#linuxcnc-devel Logs

Jun 18 2019

12:03 AM Centurion_Dan: o/
12:06 AM Centurion_Dan: I made my first PR to linuxcnc.
12:06 AM Centurion_Dan: :-p
08:47 AM -!- #linuxcnc-devel mode set to +v by ChanServ
09:20 AM seb_kuzminsky: hi Centurion_Dan, thanks for the PR :-)
09:20 AM seb_kuzminsky: what problem did you run into while building linuxcnc using pbuilder?
09:22 AM mozmck: (04:21:50 PM) Centurion_Dan: mozmck: The issue I had was clean would run and fail on dependency issues before the dependency issue were resolved by pbuilder before the build step.
09:23 AM seb_kuzminsky: hmm
09:24 AM seb_kuzminsky: the buildbot runs (or has run) pbuilder on ubuntu hardy through precise, and debian wheezy through stretch, and on all those platforms pbuilder does not run clean before satisfying the build dependencies
09:26 AM seb_kuzminsky: the process is: (1) unpack the dsc, (2) satisfy the build deps, (3) debian/rules clean
09:26 AM mozmck: I didn't remember that you were using pbuilder, but I thought it odd that no one had mentioned a problem with it before.
09:27 AM mozmck: I asked him if pbuilder was different in devuan - but I sure wouldn't think it was.
09:28 AM seb_kuzminsky: this is when building a dsc (source package) into a deb (binary package), maybe he's talking about building a dsc from the "raw" source?
09:30 AM seb_kuzminsky: "pbuilder debuild" goes straight from a directory of source code to a .deb, we don't use this pbuilder feature, maybe that's where the issue comes from
09:32 AM Centurion_Dan: I'm working from a distribution perspective so it's got to build both source and binary packages, so yes, that would explain it.
09:32 AM mozmck: huh. I've never used pbuilder. Sounds like it's not a bad idea though! I just keep a virtualbox vm set up for my builds
09:32 AM seb_kuzminsky: we build our source packages using dpkg-buildpackage
09:33 AM seb_kuzminsky: mozmck: the great thing about pbuilder is that every build happens in a clean, minimal install of the target os, so you can verify your build dependencies
09:34 AM seb_kuzminsky: Centurion_Dan: our packaging infrastructure checks out the git repo, runs 'debian/configure', then 'dpkg-buildpackage -S', then 'pbuilder build', what do you do where you saw this problem?
09:36 AM Centurion_Dan: gbp buildpackage --git-pbuilder
09:37 AM Centurion_Dan: in this case I have to manually run the debian/configure script
09:38 AM Centurion_Dan: which wouldn't work from a distrobution packaging perspective..
09:39 AM seb_kuzminsky: yeah, the debian/configure step is awkward, sorry...
09:39 AM seb_kuzminsky: as i'm sure you see, our source code can be configured to be built & packaged for different distros, and the debian/configure script is what does that customization
09:42 AM seb_kuzminsky: my understanding of debian is that the thing you upload to the distro's package builder is the dsc, not the raw git repo or `git archive` tarball, so there's room for the package maintainer to run debian/configure before making the dsc
09:54 AM Centurion_Dan: seb_kuzminsky: Indeed, that is the case, but it's also done by a Debian Developer who would probably make changes anyway.
09:56 AM Centurion_Dan: For Devuan we have a slightly different process and actually build both source and package straight out of our git repo using jenkins which in turn uses pbuilder.
09:57 AM seb_kuzminsky: ah, gotcha
09:57 AM Centurion_Dan: I'm interested in trying to get linuxCNC into Devuan as I'm planning to use it, and for me that means extra people see it exists and are more likely to try it.
09:59 AM Centurion_Dan: BIggest issue is the lack of rtai or xenomai in Debian and thus Devuan (~95% of our packages are straight out of Debian - even downloaded out of Debians archive pool)
10:00 AM Centurion_Dan: although I'm sorely tempted to add rtai to Devuan as well ;-)
10:02 AM Centurion_Dan: seb_kuzminsky: If you think my patches are inappropriate or break things, I won't be in anyway put off or offended if they get rejected.
