#linuxcnc-devel Logs

Feb 02 2022

#linuxcnc-devel Calendar

04:19 AM -!- #linuxcnc-devel mode set to +v by ChanServ
10:12 AM Guest73: good morning all
10:13 AM Guest73: I am not sure if it has been reported, but the status appears to have a bug in it on both the stable and master branches
10:14 AM Guest73: this is causing some issues with QtPyVCP backplotter on program loading
10:15 AM pere: Guest73: to figure out if it is reported, visit <URL: https://github.com/LinuxCNC/linuxcnc > and check issues and pull requests.
10:16 AM Guest73: just checked does not appear reported, i will report there
11:20 AM Guest73: i have added an issue report to the linuxcnc github page.  this is critically effecting QtPyVCP interfaces, the 2.8.2 version is the worst and is what most people are reporting on as it is most widely installed with probe Basic interface.
12:14 PM -!- #linuxcnc-devel mode set to +v by ChanServ
12:53 PM alex_jon1: hi all
12:53 PM alex_jon1: anyone knows where I can find an online repository with the emc1 code?
12:56 PM alex_jon1 is now known as alex_joni
12:59 PM alex_joni: iirc we had a rcslib and emc1 as separate projects in CVS
01:11 PM alex_joni: cradek: do you know if the old emc1 repository was migrated to github?
01:22 PM JT-Shop: hmm https://github.com/LinuxCNC 500 error
01:23 PM alex_joni: https://github.com/linuxcnc
01:23 PM JT-Shop: same 500 error for me
01:23 PM Tom_L: unless someone is archiving it i doubt you find emc1
01:24 PM Tom_L: the nist site used to have source but i can't find it
01:24 PM Tom_L: what's that archive site?
01:25 PM alex_joni: archive.org ?
01:27 PM Tom_L: yes possibly archived there
01:28 PM JT-Shop: Jon Elson has a copy
01:29 PM Tom_L: yeah he would
01:30 PM JT-Shop: http://bdi4emc.ourproject.org/Where_to_get_it.html
01:30 PM JT-Shop: I have a bdi cd somewhere
01:32 PM Tom_L: http://www.cncgear.com/EMC/BDI/
01:32 PM JT-Shop: http://www.cncgear.com/EMC/BDI/
01:32 PM Tom_L: yes
01:33 PM Tom_L: http://www.cncgear.com/EMC/BDI/Dancing_Penguin.gif
01:33 PM Tom_L: heh
01:33 PM JT-Shop: wonder why we still don't have that?
01:40 PM alex_joni: unfortunately the BDI has only the runtime deb, no sources
01:41 PM * alex_joni tries the bdi-src iso
01:42 PM JT-Shop: do you want me to email Jon and see if he still has the files?
01:43 PM alex_joni: not yet.. I still have a couple of locations to look myself ;)
01:43 PM JT-Shop: ok
01:48 PM alex_joni: found a local repository, but without cvs history obviously
01:48 PM JT-Shop: col
01:49 PM alex_joni: /rs274ngc_new_sim.ini/1.8/Thu Jan 12 07:54:35 2006//
01:51 PM ve7it: https://ve7it.cowlug.org/emc.html has old build inst with cvs commands to get sources from soureforge.... tried them, but connection seems to timeout... your browser will complain on the website... it is my site with self signed certs, not a commercial server
01:52 PM alex_joni: I think the code is stil there: http://emc.cvs.sourceforge.net/
01:57 PM ve7it: hi alex... I can confirm the links you pointed to do permit download of emc and rcslib
01:58 PM roycroft: ve7it: you probably should not use self-signed certs on any website that in public ip space
01:58 PM roycroft: browsers are getting more and more militant about refusing to connect to sites they don't trust
01:59 PM roycroft: if you're in public ip space, letsencrypt are pretty easy to install, and automatically renew if you set up a simple cron job to do the renewals
01:59 PM ve7it: no option as as my isp would have to provide the certs.... site is mostly notes to myself
02:00 PM roycroft: why would the isp have to provide them?
02:01 PM roycroft: and i'm not being argumentive here, nor telling you what to do
02:02 PM roycroft: i'm just pointing out that if the site is accessible to the public, most folks will have problems connecting to it
02:02 PM ve7it: although I have a fellow providing the dns pointer to ve7it.cowlug.org (a linux club site), the actual server is one of my home boxes and the reverse lookup points to my isp's internet modem
02:02 PM roycroft: that's ok
02:03 PM roycroft: letsencrypt can still work - you just don't do dns verification
02:03 PM ve7it: ok.. will look at that agin
02:03 PM roycroft: you can tell it to scribble a file in your documentroot, and if it can read the file it verifies the site and issues the cert
02:03 PM ve7it: thanks
02:03 PM roycroft: no problem
02:03 PM alex_joni: ve7it: thanks for checking, was doing the same here
02:03 PM roycroft: and for the record, i'm not a fan of "all websites must be encrypted"
02:04 PM JT-Shop: alex_joni, I wonder why all the file names have ,v at the end
02:04 PM roycroft: but The Internet(tm) has decreed that it shall be so, and so it now is
02:04 PM JT-Shop: rsync -ai a.cvs.sourceforge.net::cvsroot/emc/ /my/local/dest/dir/
02:04 PM JT-Shop: that did download all of emc1
02:04 PM roycroft: jt-shop: ,v files are version control files
02:04 PM roycroft: for rcs
02:05 PM JT-Shop: they came from cvs on sourceforge
02:05 PM roycroft: cvs is a front-end to rcs
02:06 PM JT-Shop: ah ok so when I did a rsync it got me the raw rcs files?
