#linuxcnc-devel Logs

Jun 27 2017

#linuxcnc-devel Calendar

12:07 AM hazzy: It's funny, I used to be frustrated that the wheezy live ISO was xfce and would always change it to something else when I got a chance.
12:08 AM hazzy: Over several installs the "when I got a chance" delay started getting longer and longer and now I use xfce exclusively, and like it very much indeed. I now use it even for non LCNC machines.
12:08 AM hazzy: I guess it is just what one gets used to that they like :)
12:08 AM hazzy: GN8
03:58 AM pcw_mesa_ is now known as pcw_mesa
07:14 AM KGB-linuxcnc: 03John Thornton 052.7 a43dd48 06linuxcnc 10docs/src/gui/touchy.txt Docs: add info about the radio buttons * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=a43dd48
07:38 AM skunkworks: there - sorry. got hexchat auto-connecting to the channels I like
07:49 AM TurBoss: edited: is remote EStop
11:27 AM jepler: I figure if you are a pro-enough user to have a strong opinion about DE, you're also up to the task of installing the debian desktop you want. If I make the image, I get to choose the default. My biases tell me that xfce is relatively low on resource usage and does a good 80% of what people need out of a DE. But also it's just the one I'm accustomed to and have used daily at least since switching
11:27 AM jepler: to debian wheezy.
11:27 AM jepler: It's not that much work to create 10 more images with all the major DEs represented (all those with task-xxx-desktop metapackages) but a lot of transfer time, a lot of disk space, and they might all have different weird problems for us to debug.
11:29 AM jepler: | (cinnamon, gnome, kde, lxde, lxqt, mate, xfce) × (i386, amd64) | = 14
11:31 AM jepler: so 12 more
11:38 AM jepler: anyway if you have a pet DE, it's just a one-line change in the script to switch to it. https://github.com/jepler/stretch-live-build/blob/master/do.sh#L21 change task-xfce-desktop to task-your-desktop
11:45 AM seb_kuzminsky: or two lines in the shell to switch an installed system from any DE to any other
11:45 AM seb_kuzminsky: jepler: thanks for working with micges to get the mesaflash project moving again
11:46 AM seb_kuzminsky: i agree that moving to 3.3 would be preferable to staying on 3.2
11:46 AM seb_kuzminsky: what do i/we need to do to get 3.3 to build packages?
11:47 AM jepler: seb_kuzminsky: I feel like mesaflash's master branch is in good shape.
11:47 AM jepler: there are no open issues except my whiny ones
11:47 AM jepler: I'd say to start with the tip of master, make a v3.3 branch and a v3.3.0-pre release tag
11:48 AM jepler: with that you should be able to make a dfsg-free tarball
11:48 AM seb_kuzminsky: i tried building packages from master but it failed to make the orig tarball - needs a dfsg tag?
11:48 AM seb_kuzminsky: oh
11:48 AM seb_kuzminsky: ok i'll try that
11:49 AM jepler: seb_kuzminsky: it's something to do with debian/rules get-orig-source
11:49 AM jepler: https://wiki.debian.org/onlyjob/get-orig-source
11:50 AM JT-Shop: isn't gnome pretty bloated for a DE?
11:51 AM seb_kuzminsky: JT-Shop: i run gnome on one of my wheezy linuxcnc machines, it works fine
11:51 AM seb_kuzminsky: bloat is in the eye of the beholder ;-)
11:51 AM seb_kuzminsky: i think pretty much any machine capable of running linuxcnc ok is also capable of running any DE ok
11:52 AM JT-Shop: ah, just what I've read from time to time
11:52 AM JT-Shop: I did try them all a while back...
11:53 AM jepler: some DEs require a compositing window manager, which requires accelerated OpenGL, which was historically a problem for realtime. I think gnome3 is in this group. But I also think realtime and accelerated OpenGL get along better than they used to.
