Newbie migrating a gantry from Mach3 to LinuxCNC
30 Oct 2019 02:41 #149105
by Bats
That, and rewriting old broken stuff with decades of accumulated cruft isn't nearly as much fun as working on new stuff. Especially when you know you'll inevitably catch flack for breaking something else that happens to depend on the broken cruft staying broken.
Of course, another option might be to just pile on a bit more cruft - leave stepconf more or less as-is and toss together a module/hack at the end (or even an independent script) that just automates the manual edits on the stepconf-created config files. It's a bit of a bandaid, horribly inelegant, and I don't know whether it would be practical for arrangements more complicated than XYYZ, but it could at least smooth out some of the bumps in the setup process for gantry users.
(And now's the point where I start feeling reeeally awkward because I just suggested an "easy" way to improve a process I have yet to even get working, on a system I have a demonstratedly poor understanding of, and requiring "simple" skills that I'm not at all sure I still possess. Maybe everyone's better off if I just try whining until someone decides to "fix" me
)
I got the ini & hal files edited to add the fourth joint, and eventually got it running (after discovering the howto had left out a key edit). It pops an "Unexpected realtime delay on task 0" error, but does let me jog the X, Y, and Z on the machine. The only problem (aside from the latency error, which can wait until the fundamentals are working) is that the fourth axis - set up as a second X for my convenience, due to machine orientation - doesn't move, or show any activity on the step or dir lines.
-Bats
Replied by Bats on topic Newbie migrating a gantry from Mach3 to LinuxCNC
I think that gets to the heart of it - and it's sort of inherent in the open source model. Nothing ever gets done until it annoys (or interests) someone enough to figure out how to do it & put in the "awful lot of work" for everyone else. Of course, from an enduser (rather than developer/contributer) perspective, that tends to get interpreted as either "learn to code" or "pester coders until you become the thing that's so annoying it needs to be fixed", either of which can be a little offputting (and I say that having spent time on all three sides of the problem).The problem is that the Linuxcnc core has been updated but Stepconf hasn't. I've long maintained that stepconf and pncconf should be updated to speak joints and just have a field where you key in the string XYYZ or whatever you want. But this breaks decades of behaviour and would involve an awful lot of work rewriting the config tools and nobody hears me.
That, and rewriting old broken stuff with decades of accumulated cruft isn't nearly as much fun as working on new stuff. Especially when you know you'll inevitably catch flack for breaking something else that happens to depend on the broken cruft staying broken.
Of course, another option might be to just pile on a bit more cruft - leave stepconf more or less as-is and toss together a module/hack at the end (or even an independent script) that just automates the manual edits on the stepconf-created config files. It's a bit of a bandaid, horribly inelegant, and I don't know whether it would be practical for arrangements more complicated than XYYZ, but it could at least smooth out some of the bumps in the setup process for gantry users.
(And now's the point where I start feeling reeeally awkward because I just suggested an "easy" way to improve a process I have yet to even get working, on a system I have a demonstratedly poor understanding of, and requiring "simple" skills that I'm not at all sure I still possess. Maybe everyone's better off if I just try whining until someone decides to "fix" me
)I have to admit, I was actually a little shocked by the process to set up a gantry system - not because the config files are particularly obtuse in themselves (although I won't pretend to understand most of what the .hal file is doing), but just because such a prevalent class of machines essentially require hand-tweaked setups on a devel branch of the software. I guess I expected it to be long-settled territory by now.But people like you and I have only ever used the joint model.....
Yup - and that's what I've been doing. I'd just started out looking at it in hopes that starting with XYZA might offer some buried options to change it to XYZ[n]. So far it's going at least moderately well.Whether an axis is linear or rotary is controlled by a setting in the ini file so it sounds like you are about to throw out stepconf and learn to edit hal and ini files from this point on.[
I got the ini & hal files edited to add the fourth joint, and eventually got it running (after discovering the howto had left out a key edit). It pops an "Unexpected realtime delay on task 0" error, but does let me jog the X, Y, and Z on the machine. The only problem (aside from the latency error, which can wait until the fundamentals are working) is that the fourth axis - set up as a second X for my convenience, due to machine orientation - doesn't move, or show any activity on the step or dir lines.
Thanks for the tip. The trick was finding the error in the first place, in the 1200 or so lines the crash dialog helpfully provides. The command line had a significantly more concise output.It is much better to open a terminal window and run linuxcnc from the command line (just type linuxcnc). Then the errors will be displayed on the console so debugging is much easier. The error is there, sometimes its cryptic to understand at the beginning but you will get better.
If someone's really bored and trying to hunt down the bug I could probably write up a quick repro (since it came from working entirely with unedited stock configs), but otherwise it's not really relevant - it was a result of doing things that didn't need to be done using a config file I shouldn't have been using in the first place - but I'm attaching the ini & hal files. I've still got more fiddling left to do on my end before giving up, but someone else may well spot a glaring error that I'm still days away from actually understanding.Copy and paste the errors to a text file and upload it here and someone will help. Also post your ini files and hal files
-Bats
Please Log in or Create an account to join the conversation.
30 Oct 2019 04:30 #149114
by Bats
I've attached the full report, but it looks like this is the meat of the problem:
Point taken, though.
In my case, I definitely haven't ruled out going the Mesa route, but, like with the Smoothstepper and Mach3, I didn't like the idea of investing in a hardware interface that would lock me into one particular software package - especially without some time actually running LinuxCNC first.
-Bats
Replied by Bats on topic Newbie migrating a gantry from Mach3 to LinuxCNC
If said someone is actually interested, I have no aversion to helping troubleshoot - I just didn't want anyone to waste their time thinking it was anything I needed a solution for. Although it looks like the significantly more concise output in this case is significantly less concise than on my other crashes - gmoccapy wanted to make a bit of noise on stdout, too.That significantly more concise output from the command line would be very helpful for somebody who is not really bored, does not have time to speculate on what might be wrong in your config files but still wants to help to zero in on the fault.
I've attached the full report, but it looks like this is the meat of the problem:
Note: Using POSIX realtime
/usr/bin/gmoccapy:325: GtkWarning: Invalid icon size 48
self.widgets.window1.show()
(gmoccapy:3683): GtkSourceView-CRITICAL **: gtk_source_language_manager_set_search_path: assertion 'lm->priv->ids == NULL' failed
pyvcp_options.hal:8: Pin 'pyvcp.spindle-speed' does not exist
Note: Using POSIX realtimeI suppose I've always had something of a weak spot for the obsolete - I think the first open source project I ever contributed to was lynx, while everyone else wanted to play with bright shiny Mosaic. Or maybe the obsolete just sort of magnetically accumulates around me - it's hard to tell, sometimes.I suspect that most gantries today are set up with Mesa hardware because the parallel port is pretty much obsolete on PC's today and therefore there is a reluctance to add features to stepconf to support such obsolete hardware.
Unless what I want is to connect via a parallel port.The mesa tools handle what you want.
Point taken, though.In my case, I definitely haven't ruled out going the Mesa route, but, like with the Smoothstepper and Mach3, I didn't like the idea of investing in a hardware interface that would lock me into one particular software package - especially without some time actually running LinuxCNC first.
-Bats
Please Log in or Create an account to join the conversation.
Time to create page: 0.108 seconds