02:06 PM ve7it: cvs -z3 -d:pserver:anonymous@a.cvs.sourceforge.net:/cvsroot/emc co -P rcslib emc works as well
02:06 PM roycroft: that would be the proper way to do it
02:06 PM roycroft: that's kind of the cvs equivalent of "git clone"
02:10 PM roycroft: foo.c,v is the file foo.c, with complete revision history
02:10 PM roycroft: rlog foo.c will let you view the revision history
02:10 PM roycroft: co -u foo.c will check out the current version without a lock
02:11 PM roycroft: you may have to install rcs to have those commands - it's not commonly installed anymore
02:11 PM roycroft: like ever since git took over the world
02:11 PM * roycroft still uses rcs every day, and git every day
02:11 PM JT-Shop: I may try that in the morning, don't want to download too much
02:12 PM roycroft: apt -y install rcs will get you the basic tools
02:12 PM JT-Shop: thanks
02:14 PM alex_joni: whoa, browsing these sources brings back some ancient memories
02:15 PM * alex_joni wonders who remembers RTHAL and ADEOS
02:15 PM JT-Shop: yeah
03:27 PM -!- #linuxcnc-devel mode set to +v by tantalum.libera.chat
03:29 PM -!- #linuxcnc-devel mode set to +vv by tantalum.libera.chat
04:19 PM Tom_L: alex_joni, wondering what your interest is in emc1?
04:27 PM lcvette: Anyone have any ideas about the Status issue?
04:28 PM lcvette: https://github.com/LinuxCNC/linuxcnc/issues/1564
04:32 PM andypugh: lcvette: It’s possible that a commit search in “state tags” might explain the changes?
04:33 PM andypugh: Otherwise, do you have the time for a git bisect? (it can be really easy if you can come up with a really quick and easy test for what is wrong)
04:43 PM lcvette: sure
04:44 PM lcvette: i'm will to help however is needed
04:44 PM lcvette: but im not a coder, im the design guy
04:44 PM lcvette: lol
04:45 PM lcvette: andypugh ^^^
04:45 PM andypugh: If you can build from source then it’s basicaly automatic.
04:46 PM lcvette: :o  not sure how to build from source
04:47 PM lcvette: i did go through the recent changes and saw some updates to alot of M code to M-code and looked like it removed alot of white spaces
04:48 PM andypugh: OK. Do you know when it was good, and when it is bad, and a very simple test to determine the status at a point in the history?
04:48 PM lcvette: the test yes, programs changed the status
04:48 PM Tom_L: they're not doing the white space for the docs are they?
04:49 PM lcvette: no it was in python code
04:51 PM andypugh: lcvette: Well, I can run the bisect here, though I would prefer you to do it :-)
04:54 PM andypugh: But: Reading the problem description. I think that the Axis satsus window actualy shows the status at the end of the preview stage, not live as the code runs.
04:54 PM lcvette: no it used to be active
04:55 PM lcvette: and some are changing
04:55 PM andypugh: So Axis is not necessarily a good way to monitor the QtPyVCP issue referenced.
04:55 PM lcvette: but not all
04:55 PM lcvette: and for instance, 2.9 the gcodes and some mcodes are updating while running the program, but in 2.8 they are not at all
04:55 PM lcvette: and this used to work perfectly fine prviously
04:56 PM andypugh: Ah, so it is _fixed_ in 2.9?
04:56 PM lcvette: no
04:56 PM lcvette: not completely
04:56 PM lcvette: and it wasn't broken maybe a couple weeks ago and beforre that its not been an issue at all
04:56 PM andypugh: Hmm. That’s awkward
04:57 PM lcvette: in 2.9 M3 M4 don't seem to want to update
04:57 PM lcvette: but M8 M9 do
04:57 PM lcvette:  it appears from my testing that most of the gcodes do update in 2.9
04:57 PM lcvette: in 2.8 nothing is updating it appears
04:57 PM andypugh: Do you have a linuxcnc-dev and git and the rest? I suspect not?
04:58 PM lcvette: all my devs are setup for working on qtpyvcp
04:58 PM lcvette: o no dev giot for linuxcnc
04:58 PM lcvette: i typically just work over a regular iso install for 2.8.2
04:59 PM lcvette: and use Joco's 2.9 bullseye builder for the py3 dev work
04:59 PM lcvette: im truly sorry i don't know more code stuff or I would gladly be helping with linuxcnc stuff also.
