#linuxcnc-devel Logs

Sep 22 2017

#linuxcnc-devel Calendar

06:09 AM jthornton: is there a guide to updating using the github? all my notes are for the old git and I see a doc error
06:11 AM archivist: been quite a few that were hunting for the old because docs are lagging :)
07:07 AM DrippityDrops1 is now known as DrippityDrops
08:11 AM jepler: jthornton: Only the original mailing list announcement, here: https://www.mail-archive.com/emc-developers@lists.sourceforge.net/msg17619.html
08:12 AM jepler: you can also get instructions for creating a new clone on the github page https://github.com/LinuxCNC/linuxcnc/ when you click "clone or download", but that will start you with a fresh clone
08:14 AM jthornton: jepler: thanks
10:47 AM seb_kuzminsky: jepler: i've been running your stretch-uspace 2.7 install on my bridgeport for a while now, it's running well and making parts, thanks!
10:47 AM seb_kuzminsky: i think the focus behavior in axis is different than it is on the wheezy-rtai 2.7 install on the same machine, but i haven't tracked it down
10:48 AM seb_kuzminsky: sometimes keyboard control of things like jog speed or machine max speed doesn't work when my fingers expected it to
10:49 AM archivist: I dont think that is a recent problem
10:49 AM archivist: seen it on memory starved earlier versions too
10:49 AM seb_kuzminsky: do you think it's a problem that comes with debian stretch, or with something else?
10:50 AM archivist: I think starved of resources the kb events get delayed
10:51 AM seb_kuzminsky: in a low memory situation the kernel can start paging out any non-mlocked process (anonymous pages go to swap, file-backed pages get discarded and re-fetched from disk later), that can sure cause processing delays
10:51 AM seb_kuzminsky: the only fix is "use less memory or add more memory"
10:51 AM archivist: not sure how to debug but seen others with similar problems over the rears
10:51 AM archivist: years
10:52 AM archivist: yup I have always assumed swapping as the cause
10:52 AM archivist: should we have a way of spotting that?
10:53 AM seb_kuzminsky: that'd probably be helpful
10:53 AM seb_kuzminsky: you can see paging activity with "vmstat 1", but that's not very helpful for most of our users
10:54 AM seb_kuzminsky: possibly we should mlock all the code we care about, to prevent that from happening
10:54 AM archivist: I see it a lot with my machine upstairs when the text editor is open as well
10:54 AM seb_kuzminsky: i'm thinking task and the guis, and maybe the "userspace" aka "non-realtime" hal components
10:55 AM archivist: worth a try
10:55 AM seb_kuzminsky: archivist: next time it happens, run "vmstat 1" in a terminal and see if the "io bi" and "io bo" columns are going nuts
10:55 AM seb_kuzminsky: if there's a lot of block-in and block-out activity, that would suggest your machine is paging
10:56 AM seb_kuzminsky: i don't think that's what i'm seeing on my machine
10:56 AM archivist: wanna bet :)
10:56 AM seb_kuzminsky: i think somehow axis just loses focus and the keypress events are going someplace else and not getting to the widgets
10:57 AM seb_kuzminsky: the machine is responsive and fast, and mouse events on the widgets work immediately, it's just keypresses that do not get handled
10:57 AM archivist: putting up other text boxes was a good way of losing focus like the print/debug
10:57 AM seb_kuzminsky: if i'm in the mdi tab, keypresses show up immediately in the mdi text entry field
10:59 AM seb_kuzminsky: it'd be interesting to have our non-realtime code monitor their performance behavior with getrusage()
11:00 AM seb_kuzminsky: if they're taking a lot of "major" (io-causing) page faults after startup, that'd be a good indicator of not enough memory and poor performance
11:03 AM jepler: seb_kuzminsky: if you find something reproducible, please open an issue. but if it's really a wheezy->stretch change, it's likely to not be our bug
