#linuxcnc-devel Logs

Jan 09 2018

#linuxcnc-devel Calendar

07:24 AM jepler: linuxcnc-build: force build --branch=jepler/master/stdint 0000.checkin
07:26 AM linuxcnc-build: build forced [ETA 2h13m03s]
07:26 AM linuxcnc-build: I'll give a shout when the build finishes
07:32 AM jepler: reboot upcoming on forum
07:33 AM jepler: whaaaatttt there's no meltdown patched kernel for ubuntu lts yet?
07:34 AM jepler: nor for this g-d spate of new php7 bugs https://www.debian.org/security/2018/dsa-4080
07:35 AM linuxcnc-build: build #4503 of 1301.rip-precise-rtai-i386 is complete: Failure [4failed compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1301.rip-precise-rtai-i386/builds/4503
07:39 AM jepler: /usr/lib/gcc/i686-linux-gnu/4.6/include/stdint-gcc.h:86:26: error: conflicting types for ‘uintptr_t’
07:39 AM jepler: /usr/src/linux-headers-3.4-9-common-rtai/include/linux/types.h:45:24: note: previous declaration of ‘uintptr_t’ was here
07:39 AM jepler: OK afk
07:45 AM linuxcnc-build: build #3118 of 1401.rip-wheezy-rtai-i386 is complete: Failure [4failed compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1401.rip-wheezy-rtai-i386/builds/3118
07:45 AM linuxcnc-build: build #2442 of 1902.clang-wheezy-rtai-i386 is complete: Failure [4failed compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1902.clang-wheezy-rtai-i386/builds/2442
08:21 AM linuxcnc-build: build #5310 of 0000.checkin is complete: Failure [4failed] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/5310
12:01 PM pcw_home: Is there some standard way to round rather than truncate integer divisions? (re issue 394)
12:04 PM seb_kuzminsky: i think a bug like that affected ilowpass a while back
12:04 PM pcw_home: easy to do
12:05 PM pcw_home: I guess adding 1/2 the divisor to the dividend would work
12:06 PM seb_kuzminsky: my memory was off by a bit: https://github.com/LinuxCNC/linuxcnc/commit/05d639606e18237fbcd758cd1b815d46b2b537b2
12:07 PM seb_kuzminsky: yeah that fix sounds right
12:19 PM KimK: I think you meant add 1/2 the divisor to the quotient, but we knew what you meant. (Do as I mean, not as I say?)
12:22 PM seb_kuzminsky: right
12:26 PM skunkworks_: hmm - i think I need a 7i76
12:26 PM pcw_home: I think I was right, that is if I add the divisor to the dividend that adds 1 to the quotient
12:27 PM pcw_home: (and I want to add 0.5 to the quotient)
12:27 PM seb_kuzminsky: oh yeah, dividend / divisor = quotient
12:34 PM KimK: Is that how it was done in the github link cited earlier? Maybe best to keep them both the same method?
12:35 PM KimK: Even though it does sound like either method should work.
12:44 PM pcw_mesa: No, the github link uses floating point (trying to avoid that here)
01:13 PM linuxcnc-build: build #1463 of 4022.deb-jessie-amd64 is complete: Failure [4failed shell_1] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/4022.deb-jessie-amd64/builds/1463 blamelist: Jeff Epler <jepler@unpythonic.net>
01:14 PM seb_kuzminsky: that failure looks like a glitch in the buildbot network - it's intermittent and i've been chasing it for a while now :-/
01:25 PM speedy: hi! i need help finishing a userspace rt component for the jog dial XHC WHB04B-6. in detail i need help connecting the last jog-dial related signals do halui.jog.* or axis.N.jog-* or similar.
01:25 PM speedy: hi! i need help finishing a userspace rt component for the jog dial XHC WHB04B-6. in detail i need help connecting the last jog-dial related signals to halui.jog.* or axis.N.jog-* or similar.
01:33 PM KimK: speedy: Hi, what help do you need?
01:35 PM speedy: hi, tnx for your reply! i already started the conversation at gitter in machinekit channe, but got no reply. thus i try to copy the same text here in. ok?
