Back
[10:51:48] <KGB-linuxcnc> 03John Thornton 052.7 a4215d3 06linuxcnc 10docs/html/index.css 10docs/src/index.tmpl Docs: remove tool tips as I couldn't get them to work consistently. * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=a4215d3
[11:14:45] <linuxcnc-build> build #1713 of 1404.rip-wheezy-rtpreempt-amd64 is complete: Failure [4failed runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1404.rip-wheezy-rtpreempt-amd64/builds/1713 blamelist: John Thornton <bjt128@gmail.com>
[11:20:13] <seb_kuzminsky> what now
[11:20:50] <seb_kuzminsky> oh, this again:
[11:20:51] <seb_kuzminsky> task: 4776 cycles, min=0.000013, max=5.081448, avg=0.002442, 3 latency excursions (> 10x expected cycle time of 0.001000s)
[11:20:54] <seb_kuzminsky> halui: emcCommandSend: no echo from Task after 5.000 seconds
[11:21:42] <seb_kuzminsky> i bet it's some variation of this that's causing the problem Marius Alksnys reported on emc-users
[11:26:04] <skunkworks> could there be an issue with 2.7 taking longer? (ie the jogging issue)
[11:26:23] <skunkworks> *keyboard jogging
[11:26:53] <pcw_home> emcPTs issue?
[11:27:23] <seb_kuzminsky> skunkworks: not impossible, but i dont know of anything that would have caused it...
[11:28:48] <pcw_home> are the task time warnings a permanent addition?
[11:29:40] <seb_kuzminsky> i added them because i thought they would help diagnose these kinds of problems, but if people dont like it we can take it back out
[11:30:01] <pcw_home> I get those occasionally with 2.7-pre7 (none longer than a few 10s of ms)
[11:30:37] <pcw_home> I think its a good thing as long as its just logged
[11:31:09] <cradek> the problem with showing numbers is that people obsess uselessly over them
[11:31:21] <cradek> I am really wary of adding more of those
[11:31:24] <linuxcnc-build> build #3372 of 0000.checkin is complete: Failure [4failed] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/3372 blamelist: John Thornton <bjt128@gmail.com>
[11:31:24] <pcw_home> yes like latency
[11:31:33] <cradek> people obsess about latency numbers a LOT more than is reasonable
[11:45:08] <cradek> 1/30th of the messages in the forum mention latency
[11:45:16] <cradek> zero mentions of "excursions" so far
[11:49:19] <pcw_home> its an interesting subject. Ideally on on Ethernet connected device, a skipped cycle due to a CRC error should not cause a significant error
[11:50:09] <pcw_home> so we are talking about a full servo thread time latency
[11:50:37] <cradek> what happens now?
[11:51:11] <pcw_home> I thin it currently over-reacts
[11:51:43] <cradek> can hm2 just give last period's values?
[11:51:50] <pcw_home> there should be a reasonable timeout set
[11:52:26] <pcw_home> and a missed cycle should be indicated in a a hal pin so PID etc can skip updates
[11:53:04] <pcw_home> Damn third caltrop on the rug I found with bare feet
[11:53:55] <cradek> legos?
[11:54:05] <seb_kuzminsky> goat's head?
[11:54:12] <cradek> DIPs?
[11:54:36] <pcw_home> seeds
[11:58:48] <pcw_home> Yeah I guess it is called goats head
[12:11:27] <seb_kuzminsky> https://en.wikipedia.org/wiki/Goat%27s_head#/media/File:Acanthospermum_hispidum_seeds.jpg
[12:12:12] <seb_kuzminsky> nature, you scary
[12:14:03] <pcw_home> Like thats probably Kallstroemia californica
[12:14:15] <pcw_home> like that,
[12:26:40] <CaptHindsight> maybe someone should go back an rewrite the wiki sections about latency
[12:27:05] <CaptHindsight> it's from a time where it was more of an issue and it's still written from that perspective
[12:34:27] <skunkworks> pcw_home, did you see
http://www.machsupport.com/forum/index.php/topic,30680.msg212978.html#msg212978
[15:10:47] <skunksleep> zlog
[17:54:31] <KGB-linuxcnc> 03Dewey Garrett 05dgarr/missinglink c6630a1 06linuxcnc 10debian/linuxcnc.files.in 10debian/rules.in debian/rules.in: make soft link for nc_files * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=c6630a1
[19:00:52] <seb_kuzminsky> dgarr: that looks great
[20:10:32] <jepler> I didn't test it, but it seems like a good idea
[20:12:58] <jepler> seb_kuzminsky: do you have enough tuits to think about getting us a jessie buildslave? There's now a preempt-rt kernel for it in debian backports, making it a very plausible choice for linuxcnc 2.7 for people with smart hardware
[20:14:54] <jepler> I suppose I should see if the package actually builds, RIP runs great
[20:18:51] <skunkworks> exciting
[20:21:12] <jepler> still struggling with the live image builder
[20:21:18] <jepler> I tell it what kernel i want, and it goes and installs something else..
