30:05] <andypugh> I am going to stae now that I wll _never_ push any changes to Axis. It isn't my code, I don't understand it. I don't care to understand it. If I see a bug I will report it and moan. Sorry.
[20:30:43] <cradek> I wish I wouldn't have said that :-/
[20:31:55] <andypugh> It was a fair comment. I don't disagree with you.
[20:32:44] <andypugh> The part of that aspect of Axis that I can't follow is where is starts to get clever with OpenGL.
[20:32:51] <cradek> jeez AXIS's 10 year birthday is coming right up
[20:33:22] <cradek> yeah to get that X flip you need the right bit of opengl aside from the gui glue
[20:34:01] <cradek> but I don't think that'll make all the numbers wrong, since they auto-orient -- so I think it's at least straightforward
[20:34:40] <cradek> might make sense to have view-switching buttons just like in mill mode, but there'd only be the two of them
[20:35:08] <cradek> but none of this fixes back-side tools, that's another issue
[20:35:29] <cradek> (easiest today is to just use negative X coordinates)
[20:36:36] <cradek> there could be X-negation that I think would not be any harder than diameter mode, but what to do about tlos might be weird
[20:37:16] <cradek> would have to think through arc IJ as well (which are not a problem in the diameter mode hack)
[20:37:29] * cradek waves his hands wildly
[20:37:44] <andypugh> If I had back-side tools I would use negative X.
[20:37:52] <cradek> I have and I did
[20:37:52] <skunkworks_> take a breath... ;)
[20:38:05] <andypugh> I use negative Z all the time anyway.
[20:38:24] <cradek> very true
[20:38:50] <andypugh> But, if all the tools are back-side, I am not sure what the "right" way is
[20:38:53] <cradek> I'm not sure if I've used back-side with ccomp but I'm sure it'd work if I thought it through and set the orientation right
[20:39:09] <cradek> use postive X for all the tools, obviously
[20:39:27] <cradek> there is NOTHING special about that setup except AXIS's preview happens to look wrong
[20:40:05] <cradek> "back-side" tool on that lathe is in front (negative X)
[20:42:10] <cradek> http://wiki.linuxcnc.org/cgi-bin/wiki.pl?BackToolLathe
[20:42:23] <cradek> here is the correct advice (if it hasn't bit-rotted and still works)
[20:42:40] <cradek> it reverses the arrows too, which I forgot is needed
[20:43:24] <cradek> hey it even works
[20:43:41] <cradek> it makes the letter Z wrong
[20:44:02] <andypugh> I have _never_ seen that before.
[20:44:54] <archivist> just wait till someone gets this type of lathe set up :) http://www.collection.archivist.info/archive/DJCPD/PD/2012/2012_08_31_Bechler_N.EA/IMG_1338.JPG
[20:45:02] <cradek> I had trouble finding it, even knowing it existed somewhere
[20:45:11] <andypugh> We first need to define "correct" then implement it.
[20:46:44] <andypugh> archivist: That's actually relatively easy (in fact all lathes are equally easy). -X == smaller stuff.
[20:46:51] <cradek> aside from Z showing up wrong, I think what that .axisrc does is correct (mirrors the preview, flips the X jog keys)
[20:48:00] <andypugh> I think if it was my machine I would just adapt to the preview being "conventional"
[20:48:07] <archivist> andypugh, one of the axes has two tools one front and one rear on a rocker
[20:48:29] <andypugh> Though the jogging might confuse
[20:48:35] <cradek> yeah was just thinking that
[20:48:53] <cradek> but reassigning those is one line in .axisrc
[20:49:08] <andypugh> archivist: Well, yes, but you still only care about diameter space.
[20:49:42] <archivist> I know just the diameter matters but a gui ....
[20:49:51] <andypugh> If axis showed both halves of the work, and a tool at both sides?
[20:50:10] <cradek> AXIS doesn't show the work
[20:50:32] <andypugh> Ok, both aspects of the path?
[20:50:43] <cradek> you mean the program and also a mirror of the program?
[20:50:49] <cradek> hmmm
[20:50:51] <cradek> interesting idea
[20:51:16] <andypugh> In fact, how hard is it in opengl to show a solid of revolution of the tool path?
[20:51:35] <andypugh> (Might mask too much internal detail)
[20:51:50] <cradek> yeah I'm not sure that's what you want at all
[20:52:10] <cradek> the outermost shell of everywhere the tool ever travels probably isn't useful
[20:52:19] <andypugh> But path and mirror seems right for a lathe, really
[20:52:50] <cradek> yeah it kinda does I guess
[20:53:10] <cradek> but only one tool, maybe rapids show only in "front"?
[20:53:38] <cradek> I'm not volunteering to do this :-)
[20:54:26] <Tom_itx> your lead in/out could also be an issue
[20:57:10] <cradek> yeah that .axisrc code is slightly bitrotted - the recentering part doesn't work
[20:57:28] <jepler> alex_joni: groan, I'm glad you found my fubar
[20:57:46] <jepler> we should put a client/server nml sim config in git if it actually works
[20:57:54] <andypugh> My brain is bitrotted. Time to sleep. :-)
[21:00:51] <cradek> I unrotted it
[21:01:23] <cradek> hi jepler
[21:01:26] <cradek> are you back home?
[21:01:53] <archivist> rather than a tool at a side, looking at the sliding head it is a tool at an angle
[21:02:25] <cradek> eh, they are all radial, the work spins
[21:02:33] <jepler> cradek: no, still up in mn for a few more days
[21:02:44] <cradek> ah
[21:02:54] <cradek> it's freezing at night here now...
[21:02:57] <archivist> I know that, but one would want a gui to look sensible
[21:04:08] <skunkworks_> mn?
[21:04:47] <jepler> skunkworks_: yeah, visiting some friends and family in the twin cities
[21:05:07] <skunkworks_> ah - neat.
[21:07:30] <skunkworks_> your about 2.5 hours from here. Sort of wrong direction
[21:08:02] <skunkworks_> https://maps.google.com/maps?saddr=Trempealeau,+WI&daddr=Twin+Cities,+MN&hl=en&sll=44.937483,-93.201&sspn=0.414123,1.012115&geocode=FY54nwIdWLSM-imBfqV2A0T5hzFa3Q8Kez-3VQ%3BFQuxrQIdmN1x-imPU7db_Sn2hzEgpK6ohsFw0A&oq=trempe&t=h&mra=ls&z=9
[21:08:02] <jepler> yeah, not sure that would interest Ingrid too much
[21:08:07] <skunkworks_> heh
[21:09:09] <skunkworks_> but wait - you could take the 'great river road' and see a bunch of eagles...
[21:18:01] <jepler> thanks for the invitation.