#linuxcnc-devel Logs

Nov 16 2021

#linuxcnc-devel Calendar

10:26 AM seb_kuzminsky: the jessie build failure above is because the buildslave VM ran out of disk
10:53 AM seb_kuzminsky: i freed up some space, it should run now
10:54 AM seb_kuzminsky: the wheezy build failure seems like a transient problem due to my vm host being too oversubscribed and the guest getting starved :-(
10:54 AM skunkworks: seb_kuzminsky: Thank you!
11:05 AM linuxcnc-build: build #3335 of 1630.rip-stretch-rtpreempt-amd64 is complete: Failure [4failed compile runtests] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1630.rip-stretch-rtpreempt-amd64/builds/3335 blamelist: Chris Morley <chrisinnanaimo@hotmail.com>, Sebastian Kuzminsky <seb@highlab.com>
11:10 AM seb_kuzminsky: boo
11:11 AM skunkworks: lol - fighting the good fight...
11:11 AM seb_kuzminsky: more timeouts... 2.8 runs on sooo many platforms, my poor vm host gets totally overloaded
11:12 AM seb_kuzminsky: i'm going to try shifting some stuff around
11:13 AM skunkworks: What kinda hardware is it running on? I don't know if I have anything newer - they would be older than dell r720's
11:18 AM linuxcnc-build: build #8346 of 0000.checkin is complete: Failure [4failed] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/8346 blamelist: Chris Morley <chrisinnanaimo@hotmail.com>, Sebastian Kuzminsky <seb@highlab.com>
11:18 AM skunkworks: seb_kuzminsky: we have some dell 2970 vintage servers we will never used...
11:23 AM skunkworks: (the r720's are running proxmox and are still decent)
11:27 AM seb_kuzminsky: thanks for the offer... i'll think about it. I just reduced the cpu count in a bunch of the 2.8-only VMs, it should get us moving again
11:28 AM seb_kuzminsky: linuxcnc-build: force build --branch=2.8 0000.checkin
11:28 AM linuxcnc-build: The build has been queued, I'll give a shout when it starts
11:28 AM seb_kuzminsky: back in a bit
11:32 AM skunkworks: I can make up a list of what we have. You can have whatever you want. I was about ready to send them to recycling.
12:18 PM linuxcnc-build: build forced [ETA 1h07m08s]
12:18 PM linuxcnc-build: I'll give a shout when the build finishes
02:06 PM linuxcnc-build: Hey! build 0000.checkin #8348 is complete: Success [3build successful]
02:06 PM linuxcnc-build: Build details are at http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/8348
02:15 PM Tom_L: hey! i see green!
02:16 PM CaptHindsight[m]: seb_kuzminsky: i talked to Alec about a Gentoo buildbot
02:16 PM CaptHindsight[m]: since it is mainly for RTAI we need to run on real hardware vs a VM
02:17 PM CaptHindsight[m]: so he will keep using his own hardware and keep posting new RTAI for LCNC on his public GITHUB
02:18 PM CaptHindsight[m]: not sure about preempt_rt Gentoo work right now
02:21 PM CaptHindsight[m]: someone could have a buildbot for Gentoo testing just to check builds vs real time performance
02:22 PM CaptHindsight[m]: but RTAI and Xenomai testing is about performance on actual hardware vs just "did it compile?"
02:27 PM seb_kuzminsky: we run rtai in VMs on the buildbot all the time
02:27 PM seb_kuzminsky: in precise and wheezy
02:27 PM CaptHindsight[m]: just to see if it builds
02:28 PM CaptHindsight[m]: vs performance testing
02:28 PM seb_kuzminsky: it doesn't get great realtime performance, of course, but it exercises all the code paths
02:28 PM seb_kuzminsky: it runs the full test suite, so it's more than just "did it build?"
02:29 PM seb_kuzminsky: realtime performance is highly hardware-dependent (or, i guess, platform dependent, to include VMs), and it's not what we're trying to assess with those VMs - there we're looking for correctness testing
02:31 PM CaptHindsight[m]: yeah for best performance you really want to tweak RTAI for each version of CPU
02:33 PM CaptHindsight[m]: goes against a general RTAI kernel for all
02:33 PM seb_kuzminsky: yeah :-/
02:34 PM CaptHindsight[m]: we'll figure something out
02:35 PM seb_kuzminsky: we're getting pulled between two different goals here: 1: supply something easy to use that works well enough for as many people as possible, and 2: make it so capable motivated folks can tweak the last few drops of performance out of their hardware
02:35 PM CaptHindsight[m]: he develops Linux now for massive and fastest local cloud services
02:35 PM seb_kuzminsky: cool!
02:36 PM CaptHindsight[m]: real crazy copy a bank of server stuff in seconds all with mouse clicks
02:36 PM CaptHindsight[m]: can't say much more
02:37 PM CaptHindsight[m]: using the fastest networks
05:35 PM skunkworks: zfs is pretty amazing that way for replication...
05:35 PM skunkworks: for us plebs..