11:54 AM seb_kuzminsky: maybe there is some VR DE with wobbly windows and reflections and echo and reverb that needs fancy graphics support
11:58 AM seb_kuzminsky: i just pushed v3.2.1 to mesaflash, just because it builds on all our platforms (wheezy, jessie, stretch, and precise)
11:59 AM seb_kuzminsky: i'll make a v3.3 branch, merge v3.2 into it, tag it, and try to build that
12:01 PM jepler: yeah but I"m pretty sure v3.2.1 is quite broken
12:01 PM jepler: the memcpy vs pci bug
12:02 PM jepler: didn't I mention this somewhere? sigh.
12:04 PM jepler: well .. making assumptions about what is in v3.2.1 anyway
12:05 PM seb_kuzminsky: ok, i won't push 3.2.1 to wlo's deb archive, i'll focus on 3.3 instead
12:10 PM jepler: 3.2.1 would be no more broken than 3.2, I guess...
12:18 PM seb_kuzminsky: uscan doesn't like that the git tag is v3.3.0-pre0 (as needed by git) and the dch version is 3.3.0~pre0 (as needed for proper sorting)
12:18 PM seb_kuzminsky: i wonder if i can rewrite "~pre" to "-pre" in the watch file
12:18 PM jepler: argh
12:24 PM seb_kuzminsky: uscan's uversionmangle can probably do it
12:24 PM seb_kuzminsky: watch files are one of those things only old unix hackers can love
12:27 PM jepler: yeah well I'd rather just delete those files from the git, but I hear there are windows users of mesaflash sooo
12:30 PM seb_kuzminsky: i think i got it working
12:31 PM pcw_mesa: Ha, I use a windows version of mesaflash for first time card init,
12:31 PM pcw_mesa: but just because its so old it doesn't care about the card name (which is all 0xFFs before the name is written)
12:47 PM seb_kuzminsky: pcw_mesa: if you were the only person using mesaflash on windows i'd suggest that we make it work for you on linux, but i guess there might be other windows users
12:48 PM pcw_mesa: I think the latest Mesaflash will work with un-named cards but I have not tried it
12:54 PM pcw_mesa: I would also have to port our batch files/ low level utilities to do initial card init on unix: setname,setmac etc (I think I ported setname before)
12:54 PM pcw_mesa: but Im happy with a 5 year old windows version of mesaflash because not much is asked of it other that writing the flash chip
01:00 PM seb_kuzminsky: i guess i need to remove the non-distributable files from the git working tree when building the dsc? during clean maybe?
01:46 PM jepler: seb_kuzminsky: somehow dpkg was supposed to do it magically?
01:47 PM jepler: seb_kuzminsky: I can look at it tonight if you're struggling. At one point I grokked how it was supposed to work after all
01:52 PM cradek_ is now known as cradek
01:52 PM -!- #linuxcnc-devel mode set to +v by ChanServ
01:54 PM seb_kuzminsky: i've muddled through it
01:54 PM seb_kuzminsky: i'll make a PR, i'd appreciate if you looked at it and told me should have done instead
01:57 PM jepler: $ VER=3.2.1 debian/rules get-orig-source
01:57 PM jepler: $ dch -v 3.2.1-1
01:57 PM jepler: I then had to change debian/source/format to 3.0 (quilt)
01:58 PM jepler: hm except that's not using the dfsg tarball
01:59 PM jepler: so dch -v 3.2.1+dfsg-1 and also remove the files from the source tree manually
01:59 PM jepler: it feels like a mess
02:46 PM Tom_itx: moving from rtai to preempt-rt are there config changes?
