#linuxcnc-devel Logs

Feb 09 2023

12:40 AM -!- #linuxcnc-devel mode set to +v by ChanServ
08:26 AM -!- #linuxcnc-devel mode set to +v by ChanServ
03:52 PM pere: hi
03:58 PM CaptHindsight[m]: https://forum.linuxcnc.org/18-computer/48079-can-the-opi5-be-configured-to-run-lcnc?start=10#gallery-3
03:59 PM CaptHindsight[m]: https://forum.linuxcnc.org/18-computer/48079-can-the-opi5-be-configured-to-run-lcnc?start=10#gallery-2
03:59 PM CaptHindsight[m]: "running the test for 45min and still 13.9us on the base thread and 19.4us on the servo thread" this the Orangepi5
04:00 PM CaptHindsight[m]: http://www.orangepi.org/html/hardWare/computerAndMicrocontrollers/details/Orange-Pi-5.html
04:07 PM andypugh__: That’s pretty respectable
04:12 PM CaptHindsight[m]: it also has three integrated Cortex M0 micros
04:12 PM CaptHindsight[m]: if there are open specs we can try to get stepgens on them like the PRU's in the BBB
04:13 PM CaptHindsight[m]: OrangeCNC and Remora are examples
04:13 PM CaptHindsight[m]: more examples
04:13 PM CaptHindsight[m]: https://forum.linuxcnc.org/18-computer/48079-can-the-opi5-be-configured-to-run-lcnc
04:14 PM CaptHindsight[m]: here's the forum thread
04:26 PM seb_kuzminsky: that's rad
04:26 PM seb_kuzminsky: pere: hi, i think i almost have a new dsc ready, testing it now
04:26 PM seb_kuzminsky: are you available to upload it, if it passes my tests here?
04:27 PM CaptHindsight[m]: andypugh__: the Opi5 is available from 4-32GB of RAM, not sure who sells to the UK though
04:29 PM CaptHindsight[m]: I think the kernel was from some project related to this similar rk3588 board https://en.t-firefly.com/product/industry/rocrk3588spc
04:29 PM CaptHindsight[m]: it is very similar to the Opi5 but the pricing I have found has been higher than Rpi5 prices
04:30 PM andypugh__: I probably have enough SBCs. I have Pi2. Pi3. Pi4 x2, Pi400, Udoo, Pine64, Beaglebone, CNCPC, CHiP….
04:31 PM CaptHindsight[m]: I know the feeling
04:32 PM CaptHindsight[m]: the rk3588 boards were delayed about a year due to parts supply issues
04:32 PM CaptHindsight[m]: they should be around for the next 3-5 years
04:33 PM CaptHindsight[m]: I have seen about 4 main variations of rk3588 boards
04:34 PM seb_kuzminsky: but it needs a non-mainline kernel?
04:35 PM CaptHindsight[m]: seb_kuzminsky: I've lost track, it might be already in mainline
04:35 PM CaptHindsight[m]: or soon will be
04:35 PM CaptHindsight[m]: the parts shortage messed up kernel and firmware development
04:36 PM CaptHindsight[m]: it does use u-boot so we can also tweak that for best RT performance, unlike the Rpi's
04:40 PM CaptHindsight[m]: seb_kuzminsky: "very basic support for rk3588 has been merged for linux-next (so 6.1) but there’s no support for pcie related stuff on Rock5b, so no ethernet or other peripherals
04:40 PM CaptHindsight[m]: Don’t expect anything more than a bootable kernel before 6.2 and even then Rock5b is probably not the first board to have most stuff working"
04:41 PM CaptHindsight[m]: so later this year
04:41 PM CaptHindsight[m]: peripherals are way behind
04:42 PM CaptHindsight[m]: https://github.com/neggles/linux-quartz64/commits/linux-quartz64
05:22 PM andypugh__: seb_kuzminsky: I was sure that the Hostmot2 encoder kept the counts in an S64 accumulator, but someone on the forum has mentioned that his hobbing spindle wraps after a few hours, and it does rather look like counts is an S32.
05:26 PM andypugh__: the HAL encoder also has this problem (though it would take a long time to accumulate 2^31 counts in the base thread) but stepgen uses a long long accumulator.
