#linuxcnc-devel Logs

Jul 09 2017

#linuxcnc-devel Calendar

09:44 AM jepler: digitalocean's function to promote a backup to a new virtual machine is not quite as handy as zfs for testing a big upgrade, because when it's successful you have to repeat the whole process again on the real machine
09:46 AM jepler: but it's still a lot better than having nothing to do when an upgrade goes badly
03:43 PM jepler: just 4 days to 1.5 billion seconds in the unix epoch
04:58 PM jepler: https://emergent.unpythonic.net/files/sandbox/linuxcnc-cdf.png
04:59 PM jepler: I was curious to know how old the average line of source code was
04:59 PM jepler: using commit time information from git blame, I produced this graph, which is intended to answer that question
04:59 PM cradek: neat
04:59 PM jepler: about 50% of the lines in src/ have been written since 2012
05:00 PM cradek: not very old, it seems
05:00 PM jepler: cradek: I wonder if you have one of those machines that pcw says the video has been broken in all modern linux releases :-/
05:01 PM cradek: I rebooted a few more times and it came up right once
05:01 PM cradek: I think it's building the nvidia driver now, but it's taking a super long time
05:01 PM jepler: hmph
05:01 PM jepler: afk
05:01 PM cradek: (I installed nvidia-driver package)
05:01 PM cradek: I think it's a not-even-super-old pcie nvidia card
05:11 PM pcw_home: The system I have problems with are all older Intel Graphics, they all eventually crash with 4.x kernels (eventually may mean weeks)
05:11 PM pcw_home: they never seem to crash with 3.18
05:49 PM KimK: jepler: What were the two big vertical "jumps" in spring 2012 and fall 2014? (Yes, a lot of work was committed all at once, but what was the goal/feature/branch?) There are some smaller jumps that might be interesting to know too. Thanks for posting this graph. I'll scroll back later.
06:08 PM jepler: KimK: probably some .glade file, they're massive and diffs (and therefore git blame) don't really do a good job of summarizing changes
09:31 PM jepler: KimK: the one in 2012 was the addition of pncconf
09:35 PM jepler: > i add missin modules. show you how use module_param.
09:35 PM jepler: oh these old commit messages are best
09:58 PM MrHindsight is now known as CaptHindsight
10:43 PM Tom_itx: is STEPGEN_MAXVEL not necessarily needed now? the docs indicate it was for older configs
10:45 PM pcw_home: I think its ok at the default of 0 (no limit)
10:46 PM Tom_itx: is it needed for the hostmot2 driver?
10:46 PM Tom_itx: it was in my original config by default
10:47 PM Tom_itx: which it pretty old
10:50 PM pcw_home: I dont think its needed
10:50 PM pcw_home: STEPGEN_MAXACCEL may be needed for the built in driver (depending on servo thread jitter)
10:50 PM pcw_home: but not if you use PID to close the stepgen loop
10:50 PM pcw_home: (or the DPLL)
10:56 PM Tom_itx: is DPLL enabled by default?
10:56 PM pcw_home: normally the DPLL in only present on Ethernet configs
10:56 PM Tom_itx: ok
10:56 PM pcw_home: (but can be on any)
10:57 PM Tom_itx: i didn't recall anything about it
10:57 PM pcw_home: its easy to have a few 100 usec of read jitter on Ethernet/slow PCs
10:57 PM Tom_itx: recovering from a hdd crash and i figured i'd try to clean up the configs a bit during the process
10:58 PM pcw_home: that makes a lot of position sampling noise when reading the stepgen position
10:58 PM Tom_itx: i'm getting some following errors right now
10:59 PM pcw_home: this gets translated into crazy/bogus stepgen speed corrections
11:00 PM pcw_home: the PID scheme is much better as far as stepgen fidelity goes
11:01 PM pcw_home: uggh, the hm2_eth debug printouts dont show the write data
11:02 PM Tom_itx: i did check the card and my config is still on it
11:02 PM Tom_itx: i can compare it to a recent bitfile can't i?
11:03 PM Tom_itx: pretty sure i have the right bitfile after the recovery
11:03 PM Tom_itx: but i'd like to check
11:38 PM pcw_home: yeah you can use verify but only if the bitfiles are identical (same creation date etc)