05:00 PM lcvette: i get myself in trouble with just the qtpyvcp stuff..lol
05:00 PM andypugh: It’s a bit late here now, but if you ould add a reference to when it definitely worked and when it definitely stopped working and a step-by-step check to determine good or bad then I can try to run the bisect.
05:00 PM andypugh: (ie, add to the bug report)
05:01 PM lcvette: its hard to nail that timeline down
05:01 PM lcvette: i have a machine that has not been updated that it still works fine on
05:01 PM andypugh: 2.7.15 good / 2.8.2 bad is enough.
05:01 PM lcvette: but its not been updated in maybe a year
05:01 PM lcvette: no idea on 2.7 never ran it
05:02 PM lcvette: qtpyvcp is preempt rt only
05:02 PM lcvette: so 2.8 and later
05:02 PM andypugh: Specific commits is better, saves a few iterations, buit major releases is fine
05:03 PM andypugh: So, Version on the good machine, version on the bad machine is fine.
05:03 PM lcvette: i would say this is a recent issue
05:03 PM lcvette: as in within the last month
05:04 PM lcvette: sek i did an install on another machine a few weeks ago that i haven't finished i will run and see if problem is shown there
05:04 PM andypugh: The main issue in running a bisect is finding a way to tell a good build from a bad one. After a while even homing a sim gets annoying...
05:21 PM lcvette: i just ran and tested all my machines in the shop and looked when they were updated all were updated within the last 4 weeks and all showed the same non updating status issue
05:22 PM lcvette: i will dig through whatever i can to try and find the issue
05:23 PM lcvette: but may be of note that i had the first report of it in early november 2021, but at the time we figured it was a video card compatibility issue with opengl
05:23 PM lcvette: but now all new installs are showing the same issue
05:33 PM lcvette: andypugh perhaps comparing that section from 2.9 to 2.8 for gcodes which update in 2.9 and not in 2.8 could be of some useful bug funding?
05:34 PM lcvette: im just not sure where to look for something like that in the linuxcnc git
05:35 PM andypugh: Do you still habe a machine that works, anywhere?
05:35 PM lcvette: i thought i did, but forgot i updated it from stretch to buster just before Christmas
05:36 PM andypugh: A guess and a good test would still work.
05:36 PM andypugh: (the good test is critical)
05:37 PM lcvette: the test i ran just now is to run the splash page Linuxcnc program with the status window opened
05:41 PM lcvette: i don't know if i can paste images here
05:41 PM lcvette: if you want i can paste to the linuxcnc chat
05:42 PM Tom_L: you're not on matrix here?
05:42 PM andypugh: You can’t, as far as I know
05:42 PM lcvette: i don't know how to get on the devel channel from matric
05:42 PM lcvette: matrix
05:42 PM Tom_L: search for linuxcnc-dev
05:42 PM andypugh: I just tried it, with the Axis GUI, and I see (for example) G1 and G0 toggling.
05:43 PM lcvette: 2.9?
05:43 PM andypugh: Is that good or bad status?
05:43 PM lcvette: in 2.8 it will not toggle at all during a program
05:43 PM lcvette: in 2.9 most toggle
05:43 PM lcvette: except M3 and M4
05:46 PM andypugh: I was looking at the status bo in the MDI tab. Is that the same?
05:47 PM lcvette: mdi works
05:47 PM andypugh: (status box)
05:47 PM Tom_L: and you found matrix!
05:47 PM Lcvette[m]: i don't know
05:47 PM andypugh: MDI tab, but running the splash screen.
05:47 PM Lcvette[m]: let em check
05:53 PM Lcvette[m]: so it looks like the mdi box is updating with gcodes F code, but not M3 M5
05:53 PM Lcvette[m]: its stuff in perpetual M5 mode
05:54 PM Lcvette[m]: 2.8.2 nothing updates at all
06:03 PM Lcvette[m]: M1, M3, M4, M5 do not appear to update in 2.9 mcode status
06:04 PM Lcvette[m]: i just rant tests checking the rest that were fairly easy to account for in the file and was getting updates in the mdi active G-codes box
08:01 PM cradek: andypugh: sorry I missed you. I feel like no matter what we do, as long as the code is floating point, there will be some crazy gcode that will cause technically wrong behavior. I just added zeroes to my .0000000001 until it broke and it seems like that's a thing that will always happen.
08:02 PM cradek: my only thought that I think makes sense is an ideal implementation will give an error when the gcode is crazy, instead of doing something unexpected. (I realize this is very hard)
08:04 PM cradek: about your patch specifically though: when I saw you got rid of the while()s I thought you might have broken multi-turns but in testing I couldn't actually find any breakage specific to that. P2 always added one full circle to whatever it was doing, P3 two circles, etc etc
08:05 PM cradek: I don't actually *understand* the patch though :-/