[20:23:59] <skunkworks> odd
[20:33:59] <jepler> actually it seems it installs the right kernel but the wrong kernel modules / udebs...
[21:13:24] <skunkworks> exit
[21:13:28] <skunkworks> heh
[21:15:09] <skunkworks> wkb
[21:15:17] <skunkworks> (wrong keyboard)
[21:22:50] <jepler> go north
[21:22:52] <jepler> open mailbox
[21:22:53] <jepler> take leaflet
[21:22:55] <jepler> go south
[21:31:08] <seb_kuzminsky> jepler: it's been on my todo list since i saw your post about jessie-backports
[21:35:10] * jepler kicks the live image builder
[21:35:12] <jepler> .. it is not behaving
[21:35:52] <jepler> seb_kuzminsky: yay, thank you
[21:35:55] <seb_kuzminsky> on the subject of buildslaves... is it time to turn off the hardy buildslaves?
[21:36:13] <seb_kuzminsky> they're only used for 2.5
[21:36:22] <jepler> are we shipping 2.7 on hardy or not?
[21:36:26] <seb_kuzminsky> should we make a final 2.5 release?
[21:36:37] <seb_kuzminsky> no, hardy has not been supported starting with 2.6
[21:36:56] <seb_kuzminsky> because of some build-dependency that hardy doesnt have
[21:36:58] <jepler> hardy is 8.04?
[21:37:00] <seb_kuzminsky> yeah
[21:37:10] <seb_kuzminsky> it was eol'd in 2013 iirc
[21:37:18] <seb_kuzminsky> or was it 2011?
[21:37:57] <seb_kuzminsky> hardy desktop eol 2011, server eol 2013
[21:37:59] <seb_kuzminsky> https://wiki.ubuntu.com/Releases
[21:38:10] <jepler> seemingly there are 66 commits since 2.5.4
[21:40:01] <jepler> too bad we don't have phone-home information about who's using what versions
[21:40:10] <jepler> <1/tau wink>
[21:40:47] <seb_kuzminsky> 1/tau?
[21:41:01] <jepler> tau being the improved circle constant
[21:41:01] <seb_kuzminsky> the 2.5 curve is flattening out, but still going up:
http://buildbot.linuxcnc.org/~seb/billions-served/
[21:41:09] <seb_kuzminsky> 2pi
[21:41:37] <jepler> yes that's the one
[21:41:47] <jepler> hm 2.6 and 2.7 not on that list
[21:41:57] <seb_kuzminsky> oops!
[21:41:59] * seb_kuzminsky hacks
[21:44:46] <jepler> I think the odds of actually making another 2.5 release are low; if someone is hit by a bug fixed in git but won't upgrade to 2.6, I'd suggest they use buildbot packages
[21:44:55] <jepler> there are some legit bugfixes in there, it's true
[21:55:56] <seb_kuzminsky> added 2.6 & 2.7:
http://buildbot.linuxcnc.org/~seb/billions-served/
[21:59:52] <seb_kuzminsky> that plot only shows downloads of buildbot packages, not downloads of packages from the official repo at wlo
[22:07:26] <cradek> we could sure make a last 2.5 release
[22:07:48] <cradek> I think that's my job but seb might have to upload/sign the packages since he touched it last
[22:46:57] <seb_kuzminsky> i'd be happy to deal with the 2.5 debian packages
[22:47:36] <seb_kuzminsky> i'd be happy to then eol 2.5-on-hardy so i can turn off the hardy buildslaves and devote those resources to jessie
[22:47:57] <seb_kuzminsky> we can still ship 2.5 on lucid, precise, and wheezy (and jessie)
[22:53:06] <cradek> cool, let's do that then, but not right now
[22:54:18] <seb_kuzminsky> agreed on both counts
[22:54:36] <seb_kuzminsky> it's really mostly just work for you, when you push the tag i
[22:54:42] <seb_kuzminsky> i'll copy the packages around
[22:55:32] <cradek> git branch -a; git tag -l; git remote -v
[22:55:46] <cradek> seb_kuzminsky: ok! I'll do it in the next few days
[22:55:57] <cradek> I agree that will be the last one.
[22:59:49] <seb_kuzminsky> no rush
[23:00:05] <seb_kuzminsky> as the man said, In your own time, comrades...
[23:00:07] <seb_kuzminsky> :-)
