#linuxcnc-devel Logs

Dec 07 2022

07:42 PM seb_kuzminsky: Tom_dev: the screenshot you posted above shows an old build, the one in debian.org
07:49 PM seb_kuzminsky: i bet you have linuxcnc-uspace.deb installed on the machine where you're testing this, and you're building from git but running the installed binaries in /usr/bin
07:51 PM Tom_dev: that looks alot better
07:51 PM Tom_dev: d = 0.000000 and circle angle = 6.283185
07:51 PM Tom_dev: d = 0.000000 and circle angle = 6.283185
07:51 PM Tom_dev: d = 0.000000 and circle angle = 6.283185
07:51 PM Tom_dev: d = 0.000000 and circle angle = 6.283185
07:51 PM Tom_dev: and a nice pretty helix
07:52 PM andypugh: OK, so fixed for you too.
07:52 PM andypugh: (I am starting to suspect that the OP is also not running the version he thinks he is)
07:52 PM Tom_L: http://tom-itx.no-ip.biz:81/~webpage/cnc/arc_test/axis_test_fix.png
07:52 PM Tom_dev: 2.10
07:53 PM Tom_dev: what was it?
07:53 PM andypugh: Starting to feel a bit more relaxed. It’s not a comfortable feeling that you have broken the second most important command in Linuxcnc
07:53 PM Tom_dev: :D
07:54 PM Tom_dev: but you also fixed it
07:54 PM andypugh: Yes, but it took nearly a year to even notice! How many thread-mills have been trashed in that time?
07:54 PM Tom_dev: something in that CART_FUZZ line?
07:54 PM Tom_dev: none of mine
07:55 PM * Tom_dev removes the print line
07:55 PM andypugh: The bug I just fixed? Ir was checking that the end point and start point are very close to each other, and if so making a full-circle arc. But the ends of a spiral are _not_ close to each other.
07:56 PM Tom_dev: ahh
07:56 PM Tom_dev: but the x and y would be in G17 etc
07:56 PM Tom_dev: just not z
07:57 PM Tom_dev: well, don't lose any more sleep over it
07:58 PM andypugh: But what complicates things is that most of the time earlier code calculates a zero-radian arc, then subtracts that from 2pi if a bit of matrix maths says that it needs to “ /* now angle is in range 0..PI . Check if cross is antiparallel to
07:58 PM andypugh: normal. If so, true angle is between PI..2PI. Need to subtract from
07:58 PM andypugh: 2PI. */“
07:59 PM andypugh: And much of the time that results in a correct spiral arc by accident.
07:59 PM Tom_dev: i assume it works in the other 2 workplanes.. i never tested those
07:59 PM andypugh: I did,
07:59 PM Tom_dev: i still have your test code
07:59 PM Tom_dev: but it ran ok here
07:59 PM Tom_dev: once i got it to run
08:00 PM Tom_dev: (z offset required)
08:00 PM andypugh: The fix is to project the start and end points to the working plane and _then_ calculate the distance between the endpoints.
08:01 PM Tom_dev: i couldn't write it but do understand what you're saying
08:02 PM Tom_dev: basically make it 2d then test it
08:27 PM seb_kuzminsky: pere: nope, i can no longer get the tests to fail in docker on unstable/sid
08:31 PM seb_kuzminsky: ugh
10:15 PM skunkworks: andypugh: I do a ton of helical plunges and have not had a problem...