02:46 PM Tom_itx: gonna try loading stretch on a ssd
02:49 PM pcw_mesa: should be no config differences at all
02:50 PM Tom_itx: great
03:00 PM seb_kuzminsky: jepler: does this look right? https://github.com/SebKuzminsky/mesaflash/commits/v3.3
03:00 PM seb_kuzminsky: mostly the final commit that fixes(?) packaging in 3.3
03:00 PM seb_kuzminsky: back in a bit
03:02 PM jepler: seb_kuzminsky: I think 'debian/rules clean' needs to invoke the real clean rule, or maybe just invoke $(MAKE) clean
03:27 PM seb_kuzminsky: oh, right
03:27 PM seb_kuzminsky: i wonder if it's time to switch all the things to gbp
03:49 PM jepler: good buckwheat pancakes?
03:49 PM seb_kuzminsky: http://honk.sigxcpu.org/projects/git-buildpackage/manual-html/gbp.intro.html
03:58 PM Tom_itx: D525 Atom, only burp so far installing stretch was I had to configure the network manually
04:00 PM pcw_mesa: mmm pancakes
04:01 PM seb_kuzminsky: mmm git :-)
04:13 PM jepler: hmmm people who pronounce "git" as "git" but "gif" as "jif" seem to be hypocrites
04:13 PM jepler: I guess I'll just have to say "jit" from now on.
04:13 PM cradek: I am not sure that follows
04:16 PM Tom_itx: max servo thread 856254 base: 95961
04:16 PM jepler: that's not great
04:17 PM jepler: I wonder what to do about localization in the strech iso I'm working on
04:17 PM Tom_itx: what should i look for in the bios? I forget...
04:17 PM jepler: with full internationalization, it's a 2GB .iso. With just the languages that we have a .po file for in linuxcnc, it's 1.6GB, and with only US English it's 1.1GB.
04:21 PM cradek: can you pick a different language when installing?
04:23 PM jepler: I don't think that's a standard prompt
04:23 PM jepler: you can install it via the package manager afterwords
04:23 PM jepler: e.g., task-french-desktop
04:24 PM Tom_itx: i was promted for a language
04:24 PM jepler: OK, maybe I'm mistaken
04:25 PM Tom_itx: i plan to run it again in a bit
04:25 PM jepler: that may change the text in the installer, or it might also pull in the related task-<language>-desktop
04:25 PM Tom_itx: turned off hyperthreading in the bios and no latency test
04:25 PM Tom_itx: stays at 0
04:26 PM Tom_itx: (not running from the livecd) rather an install
04:26 PM jepler: that's not too useful then
04:26 PM Tom_itx: heh
04:26 PM Tom_itx: i'll try from live
04:30 PM jepler: when this condition occurs, I'd be interested in the output of:
04:30 PM jepler: $ getconf -a | grep NPROC
04:30 PM jepler: $ taskset -p $$
04:30 PM jepler: (or taskset -p -c $$)
04:31 PM jepler: the first will tell me how many processors there are "configured" and "online"; the second will tell me which CPU numbers a "normal task" is trying to run on
04:31 PM jepler: LinuxCNC looks at _NPROCESSORS_ONLN and tries to run all realtime processes on the processor #(NPROCESSORS_ONLN-1) which generally works
04:31 PM jepler: .. but I've never disabled HT trying to get better latency so maybe that's not the right choice
04:33 PM Tom_itx: same results with live latency test
04:33 PM andypugh: Did anyone here see this on the forum?
04:33 PM andypugh: Recent commit of debian/configure script looks to be wrong
04:33 PM andypugh: Line 159
04:33 PM andypugh: +Debian-9.*|Raspbian-8.*) # Jessie
04:33 PM andypugh: I think it should be
04:33 PM andypugh: +Debian-8.*|Raspbian-8.*) # Jessie
04:33 PM andypugh: ( https://forum.linuxcnc.org/9-installing-linuxcnc/32955-debian-configure-commit#95014 )
04:34 PM Tom_itx: getconf returns 2
04:34 PM jepler: andypugh: argh did I get it wrong yet again?