10:03 AM Centurion_Dan: It just means that my Devuan package will probably have a few more differences.
10:04 AM Centurion_Dan: Anyway I need some sleep... it's 3am here :-o
10:04 AM Centurion_Dan: o/
12:25 PM Tom_itx is now known as Tom_L
12:42 PM CaptHindsight: Centurion_Dan: if you want to build an RTAI package for Devuan you can use this https://github.com/NTULINUX/RTAI
12:43 PM CaptHindsight: Centurion_Dan: a newer preempt_rt kernel is another thing you'll have to build for Devuan
12:44 PM CaptHindsight: IIRC the libs to build Mesa in latest devuan are kind of old so you won't be able to have it for the latest GPU's
12:47 PM CaptHindsight: ^^ builds RTAI for systemD free distros like Gentoo all the time, so that should not be an issue
12:48 PM CaptHindsight: RTAI 5 is just too broken for anything to happen with it anytime soon
01:42 PM andypugh: pcw_home: Here’s an odd thing. I had a prblem with the 7i73 in my lathe saddle (it turned out to be that the RJ45 clip had broken, so the connector was loose. But in getting there I lost track of power-on / power off and shorted an input to 24V. So that killed a pair of 5V inputs. Luckily I have another 7i73 in the mill being under-used to provide only a single encoder counter for the front-panel mpg, so I c
01:42 PM andypugh: swap.
01:42 PM andypugh: But the funny thing is, the other 7i73 counts in the reverse direction!
01:43 PM andypugh: And I don’t want to take the machine apart again to swap wires. So I will do it in HAL.
01:43 PM andypugh: Any idea why this might be, though?
04:55 PM andypugh: Aha! I figured it out, it was the nvencmode that needed to be changed from 0 and 4 to 1 and 1.
07:48 PM Centurion_Dan: CaptHindsight: we have the _rt kernels from debian. RTAI looks like a serious hack on the kernel, and from a Debian (Devuan) perspective it appears to not be something that is easily packageable. Xenomai looks a little but not much better.
07:48 PM Centurion_Dan: For them to be acceptable for
07:53 PM Centurion_Dan: Debian/Devuan kernel builds they'd need to be packaged in the form of a linux kernel with patches that are applied over the stock linux kernel source, and probably a separate source package for the libs and tools.
09:43 PM seb_kuzminsky: Centurion_Dan: i did that packaging, but sadly the resulting kernel was not stable for our needs, and the RTAI folks were unable to fix it
09:44 PM seb_kuzminsky: https://github.com/SebKuzminsky/linux-rtai-build
09:44 PM seb_kuzminsky: there's links to the various git repos in the top-level makefile
10:39 PM CaptHindsight: Centurion_Dan: the rt kernels in devuan (from Debian) are a bit dated, not new enough for AMD hardware made this year
10:40 PM CaptHindsight: Centurion_Dan: if you want to package the latest stable RTAI then use https://github.com/NTULINUX/RTAI
10:45 PM CaptHindsight: Centurion_Dan: we have Sid running on a 5.x rt kernel
10:45 PM CaptHindsight: no reason why Devuan can't run one
10:57 PM seb_kuzminsky: CaptHindsight: the test that broke rtai.org's RTAI kernel was just running linuxcnc's "runtests" in a loop
10:57 PM seb_kuzminsky: eventually it would lock up while unloading the realtime modules
10:57 PM seb_kuzminsky: does memfrob's kernel survive that?
11:16 PM Centurion_Dan: CaptHindsight: Well if you run stable the that's to be expected. Did you look at the -rt from -backports in Debian/Devuan?
11:27 PM CaptHindsight: Centurion_Dan: no we just built a custom 5.x kernel to match our hardware
11:33 PM CaptHindsight: seb_kuzminsky: memleaks 3.16.52 RTAI was his latest RTAI