#linuxcnc-devel Logs

Jun 12 2018

#linuxcnc-devel Calendar

09:25 AM JT-Shop: is there any way to have an index of the linuxcnc.org site on the main page? so folks can find dists and older versions
09:47 AM narogon: using debian stretch
09:47 AM narogon: in axis.py
09:47 AM narogon: inside the function jogspeed_incremental
09:47 AM narogon: the line ---- cursel = root_window.call(widgets.jogincr._w, "curselection")
09:48 AM narogon: root.windows... returns a tuple
09:48 AM narogon: with the first element a number and the second a empty one ex. (0,)
09:48 AM narogon: some lines below that there is a cursel = int(cursel)
09:49 AM narogon: that fails due to cursel isn't a number or string
09:49 AM narogon: I've solved the problem replacing to cursel = root_window.call(widgets.jogincr._w, "curselection")[0]
10:28 AM cradek: looks like this is bug 268, was fixed last year. Is your linuxcnc version super old?
10:30 AM cradek: https://github.com/LinuxCNC/linuxcnc/issues/268
10:30 AM cradek: does this match your bug?
10:34 AM cradek: looks like it was fixed in master and 2.7, but not 2.6
10:34 AM cradek: https://github.com/LinuxCNC/linuxcnc/commit/e583d370
10:51 AM seb_kuzminsky: according to this, we never supported 2.6 on Stretch: http://wiki.linuxcnc.org/cgi-bin/wiki.pl?MinimumSoftwareVersions
11:20 AM narogon: I'm using 2.7.11
11:20 AM narogon: it only happens with i, I to change jog increments, the rest of keys work correctly
11:21 AM narogon: at least the most frequents
11:27 AM seb_kuzminsky: narogon: it was fixed in 2.7.12
11:28 AM seb_kuzminsky: in general, if you're running a stable version of linuxcnc (ie, one without "-pre" in the version number), you can safely upgrade to the latest version of that stable series any time
11:29 AM seb_kuzminsky: we try really hard not to introduce new bugs in the stable branches, only to fix them :-)
11:38 AM narogon: seb_kuzminsky I'm only inform in order to improve!!
11:38 AM narogon: I don't complain, i know how well your work is
11:39 AM seb_kuzminsky: i didn't think you were complaining :-)
11:39 AM seb_kuzminsky: i only meant to encourage you to feel safe about upgrading often, to stay ahead of bugs like these
11:40 AM seb_kuzminsky: thanks for helping linuxcnc improve
11:40 AM narogon: I can't upgrade my version due to my own changes
11:40 AM narogon: at sometime i should work correctly with github
11:41 AM narogon: and let my version be upgraded with the main line of upgrades
11:41 AM narogon: but for now I don't do it well
11:41 AM narogon: hehehe
11:42 AM narogon: for example i'm using this patch from sittner
11:42 AM narogon: to make it works better with rt-preemt and ethercat
11:42 AM narogon: https://github.com/sittner/linuxcnc-ethercat/blob/master/patches/add-task-pll-functions.patch
11:44 AM seb_kuzminsky: narogon: ah, i see, yes that makes things harder to manage
11:49 AM seb_kuzminsky: i don't see any issues or PRs mentioning ethercat or task pll
11:50 AM seb_kuzminsky: i know we've talked about ethercat on the mailing lists, and the last i heard about that was the ethercat source code someone proposed has a license that prevented us from redistributing it
11:51 AM seb_kuzminsky: but the linuxcnc infrastructure needed to make it work better might still be possible to include (if it doesn't mess other things up)
12:19 PM narogon: this patch if there isn't any problem could be a good improvent to include
12:19 PM narogon: it creates a control of the cycle time
12:19 PM narogon: making it more constant
12:20 PM seb_kuzminsky: it looks like it tries to make the thread wake up synced to an external signal?
12:22 PM seb_kuzminsky: who calls pll_set_correction()?
12:29 PM jepler: forum kernel update coming up, small downtime
01:04 PM narogon: a component, lcec write
01:04 PM narogon: and it checks the deviation of the cycle time with a distributed clock of one of the slaves
01:05 PM seb_kuzminsky: and tweaks the pll correction to try to get the linuxcnc task to start synced with the slave's cycle?