01:37 PM KimK: OK
01:39 PM speedy: I started programming a component for the jog dial XHC WHB04B-6 [1] (an advanced version of WHB04). It is almost done, but the jog dial could work more intuitive. Thus I need help for connecting the jog-dial's signals correctly: counts, increment, decrement, step, ... So far the signals have been connected to axis.N.jog-*. This works fine in step-mode jogging but when it comes to MPG mode or CONTINUOUS mode (velocity mode) I f
01:39 PM speedy: I think the main problem is that - I am sure - my understanding of MPG, STEP, CONTINUOUS (or velocity) modes is limited or wrong.
01:39 PM speedy: In short: How can the maximum velocity of all axis be set (halui.max-velocity.value float out)?
01:40 PM speedy: What I did so far: The jog dial was connected to halui.jog-speed and halui.jog.* but the result was a jerking axis move, thus the signals have been moved to axis.N.jog-*. Besides of that, the jog dial delivers jog-counts, which cannot be directly connected to halui.jog (there is no halui.jog.counts).
01:40 PM speedy: What it is now: With the jog counts connected to axis.N.jog-counts this works fine in STEP mode but in CONTINUOUS mode the WHB04B-6 shows a percentage value for the speed (2%, 5%, 10%, 30%, 60% 100%). Thus I thought the axis.N.jog-vel-mode should be TRUE, have a constant and wide step scale of i.e. axis.N.jog-scale=2 and the machine's max-velocity should be set to the percentage value (i.e. if 60% then halui.max-velocity.value
01:40 PM jepler: seb_kuzminsky: OK, do you think that branch is OK to merge to master then?
01:40 PM jepler: I'm in no hurry so I could retrigger buildbot too
01:40 PM speedy: crux, where can a maximum velocity halui.max-velocity.value be set?
01:44 PM KimK: speedy: Have you seen this?: http://www.linuxcnc.org/docs/2.7/html/gui/halui.html Would one of the halui.feed-override.????? values work for you?
01:45 PM jepler: reading the source code, it looks like changing halui.max-velocity.counts causes halui to send a message to change the max velocity, which will later be reflected in halui.max-velocity.value
01:45 PM jepler: the halui.max-velocity.direct-value pin controls whether the setting is absolute or relative
01:45 PM jepler: but you mentioned machinekit, and I am referring to the linuxcnc master branch so their functionality could be different.
01:52 PM speedy: linuxcnc master branch: i see the point. dind't think of that both are so much different. and now i see that i was sloppy when reading the documentation ('halui.feed-override.direct-value' (bit, in) - false when using encoder to change counts, true when setting counts directly. The 'count-enable' pin must be true.) setting counts directly may help out (will test it coming weekend).
01:55 PM speedy: once the component is ready (documentation, reviewed, etc...), i could help merge it to linuxcnc - if there is interest of-course.
01:56 PM jepler: we take pull requests on github, but we don't always have qualified reviewers so they don't always get merged as quickly as we'd like
01:57 PM speedy: good. maybe i should bother machinekit for review, then come back to linuxcnc :)
01:59 PM speedy: regarding manual pulse generator mode (MPG), continuous mode and step mode, there i'd need help too (for understanding them correctly). step is clear: the machine will move the amount of cummulated jog counts. it won't stop when jog dial has stopped (for example the the jog dial has been moved faster than the machine can do).
02:00 PM jepler: I'm not an expert there, sorry I can't help.
02:01 PM speedy: the chinese device manual describes the continuous mode as the velocity mode. the machine will move as long the jog dial is rotated and stop asap when the jog dial stops. my only question which step scale should be taken for that.
02:06 PM seb_kuzminsky: jepler: yeah, looks good to me
02:07 PM KimK: speedy: Yes, whether to choose distance mode or velocity mode can be an issue, as you describe. I prefer distance mode (MPG scale works) plus operator education (don't crank far ahead of machine). Velocity mode does stop moving when operator stops cranking, but there aren't many (commercially made) machines out there that work that way.
02:07 PM seb_kuzminsky: not that i'm the gatekeeper for master
02:13 PM speedy: i see. i am asking because, since the development of the component took me much time, i'd like to make it feature complete and then leave it as is (except of bug reports of-course). but the chinese type of english documentation drives me nuts. it describes the speed mode, velocity mode and then there is an MPG mode - which i don't get what it is.
