#linuxcnc-devel Logs

Nov 28 2022

02:25 AM pere: seb_kuzminsky: 2019 edition of dxf2gcode is now in testing. did not move to the 2022 prerelease, no idea if you want a non-release uploaded. I hope you can have a look at the patches in <URL: https://salsa.debian.org/science-team/dxf2gcode/-/tree/debian/sid/debian/patches > and send upstream the ones you believe should go there, as you have been in touch with upstream before.
02:32 AM pere: seb_kuzminsky: there are still some appstream issues to fix, but I do not know how to fix them.
03:23 AM linuxcnc-build: build #2628 of 1660.rip-buster-python3 is complete: Failure [4failed compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1660.rip-buster-python3/builds/2628 blamelist: Petter Reinholdtsen <pere@hungry.com>
03:30 AM pere: andypugh: any idea how to avoid the build breaking merges? See <URL: https://github.com/LinuxCNC/linuxcnc/commits/master >
04:10 AM pere: testing increated timeout, 10->30 seconds. no idea if it will solve it.
05:14 AM linuxcnc-build: build #3030 of 1640.rip-buster-rtpreempt-amd64 is complete: Failure [4failed compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1640.rip-buster-rtpreempt-amd64/builds/3030 blamelist: Petter Reinholdtsen <pere@hungry.com>
06:21 AM linuxcnc-build: build #2789 of 1650.rip-buster-rtpreempt-rpi4 is complete: Failure [4failed compile] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1650.rip-buster-rtpreempt-rpi4/builds/2789 blamelist: Petter Reinholdtsen <pere@hungry.com>
06:21 AM linuxcnc-build: build #9432 of 0000.checkin is complete: Failure [4failed] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/9432 blamelist: Petter Reinholdtsen <pere@hungry.com>
07:25 AM linuxcnc-build: build #2790 of 1650.rip-buster-rtpreempt-rpi4 is complete: Failure [4failed compile runtests] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1650.rip-buster-rtpreempt-rpi4/builds/2790 blamelist: Petter Reinholdtsen <pere@hungry.com>
07:25 AM linuxcnc-build: build #9433 of 0000.checkin is complete: Failure [4failed] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/9433 blamelist: Petter Reinholdtsen <pere@hungry.com>
07:34 AM pere: did not help. adding more settings to .ini file to see if that help.
09:05 AM seb_kuzminsky: pere: the bad merge 96eea41a283 looks like a simple typo/"click-o"
09:06 AM pere: seb_kuzminsky: still, i would expect it to be discovered by the github CI before the merge...
09:07 AM pere: I'm trying random changes in the test-classicladder-estop-rpi4 to try to figure out why the test fail on rpi4 on the buildbot.
09:08 AM seb_kuzminsky: i think the problem was introduced by hand while running `git merge`, and github flagged it as soon as it saw it (ie when andy pushed the merge commit)
09:09 AM seb_kuzminsky: at least that's my guess, can't think of another way that could have slipped in there
09:10 AM seb_kuzminsky: the bug isn't in either of the two branches that got merged
09:10 AM seb_kuzminsky: do you have a raspberry pi or an armhf qemu vm to test on?
09:11 AM pere: my only rpi4 is busy.
09:11 AM pere: and my rpi 1 and 2 are too weak for me to try.
09:11 AM seb_kuzminsky: is the problem reproducible in a VM?
09:12 AM pere: do not know.
09:13 AM seb_kuzminsky: i'm booting up my sid armhf vm now, it's soooo sloooowwww
09:14 AM seb_kuzminsky: i wonder if anyone has run classicladder on arm, ever? since it's not in our test suite yet, it'd only get run by humans choosing it explicitly
09:25 AM pere: could be that it is simply broken on arm, yeah. time we figured out, if that is the case. :)
09:25 AM pere: I'll be offline for a while. see you later.
09:25 AM seb_kuzminsky: ok! see you :-)
09:49 AM roguish[m]: seb_kuzminsky: thank you for the explanation of the merge thing in your email. really helps a dumb mechanical understand and appreciate.
09:50 AM seb_kuzminsky: i'm glad that was helpful - i wish git didn't have such a steep learning curve, it's a super powerful tool but hard to learn :-/
10:22 AM seb_kuzminsky: ladder at least runs on bookworm arm64, which is what my pi happened to be running - i just started sim/axis/classicladder/demo_sim_cl.ini and opened the ladder editor and poked at it some, and it all seems to behave as expected
10:23 AM seb_kuzminsky: gonna try it on raspbian buster armhf, which is what the rpi4 in the buildbot is running
10:28 AM seb_kuzminsky: ok, that same config fails to start on buster, and fails in a way that looks just like pere's test in the buildbot, so it sure seems to be a problem with ladder on buster/armhf, not a problem with pere's test
10:28 AM seb_kuzminsky: and it leaves behind ipc objects in the same way
10:29 AM seb_kuzminsky: that's with some old build of 2.9
10:29 AM pere: seb_kuzminsky: aha.
10:31 AM pere: seb_kuzminsky: any clue what exactly is failing? seem like something did not start as it should?
10:50 AM seb_kuzminsky: i'm not exactly sure yet, but yes i agree with you, CL is not starting up as it should
10:57 AM pere: seb_kuzminsky: is there a way to see how it fail to start?
