#linuxcnc-devel Logs

Sep 12 2018

#linuxcnc-devel Calendar

03:52 AM Jin|away is now known as Jin^eLD
03:54 AM Jin^eLD: morning
09:19 AM hazzy-dev: Jin^eLD: TurBoss did some work in a Yacto recipe for LCNC a while back: https://github.com/TurBoss/meta-hazzy
09:19 AM hazzy-dev: IIRC the hold up was that LCNC uses Py2.7, but only the Py3 version of boost was available in Yacto
09:20 AM Jin^eLD: ah nice, so there were some attempts already
09:20 AM Jin^eLD: well, maybe python 2.7 could be backported
09:21 AM mozmck: Hasn't some work been done to make LCNC use Py3?
09:21 AM Jin^eLD: I am surprised he was able to pass the tcl checks, I would have expected a lot more patches to configure
09:22 AM hazzy-dev: mozmck: yes, rene and turboss have both been doing some work on adding support for Py3
09:22 AM mozmck: Yeah, looks like the pull requests for it are still not finished.
09:23 AM Jin^eLD: I understood jepler that py3 was not an easy task
09:25 AM Jin^eLD: TurBoss: why did you have to override do_configure and do_compile??
09:25 AM hazzy-dev: Jin^eLD: Py2 to Py3 is easy, very easy, the problem is with the python interfaces that use c++ and boost
09:25 AM Jin^eLD: maybe thats what he meant :)
11:09 AM rene_dev_: the configure script and some parts are done, the most work is probably in the c and cpp bindings
11:10 AM Jin^eLD: I see...
01:21 PM andypugh: I am a bit out of my depth on the forum.
01:21 PM andypugh: https://forum.linuxcnc.org/41-guis/35225-new-gui-for-embedded-systems#117403
01:21 PM andypugh: Someone is writing a GUI in C. (he thinks he has valid reasons) but I can’t find an example of sending an NML message via C.
01:35 PM jepler: we don't have a C API, just C++.
01:35 PM jepler: (I wish it were otherwise, but there you go...)
01:37 PM andypugh: So, you can’t write a GUI in C?
01:40 PM jepler: not without creating some kind of shim between the C++ classes and a C API of your own divising
01:40 PM jepler: devising?
01:41 PM jepler: and of course you can't build linuxcnc without C++ anyway
01:43 PM andypugh: Seems like it could be quite a thin shim.
01:45 PM rene_dev_: andypugh the lathe macros are now working, and are working great!
01:45 PM rene_dev_: I really like them
01:46 PM rene_dev_: I have no idea what the problem was, but for some reason the macros did not get any parameters
01:46 PM andypugh: He is currently operating LinuxCNC from the GUI using strings sent to halcmd. Which doesn’t seem the neatest way.
01:46 PM andypugh: rene_dev_: Wierd.
01:47 PM rene_dev_: I think they should be included in linuxcnc.
01:47 PM andypugh: Are you using the version that assembles MDI strings? (Not the version where the G-colde subs try to read HAL pins)
01:47 PM rene_dev_: the mdi version, the non-svg version
01:47 PM andypugh: OK, that really ought to work.
01:47 PM rene_dev_: not sure if the svg version works with axis/gmoccapy
01:48 PM andypugh: It will work with a touchscreen if, and only if, you add an external (preferably physical) start button.
01:48 PM andypugh: (Or posibly not, I don’t know if Norbert has added the code that tells an embedded tab that it has focus)
01:48 PM rene_dev_: I dont have a touchscreen
01:49 PM rene_dev_: currently Im using axis, but would like to switch to gmoccapy
01:49 PM pcw_mesa: andypugh: did you ever get LinuxCNC running on your RPI?
01:49 PM rene_dev_: I dont like touchy
01:49 PM andypugh: pcw_mesa: No, other projects got in the way.
01:50 PM pcw_mesa: I know how that is...
02:01 PM jepler: sigh I really don't have the bandwidth, as they say, to engage in this discussion that chris morley started. Plus I kinda promised to butt out of the role of naysayer.
02:17 PM CaptHindsight: I'm happy to listen to opinions, I never have to ask for them
02:19 PM sync: andypugh: I also think the radius thing in the macros is somehow broken or did you change something rene_dev_?
02:20 PM andypugh: jepler: So if you were to have an opinion, it would be “ney”?
02:20 PM rene_dev_: yes, thats broken
02:20 PM rene_dev_: or we dont understand it
02:20 PM andypugh: Radius works for me
02:21 PM jepler: some MK ideas are sure to be worth poaching. Their code is less likely to be applicable. The posts in that thread that I've read are far too vague on details for what Chris Morley wants to do, so it's hard to give concrete reasons for "no".
02:21 PM andypugh: What is broken for you?
02:21 PM jepler: we had "create component instances anytime" once, we took it out because of a minor bug AFAIR.
02:22 PM jepler: we can have "broken multithreaded realtime because nobody looks at the real issues but just ignores them" easily
