#linuxcnc-devel Logs

Mar 14 2018

12:28 PM dgarr: Re #384, is the pumakins.c test: "if (fabs(t1) < SINGULAR_FUZZ && fabs(t2) < SINGULAR_FUZZ){" valid ?, my notes:
12:30 PM dgarr: http://www.panix.com/~dgarrett/stuff/384_0.txt
12:36 PM seb_kuzminsky: dgarr: do you mind if i add those detailed observations to the github issue? might get more eyeballs there
01:03 PM jepler: looks like I was trying to fix the classic bug where kinematics would send an axis or joint to inf or nan and you could never get back to finiteland
01:04 PM jepler: I stand by the desire to check the return value of kinematicsInverse as well as the finiteness of all the joint positions.
01:05 PM dgarr: i agree if the pumakins test is valid -- but it is unfortunate that 2.7 continues with a detected singularity
01:07 PM dgarr: giving folks the impression that a config in 2.7 works but the same (converted) config in master fails
01:09 PM jepler: commenting out the line 'singular=1' within that condition would allow it to work. It would just change the return value and nothing else afaict
01:10 PM jepler: er, not commenting, because then it would only be set on one branch
01:37 PM dgarr: the sim config can also work with small values for [JOINT_4]HOME=.01 (for example) without triggering singular=1
02:00 PM jepler: TFW you look at github and discover that every commit in recent memory failed CI testing https://github.com/Kunena/Kunena-Forum/commits/K5.0
02:05 PM jepler: unfortuntely, the post edit bug is NOT fixed with the 5.0.14 release. It is reportedly fixed in 5.1, which is in beta and which I'm not likely to update to right the second it comes out
02:06 PM jepler: ... though with 5.0.14, released today, there's a note that it's the last 5.0.x release, which seems premature when 5.1 is not out of beta
02:23 PM jepler: I tried installing 5.1 on a snapshot of the forum and nothing's obviously broken, so that's good.
02:30 PM seb_kuzminsky: jepler: ouch
02:30 PM seb_kuzminsky: their most recent run: Ran for 3 min 50 sec
02:30 PM seb_kuzminsky: that's cute
02:35 PM rene-dev: jepler: you also can send in nans to the Feedback. But I can’t remember what happened...
04:34 PM jepler: I'm pretty sure the answer is "golly, you *could* but you sure *shouldn't*" (unless you're also sending the return value that indicates failure)