11:01 AM seb_kuzminsky: 2.8 fails in the same way, so the bug's been there for a while
11:02 AM seb_kuzminsky: the fix should go into 2.8 and be propagated up from there
11:03 AM seb_kuzminsky: pere: if you run the config you get the linuxcnc startup messages on stdout, that's the first place to look
11:04 AM seb_kuzminsky: the buildbot runs runtests with the -v flag, so you can see it in the log there too
11:05 AM seb_kuzminsky: eg, click on the most recent rpi4 failure up above, see that the runtests step is red, click on its stdio
11:05 AM pere: I did look, and it did not tell me anything.
11:05 AM seb_kuzminsky: there's an overwhelming amount of information in that log, but scroll down to the end to see that it's the classicladder/estop test that failed
11:05 AM pere: It did not even tell which process died, or if it died with a singnal or just exited.
11:06 AM seb_kuzminsky: search for the test name (classicladder/estop) to see the output from that test
11:06 AM seb_kuzminsky: the relevant part is probably this: http://paste.debian.net/1262175/
11:07 AM pere: probably, and it do not tell me anything.
11:08 AM seb_kuzminsky: heh, yeah there's not much to go on there, but that's where i'd start
11:08 AM seb_kuzminsky: oh, also i noticed that there's a core file from rtapi_app, not sure how that fits into the puzzle
11:09 AM pere: <URL: http://paste.debian.net/1262176/ > is the complete log, and did not help me.
11:10 AM seb_kuzminsky: so next i'm going to try to replicate the problem by hand
11:10 AM seb_kuzminsky: without linuxcnc running, i run `halcmd -kf` to get an interactive, empty realtime session
11:11 AM pere: my guess is that some program crash, and valgrind on the program could tell us why it crashes.
11:11 AM seb_kuzminsky: i see in the classicladder.hal file for the config i've been testing with that cl is loaded with `loadrt classicladder_rt period=1000000`
11:11 AM seb_kuzminsky: that's surprising to me - cl makes its own thread? ok, that's unique and i did not know that
11:13 AM seb_kuzminsky: lol, when i try that by hand it says: Unknown parameter `period=1000000'
11:13 AM seb_kuzminsky: so that doesn't seem right
11:17 AM pere: I copied it from configs/sim/axis/classicladder/...
11:17 AM pere: I have tried to trim away irrelevant parts, but am sure there is some fluff left.
11:20 AM seb_kuzminsky: if i load classicladder_rt by hand, and start the classicladder gui by hand, it works
11:21 AM seb_kuzminsky: i can load the sim/axis/classicladder .clp ladder config and it seems fine
11:22 AM seb_kuzminsky: i can poke the cl pins in halcmd and it seems to act right
11:22 AM seb_kuzminsky: bit ins & outs both work
11:23 AM seb_kuzminsky: that's a relief
11:24 AM seb_kuzminsky: it unloads cleanly
11:28 AM seb_kuzminsky: ok, time for me to go do something else, i'll be back later
12:20 PM pere: seb_kuzminsky: thanks for testing. so we have no idea what is wrong, then. :)
12:40 PM pere: video chat in 20 minutes on <URL: https://meet.jit.si/LinuxCNC-meeting-november-2022 >
03:19 PM pere: andypugh: tests/interp/g71-endless-loop/ and issue #1146 is the bug causing the gui program memory usage to explode.
03:21 PM pere: andypugh seb_kuzminsky: do you have pointers to the script trying to test all configs/*
03:23 PM pere: can the guis be controlled using halcmd commands? like disabling estop, powering on, homing, exiting?
04:13 PM seb_kuzminsky: pere: we don't have a script to test all the configs, because none of us knew about your trick about running in a VM, capturing screenshots, doing image recognition, and using that information to send key presses and mouse clicks
04:14 PM seb_kuzminsky: in general, GUIs can not be controlled except by a mouse & keyboard. The exception is "halui", which does not present a window or accept clicks, instead it's controlled by HAL pins.
04:15 PM seb_kuzminsky: there *are* some tests for halui (in tests/halui), but the challenge is testing the actual GUIs that humans interact with
04:17 PM pere: ok
04:25 PM pere: andypugh: look to me like the 2.9 branch do not have much build warnings.
04:28 PM andypugh: No, it’s quite manageable and nearly all docs-related: http://buildbot.linuxcnc.org/buildbot/builders/1640.rip-buster-rtpreempt-amd64/builds/3032/steps/compile/logs/warnings%20%28135%29
04:40 PM pere: andypugh: send you one pull request for two docs related ones.
04:41 PM pere: not quite sure if it is possible to write a generic configs/* test, as all interfaces seem to have different ways to exit.
04:42 PM pere: configs/sim/qtvcp_screens/qt_800x600.ini even refused to exit. would pop up a dialog asking if exiting is ok, and then ignore my 'yes' answer...
04:42 PM andypugh: They should all respond to a SIGTERM
04:44 PM pere: was hoping to use Alt-F Q or something like that to actually exercise the gui.
04:45 PM pere: time to sleep. see you later. :)
04:51 PM linuxcnc-build: build #2793 of 1650.rip-buster-rtpreempt-rpi4 is complete: Failure [4failed compile runtests] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/1650.rip-buster-rtpreempt-rpi4/builds/2793 blamelist: Petter Reinholdtsen <pere@hungry.com>
04:51 PM linuxcnc-build: build #9436 of 0000.checkin is complete: Failure [4failed] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/9436 blamelist: Petter Reinholdtsen <pere@hungry.com>
11:57 PM -!- #linuxcnc-devel mode set to +v by ChanServ