#linuxcnc-devel Logs

Jun 14 2019

04:16 AM Centurion_Dan: Hi, I'm interested in trying to build linuxCNC for Devuan... but that debian/rules makes building with pbuilder impossible....
05:33 AM Centurion_Dan: ah, fixed the clean rule and that resolved my build issue.
07:30 AM Tom_itx is now known as Tom_L
03:29 PM KimK: Centurion_Dan: Interesting, I was not aware of it. https://en.wikipedia.org/wiki/Devuan Currently #57 on DistroWatch: https://www.distrowatch.com/table.php?distribution=Devuan https://www.distrowatch.com/index.php?distribution=devuan&release=all&month=all&year=all More info: https://www.makeuseof.com/tag/debian-without-systemd-devuan/
03:32 PM KimK: Centurion_Dan: Does your change retain all existing compatibility, and if so, would you be willing to write a pull request? Thanks for working on LinuxCNC.
04:08 PM Centurion_Dan: KimK: Sure.
04:11 PM Centurion_Dan: The change just removes the autogen.sh part of the clean operation, and prefixes a hyphen to a few of the following operations so their failure doesn't cause the process to abort. Pretty trivial really.
04:11 PM Centurion_Dan: The reason for the change is that with pbuilder dependency installation comes after the clean clean stage, and it would fail in the clean stage running the configure process.
04:12 PM Centurion_Dan: everything built and appeared to work after that.
04:13 PM Centurion_Dan: To be honest, that rules generation could do with a bit of a tidy up and use debhelper to significantly simplify it.
04:14 PM KimK: Ha, maybe two PR's then? Thanks!
04:16 PM Centurion_Dan: yeah, the tidy up would take a bit of testing to ensure it retained the original functionality. I'll make a PR for the immediate fix and allocate some time to the tidy up later.
04:18 PM mozmck: I'm wondering why autogen was called before clean to start with?
04:18 PM Centurion_Dan: I also will add a PR for adding Devuan distribution support to debian/configure ;-)
04:19 PM Centurion_Dan: mozmck: it was called by clean.
04:19 PM mozmck: I know - but I wonder why? typically things like that had a reason to begin with
04:20 PM Centurion_Dan: I thought it was a bit bizarre... I suspect that it was because some of the clean commands failed if autogen hadn'
04:20 PM Centurion_Dan: t been run
04:20 PM mozmck: is pbuilder on devuan different than in debian? I have never used pbuilder
04:21 PM Centurion_Dan: that's why I prefixed the hyphens to the commands following the removed autogen
04:21 PM mozmck: That could be. Maybe it would be better if clean would just exit with an message saying autogen had to be run first.
04:23 PM Centurion_Dan: mozmck: yes, pbuilders the same as Debian. Devuan actually is ~97% Debian - with packages coming from debians archive pool, the others are just the necessary bits for a separate distribution - themes etc, and for reversing the hard systemd dependencies.
04:23 PM mozmck: Yeah, I've looked at devuan since it began but haven't played with it.
04:24 PM mozmck: I guess maybe no one has been using pbuilder then. I build packages with fakeroot debian/rules binary typically
04:25 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.
04:25 PM Centurion_Dan: gotta go. BBL
04:28 PM memfrob: Wow, I got RTAI 5.2 working on Ryzen: http://dpaste.com/0Y90SD8
04:29 PM mozmck: nice!
04:29 PM memfrob: The only thing this RTAI tree needs now is newlib plugged in but I'm surprised the test suite even works.
04:30 PM memfrob: No math support enabled yet.
04:30 PM mozmck: are you sure math support is required by linuxcnc?
04:30 PM memfrob: Yes.
04:31 PM memfrob: linuxcnc needs rtai_math.ko which provides all the math functions.
04:32 PM memfrob: Adding newlib into the RTAI source will be easy but it took me about 2 days to get the tree hacked together enough to compile on my end, need a break. This is a big step in progress though.
04:34 PM memfrob: The modifications I did to the tree are not production ready, a lot of hacks. RTAI 5.2 is a train wreck, much like the previous releases.
04:50 PM KimK: memfrob: Would it help RTAI (for others, and therefore indirectly for us) if you eventually published your RTAI improvements on GitHub? I mean if RTAI wasn't so hard to get started with, maybe more people would try it? And end up using it and liking it? I am told the original developer can be difficult, would that play into it?
04:52 PM KimK: Oops, got to go, I'll scroll back later, thanks!
04:53 PM memfrob: KimK, eventually, yes. It's not ready for Github rn, not even as a WIP branch.
04:56 PM memfrob: The most plug-and-play solution right now is to use 3.16.52 with this tree: https://github.com/NTULINUX/RTAI
04:56 PM memfrob: And when I get around to it, I'll fix up RTAI 5.2 the right way and push all my changes.
04:58 PM memfrob: 4.14.125 or whatever the latest 4.14 kernel is at the time, will have RTAI support, with the latest IPIPE kernel changes from Xenomai. Just a matter of time. Several conflicts to fix, etc.
04:59 PM memfrob: RTAI 5.1 never worked for me, no matter what I did, 5.2 actually has promise.
05:01 PM memfrob: When I fix RTAI 5.2 the right way, newlib will be included and it will be as plug-and-play as the master branch of that repo, and to the best of my knowledge, bug free.