02:30 PM seb_kuzminsky: jepler: thanks for cleaning up my 10 year old mess
03:14 PM KimK: speedy: Is there a link to the Chinglish documentation?
03:23 PM speedy: http://www.cdxhctech.com/upload/201703/20170317170802482.pdf (http://www.cdxhctech.com/html/eProduct/view_363.html)
03:24 PM speedy: these are the screens i could reproduce when reverse-engineering the protocol: https://github.com/rubienr/machinekit/tree/feature-xhc-whb04b-6/src/hal/user_comps/xhc-whb04b-6#display-examples
03:24 PM speedy: manual: see page 7
03:43 PM speedy: did somebody compile linuxcnc on beagle bone black? i had troubles (at least with machinekit). at some point the compiler run out of memory (i assume it will be the same procedure with linuxcnc). the workaround was: configure, compile as far possible. then power off, connect bbb via usb, chroot with arm interpreter enabled and compile the rest. then, again, further compiling on the running bbb. but this procedure sounds odd.
04:42 PM jepler: we recently took a patch to reduce peak memory usage while compiling, but it may still not fit in a bbb without swap. nobody in this project builds for bbb as far as I know. Our one ARM system in the buildbot is an armhf but it has 2GB RAM or something so it muddles through.
04:49 PM seb_kuzminsky: speedy: i built linuxcnc for bbb a while back
04:49 PM speedy: i wonder how the debian packages are built. as far i see, the configure step relys on kernel extensions and build environment which are set on runtime.
04:49 PM speedy: seb_kuzminsky: how did you overcome the mem issue?
04:49 PM jepler: our ARM package is for preempt-rt, which does not have a dependency on a specific kernel
04:50 PM seb_kuzminsky: i added a swap partition on the sd card
04:50 PM speedy: hmm, but configure checks a lot of things.
04:51 PM seb_kuzminsky: speedy: http://linuxcnc.org/docs/devel/html/code/building-linuxcnc.html#_building_debian_packages
04:51 PM seb_kuzminsky: much of what debian/configure looks for how to generate the debian package metadata, build depends etc
04:51 PM jepler: afk
04:52 PM seb_kuzminsky: seeya jeff
04:52 PM speedy: seb_kuzminsky: damn, i did not have this fancy idea :)
04:53 PM speedy: seb_kuzminsky: 3.2. Building Debian Packages won't cross compile and pack, right?
04:54 PM speedy: seb_kuzminsky: what was the swap size?
04:55 PM seb_kuzminsky: probably 2 gig, i dont remember
04:55 PM seb_kuzminsky: i know that linuxcnc builds in 2 gigs of ram
04:55 PM speedy: i see. will give it a try next time.
04:57 PM seb_kuzminsky: cool
04:57 PM seb_kuzminsky: i bet it will work, but slllloooowwww
04:57 PM seb_kuzminsky: somewhere i have my notes on setting up a linuxcnc dev environment on the bbb, i should find it and publish it
04:59 PM speedy: i don't mind as long there is a way to easily compile in the bbb. i always put my notes in a github repo or gist.
05:03 PM speedy: i must say, i am impressed by linuxcnc on that weak bbb thingi. even if it is rather slow when it comes to visualizing the 3d-path but it works.
08:03 PM jepler: OK, meltdown mitigation kernel is now available on ubuntu lts. rebooting flo in a second..
08:03 PM jepler: downloading from http://archive.ubuntu.com/ubuntu is slow, wonder if they're experiencing a larger than normal number of updates
08:03 PM jepler: Fetched 69.5 MB in 44s (1,556 kB/s)
08:03 PM jepler: but also how crazy is it that 1.5MB/s now feels slow
08:06 PM jepler: also rebooting the machine where my irssi lives... bbiab
08:06 PM skunkworks_: 1MB/s is too slow for youtube now..
08:07 PM jepler: welp
08:14 PM jepler: [ 0.000000] Kernel/User page tables isolation: enabled
08:37 PM rob is now known as Guest41912