Jun 25 2019
03:41 PM mozmck: andypugh: run from line has been broken for a long time.
03:44 PM andypugh: Yes, but that doesn’t mean that we shouldn’t look at how it ought to work.
03:44 PM andypugh: And he seems to be describing something even odder.
03:44 PM mozmck: Correct. His description sounds similar to what I found. I'm trying to find the bug report now
03:44 PM mozmck: I implemented a simple run from line function to work around it because no one at the time could figure it out.
03:47 PM mozmck: https://forum.linuxcnc.org/38-general-linuxcnc-questions/30276-run-from-line-and-v-2-7-3?start=10
03:49 PM mozmck: https://sourceforge.net/p/emc/mailman/message/34839325/
03:54 PM Tom_L: does it work the same?
03:55 PM Tom_L: even with remap?
03:56 PM mozmck: I haven't used remap. What my function does is if it is enabled in the INI, run from line will simply start at the given line without going through the code to set state.
03:56 PM mozmck: You have to be careful where you start, but for us it has worked quite well for several years.
03:57 PM mozmck: I plan to put it in 2.8 as soon as I get a chance. Things have gotten busy here for a spell
03:57 PM Tom_L: i've had a few issues getting the spindle to stay on from some lines
03:57 PM Tom_L: just add an M3 on the line above and run from it
03:58 PM Tom_L: my issue with remap was that it caused multiple tool changes to prior used tools in the file
04:01 PM mozmck: My problem was that a probe routine in an external file subroutine causes really bad stuff if you try to run from any line after the call to that sub
04:54 PM andypugh: I generally don’t even try to start from anywhere but a tool change.
06:23 PM Tom_L: when it's a largeish file and i know where i need to start, i will but now i know to add an M3 in
06:54 PM CaptHindsight: RTAI 5.2 is working here
06:55 PM CaptHindsight: first time stable
06:55 PM CaptHindsight: 4.14 kernel
06:59 PM memfrob: Hey everyone, I just got RTAI 5.2 working, no math support yet, but latest 4.14 kernel (4.14.130, posted today) with the latest IPIPE code from Xenomai, would anyone be willing to try the testsuite?
06:59 PM memfrob: https://github.com/NTULINUX/RTAI/tree/for-next
07:00 PM jthornton: cool
07:00 PM memfrob: On a Debian sid Ryzen system, latency is all over the place, on my AMD Fusion APU, latency is 10 microseconds on Gentoo.
07:00 PM andypugh: memfrob: I will have a go tomorrow night. (It’s too late here now)
07:00 PM memfrob: Thank you andypugh!
07:01 PM andypugh: Just emailed myself the link :-)
07:05 PM memfrob: Just note that it will _not_ work with LinuxCNC -- I have some more cleaning up to do and I'll add newlib libm into the source tree later on, for now I need to make sure the kernel/latency and base functionality is working.
07:06 PM memfrob: Building it is simple: ./autogen.sh ; make menuconfig ; make ; sudo make install ; sudo bash /usr/realtime/testsuite/run
07:07 PM memfrob: I also need to blacklist several kernel options as a lot of them are problematic, such as IOMMU and several security/hardening options.
07:09 PM memfrob: http://dpaste.com/21GCFWH -- My latency scores on Ryzen
10:34 PM memfrob: Well I got newlib building with the tree but it doesn't work.. lots of undefined symbols in the modules, practically all of them.
10:34 PM memfrob: https://github.com/NTULINUX/RTAI/commit/de6678e5178752618478fde106a026e7f60fd8a4
10:35 PM memfrob: http://dpaste.com/10316MS
10:36 PM memfrob: I ran into these problems before many times years ago, I'm done for the night.
10:58 PM pcw_home: ~