05:27 PM andypugh__: The smart-serial encoder module, on the other hand, does store the total counts in a long long.
05:29 PM andypugh__: seb_kuzminsky: Now that we have 64-bit types in HAL (in master at least) the simple fix would be to switch the counts and rawcoiunts to these.
09:43 PM pere: seb_kuzminsky: yes. :)
09:44 PM seb_kuzminsky: sweet :-)
09:48 PM seb_kuzminsky: andypugh__: yeah, i like that idea
09:48 PM pere: andypugh__: so 64 bit hal finally have a use case. :)
10:09 PM pere: seb_kuzminsky: any timeline for me?
10:17 PM seb_kuzminsky: just finished!
10:17 PM seb_kuzminsky: http://highlab.com/~seb/debian/linuxcnc/2.9.0~pre1+git20230208.f1270d6ed7-1/
10:17 PM seb_kuzminsky: i disabled your d/patch since the m4 change is already in the main repo
10:18 PM seb_kuzminsky: but there was a second change, to src/configure.ac that's not in our git, also disabled, so i hope that's ok
10:18 PM seb_kuzminsky: it didn't look crucial, and it wasn't mentioned in the patch comments so maybe it was unintentional?
10:29 PM pere: looking
10:33 PM seb_kuzminsky: pere: ugh, i forgot to mark it released in d/changelog, hang on let me redo it real quick
10:36 PM seb_kuzminsky: ok, *now* it's ready
10:38 PM pere: 'dget http://highlab.com/~seb/debian/linuxcnc/2.9.0~pre1+git20230208.f1270d6ed7-1/linuxcnc_2.9.0~pre1+git20230208.f1270d6ed7-1.dsc' give me a 404?
10:39 PM seb_kuzminsky: hmm
10:39 PM pere: http://highlab.com/~seb/debian/linuxcnc/2.9.0~pre1+git20230208.f1270d6ed7-1/linuxcnc_2.9.0~pre1+git20230208.f1270d6ed7.orig.tar.xz is the problem.
10:41 PM seb_kuzminsky: fixed
10:51 PM pere: seb_kuzminsky: do not remember the configure.ac fix, but it seem to be useful.
10:51 PM pere: seb_kuzminsky: is it intentional to not close the NMU bug?
10:53 PM seb_kuzminsky: err, no, leaving the NMU bug open is not intentional
10:53 PM seb_kuzminsky: lemme look...
10:54 PM seb_kuzminsky: should i just add a line in d/changelog that says "Closes: #1023651"?
10:55 PM pere: yeah. I tend to put it next to 'new upstream version XXX.YYY' in my changelogs.
10:56 PM seb_kuzminsky: ok, hang on
10:57 PM seb_kuzminsky: that reminds me, i'll add the "new upstream version" too ;-)
10:57 PM pere: I thought I put everything in my patch into the upstream source. suspect this part as 420923f0d015f9f534d54141cff0cadf52bc66cf.
11:00 PM seb_kuzminsky: ok, try one more time please
11:01 PM seb_kuzminsky: ah, that's in master, not 2.9
11:02 PM pere: aha
11:05 PM pere: seb_kuzminsky: not sure it is required if the dependency is available.
11:05 PM seb_kuzminsky: it's been working without it, so i say lets ship it as is
11:11 PM pere: uploaded.
11:12 PM pere: seb_kuzminsky: is it an idea to share your upload git repo with the project to allow more people uploading?
11:12 PM seb_kuzminsky: yay, thank you!!
11:12 PM seb_kuzminsky: yeah, it should be on github under the linuxcnc organization, not my personal account
11:18 PM seb_kuzminsky: ok, done: https://github.com/LinuxCNC/linuxcnc-gbp
11:18 PM pere: very good.
11:18 PM seb_kuzminsky: i have some notes on how i use this repo, i should add them as debian/README.source
11:18 PM pere: yeah
11:19 PM pere: and if I am not around, smoe should be able to help with uploading.
11:19 PM seb_kuzminsky: also, i don't know if i'm using gbp correctly. it's a bit weird since linuxcnc.git has debian/, but linuxcnc-gbp.git wants different debian/
11:19 PM seb_kuzminsky: anyhow, thanks pere, i'm heading out for tonight
11:24 PM pere: thanks for the work, look forward to seing the autopkgtest results. :)