05:02 PM memfrob: Some hardware isn't RTAI-friendly and there's nothing I can do about that, but for the systems it works on, it will always work on. No funny business, compiling errors, or any other build issues.
05:03 PM memfrob: The lead maintainer of RTAI writes the absolute worst Makefiles, half-working configure file, etc. they always need a full re-write.
06:19 PM Centurion_Dan: memfrob: I'm new to RTAI. How does it compare these days to rt_preempt in terms of jitter and latency?
06:29 PM memfrob: Centurion_Dan, well with PREEMPT_RT on this ryzen system, latency can hit as high as 200 microseconds. With RTAI, max usually never exceeds 20-25 microseconds in worst conditions.
06:30 PM memfrob: PREEMPT_RT on Intel boards can actually run better than RTAI can so it all depends on where you're starting from and with what hardware.
06:30 PM Centurion_Dan: That sure makes it worth while!!
06:32 PM Centurion_Dan: I hear that Ryzen is a pretty horrible arch compared to the previous generation with lots of bugs.
06:33 PM memfrob: No idea where you heard that, the only problem with Ryzen has been DDR4 speeds.
06:33 PM Centurion_Dan: I've got a machine here that is normally used as a digital audio workstation on a FX3850 - 8 core, and it's pretty reliable latency wise.
06:34 PM memfrob: 3.6 GHz is about the max Ryzen can get right now on pretty much all boards with no issues. With Ryzen 3000 coming out next month, DDR4 speeds will be able to reach 5 GHz.
06:34 PM Centurion_Dan: actually it's one I built for a colleague to run ardour, 64 channels of audio i/o.
06:35 PM Centurion_Dan: Is there a good independent latency test tool/
06:35 PM Centurion_Dan: ??
06:36 PM memfrob: OT: How good are you at ardour? Can you make songs like scary monsters and nice sprites?
06:36 PM memfrob: LinuxCNC's latency tool is what I use, RTAI has a built-in latency test though
06:37 PM Centurion_Dan: I'm pretty amature, but my friend is a semi-pro audio engineer, and he's mastered some pretty nice stuff.
06:39 PM Centurion_Dan: OT: I have the machine sitting here to use as a cinelerra render node, and because I had to fix some video and alsa routing issues.
06:40 PM memfrob: How can you have an alsa problem when there's .asoundrc ?
06:40 PM Centurion_Dan: memfrob: so whats required to build RTAI for linuxCNC - a kernel module and some userspace libs?
06:40 PM memfrob: RTAI and an RTAI-enabled kernel
06:41 PM memfrob: https://github.com/NTULINUX/RTAI/blob/master/README.INSTALL
06:41 PM Centurion_Dan: The asoundrc needs to be tweaked to utilise all the channels on the 2x hammerfall cards.
06:42 PM memfrob: I never heard of hammerfall, I use creative hardware, the support is absolute trash in Linux unless you're running the latest kernel git
06:42 PM Centurion_Dan: memfrob: is the sort of thing that might be worth packaging for Devuan, or is it in a pretty bad state still?
06:43 PM memfrob: My stable tree (master branch) is perfectly suitable for packaging.
06:43 PM Centurion_Dan: RME Hamerfall cards are pretty much the pro-audio go to these days and really well supported under linux.
06:43 PM memfrob: What audio driver does the kernel use? Oxygen HD or something?
06:45 PM Centurion_Dan: \me goes and looks...
06:46 PM memfrob: I was looking at an ASUS STRIX RAID 8 because it uses the C-Media driver which has great support and has enough channels for a Razer Tiamat 7.1 V2
06:47 PM Centurion_Dan: snd_hdspm is the module by the looks.
06:47 PM memfrob: Oh wow, a native driver!
06:48 PM memfrob: Thought it'd be intel hda generic
06:48 PM Centurion_Dan: yup. been in the mainline for a long long time.
06:48 PM Centurion_Dan: so it's mature...
06:49 PM memfrob: kernel support was just added for the AE-5 so I'm gunna pass on that.
06:49 PM memfrob: I don't know what it is about Creative chips that make them so difficult to get going in Linux.
06:49 PM memfrob: I have a sound blaster z and it rarely works in linux, the bus always gets in a weird state.
06:50 PM Centurion_Dan: you should be able to find RME hdsp pci-e cards for a reasonable price.
06:50 PM Centurion_Dan: and they are rock solid.
06:50 PM Centurion_Dan: gotta go get lunch ready... BBL.
06:51 PM memfrob: See ya!
10:20 PM memfrob: RTAI 5.2 -> Initial clean-up -> 1114 files changed, 4875 insertions(+), 211523 deletions(-)
10:23 PM memfrob: At least 30,000 more lines to delete.
10:25 PM memfrob: Scratch that, just took care of 262714 lines
10:26 PM memfrob: I'll work on this more later, gotta wind down, take care everyone!
10:26 PM Tom_L: good progress i see
10:30 PM memfrob: Thank you! :D
10:32 PM Tom_L: way above my paygrade but i appreciate it
11:27 PM memfrob: What I have so far: https://github.com/NTULINUX/RTAI/tree/for-next
11:27 PM memfrob: Night everyone!