#linuxcnc-devel | Logs for 2013-07-23

[02:51:33] <KGB-linuxcnc> 03chrisinnanaimo 05master 6ec4c49 06linuxcnc 10configs/sim/gscreen_custom/industrial.glade 10src/emc/usr_intf/gscreen/gscreen.py * gscreen -add jog mode LED indicator
[02:51:33] <KGB-linuxcnc> 03chrisinnanaimo 05master 6af8a34 06linuxcnc 10src/emc/usr_intf/pncconf/pncconf.py * pncconf -add debugging code
[02:51:37] <KGB-linuxcnc> 03nieson 05master 864f7ef 06linuxcnc 10configs/sim/ 10gmoccapy/gmoccapy.glade 10gmoccapy/gmoccapy_handler.py 10gmoccapy/release_notes.txt * gmoccapy - bugfix in restart dialog and spindle switching with logging on
[03:11:00] <KGB-linuxcnc> 03chrisinnanaimo 05master 2671b21 06linuxcnc 10src/emc/usr_intf/gscreen/gscreen.py * gscreen -fix run-at-line dialog bugs
[07:10:01] <mhaberler> jepler: ulapi autoloading works as expected, I was able to back out the change to runtests
[10:11:34] <cradek> cmorley1: sorry, I forgot you told me you were waiting on a release. I think elson's stuff is all done now, so I'll get started as soon as I can.
[10:16:58] <seb_kuzminsky> 2.5.3? awesome :-)
[10:17:02] <seb_kuzminsky> let me know if i can help
[10:17:40] <cradek> thanks, but you already did your part
[10:18:14] <seb_kuzminsky> i did?
[10:18:19] <cradek> buildbot
[10:18:27] <seb_kuzminsky> oh
[10:18:34] <seb_kuzminsky> :-)
[10:19:58] <seb_kuzminsky> give me a heads up when you're ready to push the release build and i'll keep an close eye on it for that source fetch bug we keep running in to
[10:20:29] <cradek> ok
[10:20:43] <cradek> I don't remember whether we thought we solved it
[10:20:56] <seb_kuzminsky> i think we did, but i've thought that before
[10:21:17] <cradek> for 2.5.2 I think something got confused because bb wasn't idle when I pushed, or something like that
[10:23:23] <seb_kuzminsky> that shouldn't cause a problem
[10:24:54] <seb_kuzminsky> i think the problem was that "apt-get source" by default thinks the package name you ask it for is the binary package you want source for, and uses fuzzy matching to find the source package name
[10:25:17] <cradek> oh that's right
[10:25:23] <seb_kuzminsky> i think you suggested to add "--source-only" to the "apt-get source" command line, and the problem went away after that
[10:25:37] <cradek> oh yay
[10:25:50] <cradek> you didn't put the libmodbus change in 2.5 did you?
[10:25:51] <seb_kuzminsky> at least that's what i think my terse commit message from march means ;-)
[10:25:57] <seb_kuzminsky> no
[10:26:24] <seb_kuzminsky> i didnt touch 2.5 for the libmodbus change at all
[10:27:20] <seb_kuzminsky> 277 commits in 2.5 since 2.5.2
[10:27:46] <seb_kuzminsky> and much of that is useless weblate crap
[10:27:52] <cradek> 133
[10:28:18] <cradek> brb
[10:34:16] <seb_kuzminsky> i monkeyd with the Developers' Manual pdf while we were at fest, 85539c2bcca82d6265205eb9a5f10988b1fbdbe1
[10:34:22] <seb_kuzminsky> (that's in 2.5)
[10:34:51] <seb_kuzminsky> i think it's better now but maybe another pair of eyes would be good
[11:11:10] <jepler> I'm sad the weblate experience was not a positive one.
[11:15:17] <cradek> I'm not sure it wasn't?
[11:25:42] <seb_kuzminsky> i'm not sure either
[11:26:30] <seb_kuzminsky> my whining was mostly because it generated tons of git commit (which is not actually a bad thing), and because all the ones i spot-checked are for a language we don't build
[11:28:54] <cradek> oh, hmm
[11:29:32] <cradek> I only found 18ish things that want changelog entries
[11:30:25] <cradek> halui jogging wasn't broken in 2.5
[11:30:53] <cradek> (I mean the behavior wasn't broken - the code was wrong)
[11:31:11] <seb_kuzminsky> yes
[11:34:55] <jepler> seb_kuzminsky: ugh, really? missing makefile juice?
[11:35:19] <seb_kuzminsky> zh_HK is chinese, right? we don't have that
[11:35:42] <jepler> TARGETS += $(patsubst po/%.po, \
[11:35:42] <jepler> ../share/locale/%/LC_MESSAGES/linuxcnc.mo, \
[11:35:42] <jepler> $(wildcard po/*.po))
[11:35:43] <jepler> it's supposed to just wildcard all the po files
[11:36:05] <seb_kuzminsky> oh hrm
[11:36:23] <seb_kuzminsky> nevermind, i was thinking of our docs, not the localization of our software
[11:36:29] <seb_kuzminsky> ignore me
[11:36:33] * seb_kuzminsky gets more coffee before typing again
[11:41:53] <seb_kuzminsky> cradek: got a schedule in mind for 2.5.3? just curious
[11:42:29] <cradek> next day or two I hope
[11:43:04] <seb_kuzminsky> oh wow!
