#linuxcnc-devel Logs

Oct 05 2017

10:40 AM skunkworks: jepler, is there a reason you picked 4.9.0-3-rt-amd64 #1 SMP PREEMPT RT Debian 4.9.30-2+deb9u2
10:42 AM skunkworks: for stretch?
10:42 AM skunkworks: I just tried linux-4.9.47.tar.xz patch-4.9.47-rt37.patch
10:42 AM skunkworks: it locks up this laptop
10:42 AM seb_kuzminsky: skunkworks: i think because that kernel was already packaged for stretch (by debian.org), and showed reasonable stability and performance on a wide(?) range of hardware
10:42 AM skunkworks: oh - cool
10:42 AM skunkworks: ok
10:56 AM skunkworks: Hi seb
10:57 AM seb_kuzminsky: hola
11:00 AM skunkworks: actually - I don't know for sure - it was either the kernel - or virtualbox that caused it to be unstalbe.
11:00 AM skunkworks: I think it booted a couple times before I installed virtualbox.
11:01 AM skunkworks: but 4.9.30 seems to be working just fine
11:09 AM jepler: yes, I chose the kernel that was packaged by debian
11:10 AM jepler: if there's something that's at least as unfun as building a debian distribution, it's doing debian packaging of a patched kernel
11:30 AM jepler: .. the kernel is up to deb9u5 and has security fixes, maybe I should spin another CD
11:48 AM skunkworks: http://electronicsam.com/images/KandT/testing/FunWithLinux.png
11:48 AM jepler: you poor guy, forced to use software from all of microsoft, oracle, and autocad
11:48 AM skunkworks: :)
11:49 AM skunkworks: it's not that I am forced to.. Just easier. ;)
11:51 AM skunkworks: just updated to u5
11:51 AM Roguish: skunkworks: hey, how is the fusion360 post for linuxcnc?
11:51 AM Roguish: anything special or wierd?
11:51 AM skunkworks: Roguish, I am really green with fusion360. Only drawn a few things - and have not generated actual code.
11:51 AM Roguish: ok. sort o fmy level also.
11:52 AM skunkworks: I think it is probably pretty good - tormach pushes fusion for pathpilot
11:58 AM seb_kuzminsky: i've been using solvespace and pycam, all open source, all in linux, that's a pretty terrible tool chain too :-/
12:02 PM skunkworks: heh
12:04 PM skunkworks: solvespace looks pretty cool
12:04 PM skunkworks: I don't think I have heard of that
12:05 PM jepler: I've heard of solvespace, I should try it again.
12:05 PM jepler: 2.3 is in debian stretch, no excuses not to
12:29 PM seb_kuzminsky: yeah, the stretch version works quite well
12:32 PM seb_kuzminsky: i mostly use it to make constained, dimensioned 2d drawings, export them as SVG, import that into pycam, and make g-code from that using the "engraving" process
12:33 PM seb_kuzminsky: i'm not fancy enough to do real 3d work...
12:33 PM jepler: while I'm only going to play with my computer controlled glue gun, so I want fully 3D models
12:40 PM seb_kuzminsky: i've made some 3d models in solvespace and it worked as far as i can tell
12:40 PM seb_kuzminsky: it does stl export
12:44 PM jepler: yes, I saw that it does
12:49 PM seb_kuzminsky: it's got a ui that feels like it was designed by a programmer...
01:10 PM hazzy: solvespace looks very promising
01:11 PM seb_kuzminsky: it feels more energetic and vibrant than freecad, which is the other reasonable open source option that i know about
01:11 PM hazzy: It is the only cad package were I have been able to make some kind of model in the first 30s
01:11 PM hazzy: It is very intuitive
01:12 PM seb_kuzminsky: solvespace? yeah
01:12 PM seb_kuzminsky: and the keyboard shortcuts make it pretty fast after you've spent ~30 minutes with it
01:12 PM hazzy: yes
01:12 PM hazzy: They are the same shortcuts I have spent a lot of time adding to solidworks
01:13 PM hazzy: So I feel right at home
01:13 PM seb_kuzminsky: i think freecad is easier to transition to if you're coming from solidworks or probably any of the big windows cad programs
01:13 PM Tom_L: does it have cam too?
01:13 PM seb_kuzminsky: it's got basic cam, but i haven't played with it
01:13 PM seb_kuzminsky: it might be awesome, for all i know
01:13 PM Tom_L: looks like 2.5D
01:50 PM seb_kuzminsky: any reason not to use high flute-count endmills? i've got a slow spindle (3500 rpm) but plenty of power
01:51 PM seb_kuzminsky: price, i guess, the more flutes, the more expensive the endmills are
01:51 PM seb_kuzminsky: though maybe they last longer since the wear is spread over more cutting edges?
02:13 PM mozmck: Off the top of my head from memory, I think the higher count flutes can give a smoother finish, but aren't as good for hogging out material.
02:30 PM Tom_L: 4 flute are generally used for steel, 2 or 3 for aluminum
02:30 PM Tom_L: large roughing endmills are sometimes even 6 or more flutes but are high helix
02:30 PM mozmck: Interesting
02:31 PM Tom_L: if you use 4 flute on aluminum the chips are too big to evacuate and clog and gauld the cutter
02:32 PM mozmck: I was thinking part of it was chip clearance. I guess if you go slower it would work?
02:32 PM hazzy: Leading to stir welding, I have mucho experience ...