02:22 PM andypugh: Yes, I want to steal the MK GPIO driver for the RaspPi
02:22 PM jepler: (both of those are much easier with uspace realtime fwiw)
02:22 PM rene_dev_: uspace realtime is the future imho
02:22 PM jepler: john kasunich ported the mk bb gpio driver with little effort a few years ago
02:22 PM rene_dev_: because ethernet is the one and only way to go
02:23 PM jepler: we can do work to modularize, i.e., move hal into a separate git repo. but will it get used? after all, we have 100% support for a decade for building out of tree HAL components and LinuxCNC UIs, but nobody uses it
02:23 PM jepler: (it's underdocumented)
02:23 PM jepler: get the main benefit by taking a weekend to split the linuxcnc binary package up usefully
02:23 PM jepler: etc etc
02:24 PM sync: andypugh: it did not retract in the right direction
02:24 PM sync: I wanted to make an outside radius and it would retract *into* the work
02:25 PM andypugh: sync: Did you select the correct radio button?
02:25 PM andypugh: And where was the tool tip when you started the macro?
02:26 PM andypugh: (radius starts at the starting Z and ends at the starting X, so you need to be +X and +Z of the corner before starting)
02:27 PM sync: I think the tool tips were broken in that screen
02:27 PM sync: but yes I had the right radio button clicked
02:27 PM andypugh: jepler: This is probably my thinking, it’s a nice idea, but a lot of developer work for two potential users.
02:31 PM sync: yes, I positioned the tool right on the edge of the work in both axis
02:31 PM sync: and it turned a radius correctly
02:31 PM sync: just retracted into the work
02:31 PM CaptHindsight: using the Vulkan API is my real only next interest
02:31 PM CaptHindsight: and thats just for being able to support high res vs wire frames in a GUI
02:36 PM CaptHindsight: I'd likely be the only one to use it so it would sit on github next to the RTAI fork
02:42 PM jepler: "newinst" was limited to things without loadrt flags, which was a big limitation. with uspace, we could potentially reuse the same logic that we are parsing loadrt flags with
02:48 PM jepler: everybody's too tired to rewrite axis/gremlin for merely 8 (12?) years old OpenGL, who's going to come down from heaven to give us Vulkan support?
02:50 PM JT-Shop: Vulkan looks pretty good
02:52 PM CaptHindsight: JT-Shop: are you working with hazzy-dev on a new GUI?
02:53 PM JT-Shop: no, I'm working on configuration tools for 2.8
02:53 PM JT-Shop: we just share programming problems with each other
02:54 PM CaptHindsight: working/sharing
02:55 PM CaptHindsight: I don't really "need" Vulkan for my next project but it might fund it
02:57 PM CaptHindsight: lots of little Linux projects that need a thing or two that eat up time and energy
03:01 PM * JT-Shop thinks the welds have cooled enough to do the next step
03:02 PM mozmck: jepler: I seem to remember there were some license issues hindering ZeroMQ from being used instead of NML - maybe gpl3?
03:06 PM CaptHindsight: From our Trusted Partners
03:06 PM CaptHindsight: Miru (Hong Kong) maintains the OpenPGM multicast library. Areas of expertise: ØMQ+OpenPGM. Contact Steven McCoy.
03:06 PM CaptHindsight: Bear Metal (Europe) is a core contributor to ZeroMQ. Areas of expertise: Monitoring, custom bindings and transports as well as anything Ruby specific. Contact Lourens Naude.
03:06 PM CaptHindsight: Dual Licensing not Required
03:06 PM CaptHindsight: ØMQ is available under the LGPLv3 with a static linking exception including for iPhone apps. We do not offer a commercial license alternative.
03:07 PM CaptHindsight: oopps only meant to past the last line
03:11 PM mozmck: Yeah, seems like the LGPLv3 and GPL2 were not compatible or something like that. I'll have to go back and find the discussions I saw on that. Or maybe it was a different library...
04:43 PM Jin^eLD is now known as Jin|away
04:51 PM jepler: GPL2-only is sure a problem that linuxcnc faces
04:52 PM jepler: but we have one ex developer who has stated in private he will not consent to relicense as gpl3 or gpl2+
04:52 PM jepler: which puts us in a tough spot
04:56 PM andypugh: How much of his work remains?
04:57 PM andypugh: It poses an interesting question about relicensing work from developers who are, for example, dead..
04:57 PM jepler: it's hard to say
04:58 PM andypugh: (reading that back, I was not proposing a radical solution to our licencing problem)
04:58 PM jepler: Indeed. It's one of the reasons some projects choose copyright assignment that are NOT evil
09:17 PM jepler: sigh, so much use of strcpy / strcat
10:11 PM jepler: well that ended up more work than I'd imagined at the outset, for less real impact