01:28 PM narogon: yes
01:29 PM narogon: in ethercat is very important the master (linuxcnc) cycle is connected to the slaves
01:34 PM narogon: so each cycle the distributed clock (most frequently the first slave) send its clock time in order to resync all the elements
01:50 PM seb_kuzminsky: do you have a link to the component that lcec wrote that manages linuxcnc's cycle time using input from the distributed clock from the slave?
01:51 PM rene-dev_: seb_kuzminsky I dont see how that breaks anything, when the function isnt called
01:51 PM rene-dev_: https://github.com/sittner/linuxcnc-ethercat/blob/407698b6d9eb3bd0e964370dd8c92cbc04be91b4/src/lcec_main.c#L1317
01:52 PM rene-dev_: thats really clever. I wonder if pcw_home ever thought about using that.
01:53 PM seb_kuzminsky: i agree that addition probably won't break any existing systems, and that's very good
01:54 PM rene-dev_: we have similar issues with the stmbl, and thought about having a pll in the drive that tracks linuxcnc, but that just doest make sense
01:55 PM narogon: yes
01:55 PM narogon: sittner has it in github
01:56 PM narogon: https://github.com/sittner/linuxcnc-ethercat
01:56 PM seb_kuzminsky: it would be good if this feature came with a clear example of how to use it, and maybe documentation
01:56 PM narogon: https://github.com/sittner/linuxcnc-ethercat/blob/master/src/lcec_main.c
01:56 PM seb_kuzminsky: i see it, that's cool
01:57 PM seb_kuzminsky: i wonder if it would make sense to make the pll correction a hal pin on the thread object? rather than do it with a standalone component and an rtapi call
01:59 PM seb_kuzminsky: each thread currently has a .time pin and a .tmax param that both get updated every time the thread runs
02:01 PM narogon: this component is what receive the slave time, and manage all the network
02:01 PM seb_kuzminsky: might make sense to move lcec's .pll_* pins to the thread, and do that update in hal_lib.c:thread_task()
02:02 PM seb_kuzminsky: narogon: right, i'm not proposing changing that, only how the pll offset gets into linuxcnc
02:02 PM seb_kuzminsky: i'm suggesting embedding it more deeply into the guts of linuxcnc, via HAL pins instead of an rtapi function
02:03 PM narogon: it could be so interesting you could talk it with sittner
02:03 PM seb_kuzminsky: lcec would stil lget the slave time, but it would just put the time info on those hal pins (instead of doing the math and calling into rtapi)
02:03 PM seb_kuzminsky: then the hal thread would do the math itself and adjust its timing as indicated
02:04 PM narogon: I put him in the correct direction but he was the one who programs everything. https://github.com/sittner/linuxcnc-ethercat/issues/45
02:04 PM narogon: It gets out of my capabilities
06:04 PM andypugh: cradek: The above is not particularly important to me, as I am actually working on not needing to home, and clearly have not felt the need to unhome until now when I am testing homing behaviour.
09:50 PM jepler: the new kunena version has "18 enchaments". Shall we upgrade the forum again?
09:51 PM jepler: here we go
10:07 PM skunkworks: jepler _ chr
10:07 PM skunkworks: jeeze
10:09 PM skunkworks: jepler - chrome mobile seem to be missing the search/new message menu
10:11 PM skunkworks: it used to be a burger menu
10:11 PM skunkworks: *hamburger
10:32 PM jepler: skunkworks: bummer
10:34 PM jepler: I agree, it reproduces for me in FF desktop while using responsive testing mode
10:34 PM jepler: your device has to be 980px across for that bit to appear :-/
10:43 PM jepler: https://github.com/Kunena/Kunena-Forum/commit/8af33b790412035f8f9b57218cd7f998c794a955 they screwed it up on purpose
10:46 PM jepler: and it's designed so that any changes will be overwritten on update so it's useless to touch it
10:46 PM jepler: OK I changed it but next update it'll be ruined again because they're useless
10:46 PM jepler: just remind me to edit com_kunena/template/crypsis/layouts/widget/menubar/default.php