04:34 PM Tom_itx: for both
04:34 PM jepler: sigh
04:35 PM Tom_itx: taskset returns pid 1082's current affinity mask: 5
04:35 PM andypugh: Same chap has a query about whether libxenomai-dev is needed for RTAI or Preempt-rt
04:37 PM jepler: Tom_itx: OK, that's unusual. mask 5 is the same as list 0,2. But since _NPROCESSORS_ONLN is 2, it will try to set the affinity list to 1, or mask 2. This means that the realtime task can no longer run on any online CPU, so it'll never run.
04:37 PM jepler: andypugh: no, of course not
04:38 PM Tom_itx: so disabling hyperthread for better latency is a no no on stretch
04:38 PM jepler: in master branch, uspace can support 3 (4) types of realtime: preempt-rt, xenomai, and rtai (and non-realtime).
04:38 PM jepler: debian/configure tries to enable any of them that look like they might work
04:38 PM jepler: but for any specific machine you probably want just one
04:38 PM jepler: andypugh: ^^^
04:38 PM andypugh: He opines that: “The debian/configure scripts checks presence of libxenomai-dev installation and ask to enable deb-src repositories for compiling."
04:38 PM jepler: yes that's correct
04:39 PM andypugh: OK, I will reply that this is expected
04:40 PM jepler: if there is a libxenomai-dev binary package known to apt-cache search, THEN debian/configure has to find its associated source package
04:40 PM jepler: .. in order to know the correct list of Debian architectures where libxenomai-dev will be available
04:41 PM jepler: if your /etc/apt/source.list configuration has libxenomai-dev as a BINARY PACKAGE but cannot find its ASSOCIATED SOURCE PACKAGE then your debian is misconfigured
04:41 PM jepler: hoever, this seems to happen in practice in the jessie -> stretch update, because you might have leftover information about the jessie version of libxenomai-dev, but no information about the source package because libxenomai-dev is NOT IN debian stretch
04:42 PM jepler: or you might have it because your distro is just misconfigured to begin with, by not having deb-src lines
04:43 PM Tom_itx: with hyperthreading, getconf returns 4 for both
04:43 PM Tom_itx: taskset returns pid 1082's current affinity mask: f
04:46 PM KGB-linuxcnc: 03Jeff Epler 05master f9aab02 06linuxcnc 10debian/configure debian/configure put back Debian 8 * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=f9aab02
04:48 PM Tom_itx: wow, base: 90971 servo: 1948192
04:51 PM jepler: Tom_itx: I filed a github bug, if you have an account you may want to subscribe to it
04:51 PM Tom_itx: i don't
04:52 PM Tom_itx: err maybe i do.. i never use it
04:52 PM Tom_itx: trying on an asrock board now
04:52 PM jepler: no big deal, just the opportunity to get notified when there is activity related to the bug
04:52 PM jepler: afk
04:54 PM Tom_itx: asrock returns servo 18250 base 44275 so far
04:54 PM Tom_itx: alot better than the D525
04:55 PM Tom_itx: Q1900M Pro3
05:11 PM jepler: can't reproduce locally, with HT off my system has _NPROCESSORS_ONLN 4, affinity list 0-3, so the (N-1) algorithm works fine
05:12 PM Tom_itx: i'm not exactly sure where HT is in the asrock but i changed a couple things in the bios and latency always runs
05:13 PM Tom_itx: i tried the intel D525 because it was a popular board there for a while
05:14 PM pcw_mesa: in general faster CPUS are better for Preempt-RT
05:14 PM Tom_itx: it may not be a good candidate for an update
05:17 PM pcw_mesa: d525s are pretty slow, best Ive seen with Preempt-RT is maybe 70 usec on the servo thread (no base thread)
05:17 PM pcw_mesa: also D525s (and older hardware) may do better with 3.xx kernels
05:19 PM pcw_mesa: (the many Core2 Duo junk machines I have, have much better latency on 4.1 or 3.18 vs 4.9 and some will crash every few days with 4.xx kernels but are totally stable with 3.18)