11:15 AM archivist: which reminds me of http://www.overclockers.com.au/image.php?pic=images/newspics/22sep17/19.jpg
11:17 AM seb_kuzminsky: jepler: i'll try to reproduce it
11:18 AM seb_kuzminsky: i'm a bad citizen of our open source community, because i was more interested in making my part than in reproducing the bug i saw :-/
11:18 AM seb_kuzminsky: anything that goes through rtapi_app_main() runs mlocked(), but all the non-realtime stuff, including important things like task and the guis, do not
11:20 AM seb_kuzminsky: has anyone built a touch probe? i'm thinking of building something like this: http://web.archive.org/web/20121126174517/http://www.indoor.flyer.co.uk:80/probe.htm
11:20 AM seb_kuzminsky: uh oh, archive.org is down, we need a meta-archive.org
11:21 AM seb_kuzminsky: fine, something like this, then: http://www.homemetalshopclub.org/projects/touch_probe/touch_probe.html
11:22 AM archivist: seb_kuzminsky, the real ones press the balls into the substraight
11:23 AM archivist: I have been playing with real but old renishaw and other probles, the hard part can be the reading of the contacts
11:25 AM archivist: made a fartuino reader that looks for change in resistance on the probe touching, renishaw does something similar
11:26 AM seb_kuzminsky: it looks for electrical contact between the probe stylus and the surface being probed? i thought renishaw used physical displacement of the probe stylus to open an NC-MO circuit
11:26 AM archivist: as the balls/cylinders wear the contact resistance rises, takes that into account
11:27 AM seb_kuzminsky: huh, i'd expect it to be overpowered by the pull-up resistor
11:27 AM archivist: they look for resistance change that happens befor it opens
11:27 AM seb_kuzminsky: gotcha
11:27 AM archivist: more sensitive
11:29 AM archivist: I also have a force sensor in the code to measure the contact force http://www.archivist.info/cnc/Arduino_Probe_Interface/
11:29 AM archivist: one day I might polish it up and finish
11:30 AM pcw_home: I think I'd make one with ceramic balls and mount accelerometers on the stylus and housing and use the difference as the trigger
11:32 AM archivist: the modern use strain gauges
11:37 AM pcw_home: pretty old tech
11:37 AM pcw_home: cheap accelerometers are relatively new
11:38 AM archivist: not seen that in a probe yet
11:38 AM jepler: yeah if you pursue the idea I'd love to hear how it turns out
11:39 AM jepler: I wonder, in the absence of contact, how much the stylus acceleration would trail the housing; and whether that would change over time as the components aged
11:40 AM pcw_home: I would still use contact
11:41 AM archivist: this was me measuring contact force http://www.collection.archivist.info/archive/DJCPD/PD/2017/2017_01_08_arduino_probe/IMG_2165.JPG
11:41 AM pcw_home: it allows quick settling without active suspension
11:52 AM linuxcnc-build_: build #3296 of 1405.rip-wheezy-armhf is complete: Failure [4failed compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1405.rip-wheezy-armhf/builds/3296 blamelist: Norbert Schechner <nieson@web.de>
12:19 PM linuxcnc-build_: build #5146 of 0000.checkin is complete: Failure [4failed] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/5146 blamelist: Norbert Schechner <nieson@web.de>
12:25 PM jepler: it's 3d printing, so let's just ram the hotend into the table and use an accelerometer to notice it https://www.youtube.com/watch?v=pgSxb4t1gLo https://www.youtube.com/watch?v=dM--z7qAZ60
12:27 PM jepler: I guess they productized it, "HE280 metal hotend with acceleromter probe" https://www.seemecnc.com/products/he280-hotend-with-accelerometer-probe
12:28 PM jepler: also in a pinch it looks like you can use it as a tricopter
03:52 PM skunkworks: just about asked a stupid question... What voltage thcad did I have.. (then read there is a calibration sticker on the bottom)
05:44 PM skunkworks: scale works well - encoder -> scale -> actual voltage