Mick I hope you don't mind I pushed this back to the list.
I too was hoping to make inroads for modularity, future proofing, and lowering the bar for future developers. I'm not an expert of course but the multi core approach did seem to make sense for the future. It was also conveniently already done and already modular. If we had made our HAL similarly modular then one could conceivably used either library. I also thought that by using a HAL library maintained by someone else, the small amount of active developers could concentrate on other things, while still expanding the utility of HAL. It seems, at least on the linuxcnc side there is little interest in the work/idea of using MK's HAL. There are multiple reasons that i draw this conclusion, but it doesn't really matter. I do appreciate, everyone's participation in the discussion. I feel it's good to stop and ask these kinds of questions from time to time. Chris M ________________________________ From: schoone...@gmail.com <schoone...@gmail.com> Sent: September 14, 2018 5:19 PM To: EMC developers; Chris Morley Subject: Re: [Emc-developers] Breakout of HAL/ machinekits's HAL For some reason this did not go and has not hit the list Chris I have no interest in evangelising Machinekit, it is there, if people don't like it or are in other ways resistant, I'm not bothered. I was merely outlining a possibility. The big issue regards the CNC stack IMHO is not whether we need joints/axes, but getting rid of NML and making the whole CNC stack more modular. NML was written by C programmers operating on single core machines with linear programs, who never expected anything else to exist. It is littered with global variables, structures and many other things that make any sort of OOP impossible as it stands. The assumptions are similar to those in components we changed in Machinekit, where for example, data exchange between threads was reliant upon an assumed model of preemption giving priority to one thread and it being allowed to complete before the other one began. Michael showed that they fell apart when exposed to cached multi-cored processors and started producing some very strange results. If you want any help, contact me, but you know your own code base better than me. Machinekit is GPL2 AFAIK, ie. there is no binary stipulation. The problem with GPL2+ as I see it, is it allows later bad versions to supercede it, at the users arbitration of what version is most advantageous to them. (Linus certainly seems to think GPL3 is bad) Good luck Mick From: Chris Morley <chrisinnana...@hotmail.com><mailto:chrisinnana...@hotmail.com> To: EMC developers <emc-developers@lists.sourceforge.net><mailto:emc-developers@lists.sourceforge.net> Subject: Re: [Emc-developers] Breakout of HAL/ machinekits's HAL Message-ID: <by2pr05mb224801f1ed450b05292c91e2c0...@by2pr05mb2248.namprd05.prod.outlook.com><mailto:by2pr05mb224801f1ed450b05292c91e2c0...@by2pr05mb2248.namprd05.prod.outlook.com> Content-Type: text/plain; charset="us-ascii" We are getting off topic here.... I was suggesting we separate our HAL and cnc stack with the idea of possibly using MK HAL. NML is part of the cnc stack. That would be future work, I'd someone wanted to tackle it. Though just the act of splitting hal from CNC without adopting anything would make future work easier. Chris M Can you tell us about license of MK HAL work. AFAIK Linuxcnc is GPL2 for some and GPL2+ for most(?) _______________________________________________ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
