These are chat archives for machinekit/machinekit

3rd
Dec 2015
John Morris
@zultron
Dec 03 2015 04:37
Oops, @kinsamanka you're right about the maintainer. The debian/changelog entry should have the name of the person who merged the commit. Somehow I thought you were talking about the Maintainer in debian/control.
John Morris
@zultron
Dec 03 2015 05:07
Ok, the debian/changelog entry looks a little better now. If we set the PKGSOURCE variable to something like mk.travis the version-release would be a slightly smaller mouthful:
0.1.1449118690-1mk.travis.pr814.git0d5bd75f~1wheezy
Bas de Bruijn
@luminize
Dec 03 2015 05:44
Hi guys, yesterday evening the Travis run tests failed on PR #813
but this morning everything seems ok suddenly
But the poster hasn't added another commit.
so what's going on there?
GP Orcullo
@kinsamanka
Dec 03 2015 06:02
You can restart the travis tests
There's an intermittent bug on the regressions
But #813 also failed on the buildbot?
Bas de Bruijn
@luminize
Dec 03 2015 06:20
Yes Yesterday, both failed but magically Travis not anymore
John Morris
@zultron
Dec 03 2015 07:19
It looks like we have an intermittent bug.
Hmm, looks like the old race conditions we used to see where output is truncated.
Michael Haberler
@mhaberler
Dec 03 2015 07:24
Which one, do you have result/street?
#795 looks immediately suspicious. :worried:
Actually, the latter is slightly different, even though the same tests/mux failed....
@mhaberler that ring a bell? Unexpected realtime delay on RT thread 1
...leading to lost output
GP Orcullo
@kinsamanka
Dec 03 2015 07:40
Shall we continue with packagecloud.io?
Michael Haberler
@mhaberler
Dec 03 2015 07:41
@zultron - what is the rationale #795 ? fixing behaviour when a runtime prereq is missing? why is rsyslog missing? shouldnt we test in an environment which matches what is being installed?
@kinsamanka - I guess @zultron needs map out how deb.dovetal-automata.com will b fed first - packagecloud.io certainly would be ok for any other folks running builds
GP Orcullo
@kinsamanka
Dec 03 2015 07:43
I’ve reached my limits yesterday :unamused:
Michael Haberler
@mhaberler
Dec 03 2015 07:43
oh ok, this is going on in the issue as well
well I think we should stick to relevant decisions discussed and mapped in the tracker and leave this chat for operational coordination - this is super confusing
John Morris
@zultron
Dec 03 2015 07:48
#795 allows MK not to choke when rsyslog isn't fast enough to keep up with log messages. Feel free to take it out. Those syslog_async settings need to be looked at more closely & something less hackish put in place.
Not sure whether that's the culprit now, though; the output used by the regression tests isn't piped through syslog_async, is it?
Off to bed, g'night
Michael Haberler
@mhaberler
Dec 03 2015 07:53
"the output used by the regression" you mean like the output of linuxcnc (DEBUG=0x...)? no, that is rcs logging, does not go to syslog
I fail to see the actual error in that huge log - what is actually the issue?
John Morris
@zultron
Dec 03 2015 17:46
The tests/mux thing couldn't be related to #202, could it?