#linuxcnc-devel Logs

May 28 2020

#linuxcnc-devel Calendar

03:10 AM rmu|w: regarding the "crazy move" bug 865: in my comment https://github.com/LinuxCNC/linuxcnc/issues/865#issuecomment-632688844 you can see that the last g-code line before abort that was appended to the TP was line nr. 2557
03:15 AM rmu|w: after the abort, 1001 queued items are discarded (everything up to line 2557), but linear move of line 2558 is appended
03:15 AM rmu|w: that is the "crazy move"
03:15 AM rmu|w: wherever this comes from, taskPlanReset seems to clear that queue, and taskPlanClose with lazy_close = 1 does not.
03:15 AM rmu|w: i cannot easily bisect on my test platform (rpi4) because hw support for the 7c80 would disappear rather quickly
05:49 AM cerna: And nos with 6dec6b35ef26ddc14ad311b8c24a1e8cd361bf2f whole buildbot is dead.
05:54 AM cerna: * And now with 6dec6b35ef26ddc14ad311b8c24a1e8cd361bf2f, the whole buildbot is dead.
05:55 AM cerna: And the 0000.checkin has 20 scheduled jobs, but is idle.
06:35 AM jepler: seb_kuzminsky: can you check into this please ? ^^
06:36 AM jepler: cerna: I am sure the buildbot has never had as many API requests as you've been sending it. Please consider disabling this temporarily or rate limiting it, just in case it is a part of whatever is going on.
07:09 AM cerna: jepler: I have just shut it down.
08:52 AM jepler: Thanks. I don't know for sure it's causing a problem but for now it's also not helping anybody
09:44 AM cerna: On the http://buildbot.linuxcnc.org/buildbot/grid the 4020.deb-jessie-rtpreempt-amd64
09:44 AM cerna: is signalling state of "building", but on http://buildbot.linuxcnc.org/buildbot/builders/4020.deb-jessie-rtpreempt-amd64 there is "No current builds" and "No Pending Build Requests" - I am not sure, but maybe it has a meaning and the problem originates from there.
09:47 AM cerna: BTW, that status checking script should not cause any rate problems, I think, because for every new push to branch (git SHA) it is checking only once (when everything is working normal) and then it polls only the Github for new pushes.
09:48 AM cerna: However, when it finds error on buildbot, then it doesn't update the tested git SHA and in next run it tests it again.
09:50 AM cerna: So I think the buildbot hugh-up was caused by something else. But truth is - if something caused the error, then cyclically checking the state could finish it off.
09:59 AM pcw_home: andypugh: Can you check if the problem is in 2.8? No, its not in 2.7 or 2.8, only master
10:14 AM pcw_home: ^^^^^^^^^^^^^^ Crazy move issue
10:21 AM cerna: Looks like something started happening: http://buildbot.linuxcnc.org/buildbot/buildslaves
10:23 AM pcw_home: rmu|w: I can try git bisect if I can find out where to start
10:54 AM rmu|w: i think first we should check if it is in latest 2.8 and 2.7
05:18 PM andypugh: pcw_home: Well, going back to the 2.9 tag would be logical. 10561703fc39f959f759d46ec473f221c8e00f24
05:27 PM pcw_home: That's the first 2.9 tag?
05:34 PM andypugh: Yes
05:40 PM pcw_home: OK Ill grab that and try this tomorrow