02:32 PM Tom_L: lower chipload
02:32 PM Tom_L: typically aluminum is .004" or more per tooth pushing up to .010"
02:33 PM Tom_L: https://www.mscdirect.com/product/details/51710861
02:34 PM Tom_L: roughing endmills have 'wavy' flutes and look similar to a tap
02:34 PM Tom_L: larger ones 3/4" and up will likely have 6 or more flutes
02:37 PM Tom_L: http://www.lakeshorecarbide.com/ has good cutters
02:56 PM andypugh: Ref: https://github.com/LinuxCNC/linuxcnc/issues/238
02:57 PM andypugh: Currently the reason that the first attempt at a fix went wrong was due to the way that the ioaddr modparam is pre-loaded with likely addresses: https://github.com/LinuxCNC/linuxcnc/blob/master/src/hal/drivers/mesa-hostmot2/hm2_7i43.c#L45
02:58 PM andypugh: So without counting config string the driver has no way to know when to stop looking, and generally will fail to find a board at 0x278 even if it finds one at ox378
02:59 PM andypugh: I propose to remove the 0x278 from the string. This plausible-guessing behaviour is undocumented, and I wouldn’t expect that anyone is relying on it.
03:01 PM andypugh: (The alternative, possibly safer if anyone _is_ relying on the current behaviour) is probably to keep a flag about whether there were any ioaddr specified, guess some numbers if not, and and then alter the behaviour on failure depending on the flag.
03:02 PM andypugh: But that seems elaborate for the corner-case of someone with 2x7i43, one on 0x378, one on 0x278 who _also_ doesn’t specify the ioaddr in his loadrt line.
03:21 PM andypugh: seb_kuzminsky: I am proposing to put this in 2.7 as a bugfix, so woud that possible regression worry you?
04:36 PM hazzy: How does one compile hal comps with a RIP?
04:36 PM hazzy: I am using `halcompile --compile mpg.comp`
04:36 PM hazzy: But I get this error:
04:37 PM hazzy: gcc: error: comps: No such file or directory
04:38 PM jepler: halcompile/extralib/test.sh:halcompile --compile extralib_test.comp
04:38 PM jepler: that is the same as our testsuite does
04:38 PM jepler: since our testsuite passes, we believe it works
04:39 PM jepler: note that of course you have to ". scripts/rip_environment" or "halcompile" will invoke an installed version if there is one
04:40 PM jepler: jepler@rat:~/src/linuxcnc/master$ . scripts/rip-environment
04:40 PM jepler: jepler@rat:~/src/linuxcnc/master$ ./scripts/runtests tests/halcompile/extralib/Running test: tests/halcompile/extralib
04:40 PM jepler: Runtest: 1 tests run, 1 successful, 0 failed + 0 expected
04:40 PM hazzy: ok thanks! I was using the halcompile found in linuxcnc-dev/bin
04:40 PM jepler: you can try running that exact test. If it succeeds but building your "mpg.comp" fails, then look for problems in your comp.mpg.
04:41 PM hazzy: Ok, the test run fine
04:48 PM jepler: If you set the environment variable V=1 you can see the commands that halcompile --compile invokes
04:49 PM jepler: e.g., $ V=1 bin/halcompile --compile tests/halcompile/extralib/extralib_test.comp
04:49 PM jepler: (this is not documented anywhere, sadly)
04:49 PM hazzy: Ok will do. I think I am missing something simple. I never have had problems compiling these same comps before
04:49 PM hazzy: Here is what I have tried:
04:49 PM hazzy: . /home/kurt/Sources/linuxcnc-dev/scripts/rip-environment
04:50 PM hazzy: halcompile --compile mpg.comp
04:50 PM hazzy: gcc: error: comps: No such file or directory
04:50 PM hazzy: Does it matter the dir I compile the comps in?
04:51 PM hazzy: In the past I have always done it on the Desktop.
04:51 PM jepler: -option extra_link_args "-lm";
04:51 PM jepler: +option extra_link_args "-lm comps";
04:51 PM jepler: If I change extralib_test.comp like this ^^^ I get the error: gcc: error: comps: No such file or directory
04:52 PM jepler: so an error in your file mpg.comp could fully explain what you are seeing
04:53 PM hazzy: I can compile fine with the installed 2.7 LCNC though ...
04:54 PM hazzy: Just not the using the RIP
04:54 PM jepler: is your RIP also 2.7?
04:54 PM hazzy: No, recent 2.8
04:55 PM jepler: well I don't know what is wrong, but I believe the problem is in this file "mpg.comp" which I don't have
04:55 PM hazzy: I think you are probably right, will fiddle some more. thank you!
05:05 PM skunkworks: heh - the limit3 run away bug seems to be easy for me to trigger :)
05:11 PM seb_kuzminsky: i wrote tests that trigger the two known limit3 bugs, but then i never got around to fixing limit3 :-/
05:13 PM seb_kuzminsky: andypugh: i dont know if anyone's relying on that behavior or not...
05:14 PM andypugh: I guess we can’t know, so the only safe thing to do is assume that they are.
05:14 PM andypugh: It’s all init code, so no penalty
05:16 PM seb_kuzminsky: yeah, that would be safer
05:16 PM seb_kuzminsky: i'd hate to break 2.7 *again*
05:16 PM seb_kuzminsky: once, no twice was enough for me
06:34 PM andypugh: Well, this time I do have the hardware
06:34 PM andypugh: So can actually test.
08:19 PM jepler: the workaround is to specify any non-empty config= ? I'd be tempted to leave 2.7 alone, too