05:20 PM pcw_mesa: the crashing is a unfixed video issue with older integrated intel video and 4.xx kernels
05:23 PM Tom_itx: asrock staying around base: 141032 servo: 20189
05:33 PM jepler: I can reproduce something like Tom_itx's no-HT problem by: $ sudo sh -c 'echo 0 > /sys/devices/system/cpu/cpu6/online'
05:34 PM jepler: since I have 8 threads, this makes _NPROCESSORS_ONLN=7, and (N-1) is the number of a CPU that is not online
05:34 PM jepler: latency-test stays stuck at 0
06:00 PM jepler: o.O wow this is some good latency (not stuck at 0)
06:00 PM jepler: 4.4us latency servo, 7.7us latency base
06:00 PM jepler: .. rtapi running on CPU 7, CPU 6 (other half of HT pair) offline
06:00 PM jepler: I even did 30 seconds of glxgears and firefox
06:08 PM jepler: hm, no, that must not be it; the HT sibling of 7 is 3
07:13 PM jepler: huh I wonder how to have travis-ci *NOT* pre-configure a bunch of non-Ubuntu package servers
07:13 PM jepler: W: Failed to fetch http://ppa.launchpad.net/chris-lea/redis-server/ubuntu/dists/precise/InRelease Could not connect to ppa.launchpad.net:80 (, connection timed out
07:14 PM jepler: noticed because it happens to be failing right now
07:14 PM jepler: also this seems weird W: http://security.ubuntu.com/ubuntu/dists/precise-security/InRelease: Signature by key 630239CC130E1A7FD81A27B140976EAF437D05B5 uses weak digest algorithm (SHA1)
07:46 PM jepler: next time's the charm
08:25 PM jepler: seb_kuzminsky: I found that I needed this stanza in debian/rules:
08:25 PM jepler: .PHONY: override_dh_auto_clean
08:25 PM jepler: override_dh_auto_clean:
08:25 PM jepler: rm -rf winio32 WinIo32.* libpci.dll libpci
08:25 PM jepler: dh_auto_clean
08:25 PM jepler: otherwise, "debuild"ing twice in a row would give errors about files left over
08:26 PM jepler: seb_kuzminsky: I'm not sure about where the 'watch' file should point, tbh. clearly you need to point it at yourself during testing..
08:27 PM jepler: (and many debian packages fail to "debuild" twice in a row due to left over files so it may be NBD. you don't notice it if using sbuild or any of those alternatives)
08:29 PM KGB-linuxcnc: 03Jeff Epler 05master 5d6966a 06linuxcnc Merge branch 'remove_trivkinstype' of https://github.com/phillc54/linuxcnc * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=5d6966a
08:29 PM KGB-linuxcnc: 03Jeff Epler 05master d182cfa 06linuxcnc 10src/emc/usr_intf/axis/scripts/axis.py Merge branch 'issue_284' of https://github.com/phillc54/linuxcnc * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=d182cfa
08:29 PM KGB-linuxcnc: 03Jeff Epler 05master b43d799 06linuxcnc Merge branch 'python-interface' of https://github.com/phillc54/linuxcnc * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=b43d799
08:47 PM skunkworks: wow - I must be blind - where are the screen saver settings in xfce?
08:53 PM jepler: skunkworks: you aren't blind, there aren't really screen saver settings as such
08:53 PM skunkworks: hmm can I blank the screen? I do like the password screen on my personal system. :)
08:54 PM skunkworks: I will google
08:54 PM jepler: as far as power saving, applications > settings > power manager
08:55 PM jepler: as far as whether it locks, the locker in *new* installs of debian stretch is "light-locker". you can enable/disable it via applications > settings > session and startup > application autostart > checkbox next to "screen locker"
08:55 PM jepler: at least .. that's what I think I know
08:55 PM jepler: good luck googling
10:01 PM seb_kuzminsky: imma push the mesaflash 3.3.0~pre0+dfsg-1 (yes really) packages to wlo unless someone points out the error of my ways in the next few minutes