[11:43:05] <cradek> mayyyybe I'll want to build it tomorrow night
[11:44:28] <jepler> > As specified in the Parallella reference manuals, the Parallella board is sold as a development board to be used only by qualified individual.
[11:44:50] <jepler> I wonder what in the world this means
[11:45:05] <cradek> there's only one person who's qualified to use it?
[11:46:14] <skunkworks> wow - I just installed 8.04 on a pentum 3 - 733.. around 15us latency.
[11:48:01] <pcw_home> wonder if they lose money on every one so dont want to sell many at this point
[12:00:13] <cradek> skunkworks: P3s all seemed to work great
[12:00:55] <cradek> my dual P3-1000 has about 5us latency when using isolcpu
[12:12:49] <memleak> I get around 5us seconds too with gentoo and a completely custom kernel / RTAI build :D
[12:14:10] <seb_kuzminsky> memleak: on an old P3?
[12:14:57] <memleak> oh.. I missed that part.. AMD Phenom II X4 955 Quad-core 3.2 Ghz CPU
[12:15:04] <memleak> :/
[12:15:12] <memleak> quite the opposite!
[12:18:00] <skunkworks> heh
[12:43:27] <seb_kuzminsky> that's great memleak - very encouraging :-)
[13:09:56] <CaptHindsight> jepler: it's more of a "we are warning you that it's a dev board and we don't know what does and doesn't work and the software is not by any means stable, so don't buy one and complain that it doesn't work for you and you need your hand held"
[13:11:52] <CaptHindsight> "We are pleased so say that we were able to keep the entry point at $99 without losing money!"
[13:13:18] <CaptHindsight> their kickstarter was funded ~9 months ago
[13:13:39] <CaptHindsight> http://www.kickstarter.com/projects/adapteva/parallella-a-supercomputer-for-everyone
[13:14:30] <pcw_home> They must have a heck of a deal from Xilinx (the 7020 is ~$200 in low qty)
[13:18:58] <CaptHindsight> Xilinx doesn't seem to be interested in any low volume applications for ZYNQ, similar to Samsung, Nvidia, Qualcomm etc
[13:23:05] <CaptHindsight> Parallella is trying to capture many of the fans if the Rpi, i.mx6 board
[13:23:39] <CaptHindsight> and similar low cost ARM boards
[13:25:10] <pcw_home> Not so sure. It does not have hacker friendly I/O connectors and seems more aimed at big SMP systems
[13:28:42] <pcw_home> But if they can actually continue to sell it at $99.00, its a real bargain
[13:31:29] <jepler> yeah, I was one of the KS backers though I have no idea what I'll do with the boards
[13:37:58] <pcw_home> Thats Its design is tilted toward SMP is indicated by the fact that it has more epiphany pins
[13:37:59] <pcw_home> than FPGA pins available. Still amazing for the price even with just 24 FPGA pins
[13:41:25] <KGB-linuxcnc> 03chris 05v2.5_branch ce9835a 06linuxcnc 10VERSION 10debian/changelog * Release 2.5.3
[13:41:26] <KGB-linuxcnc> 03chris 05signed tags a679654 06linuxcnc 03v2.5.3 * Release 2.5.3 (tagged commit: ce9835a)
[13:42:57] <cradek> ALL NB: Please do not go announcing this everywhere without disclaimers. It might still be a couple days to finish the release. Premature announcements caused a bunch of questions and confusion last time.
[13:47:49] <jepler> cradek: got it
[13:52:29] <cradek> but also: yay!
[13:52:40] <jepler> cradek: thank you for your work as RM
[13:53:05] <cradek> welcome!
[13:55:32] <mhaberler> jepler: still time for a guru-mode question?
[13:55:48] <jepler> mhaberler: yeah apparently I leave on thursday, not tuesday as I thought
[13:55:54] <jepler> I was sad, had expected a 1-day workweek.
[13:55:59] <mhaberler> superb ;)
[13:56:42] <mhaberler> my assumption was that linking with shared libs is transitive - i.e. I link liblinuxcnchal.so against liblinuxcnculapi.so which has the interceptor as discussed
[13:56:45] <mhaberler> on i386 this works fine
[13:57:00] <mhaberler> on a x86_64 wheezy box I get this:
[13:57:11] <mhaberler> usr/bin/ld: objects/hal/components/sampler_usr.o: undefined reference to symbol 'rtapi_switch'
[13:57:39] <jepler> the i386 box is also wheezy?
[13:57:46] <mhaberler>
[13:57:47] <mhaberler> and /usr/bin/ld: note: 'rtapi_switch' is defined in DSO /src/emc2-ub/lib/liblinuxcnculapi.so.0 so try adding it to the linker command line
[13:57:56] <mhaberler> no, 10.04/atom
[13:58:38] <mhaberler> same submakefile.. seems the linker on wheezy ignores the fact that I already linked hallib agains ulapilib
[13:58:56] <mhaberler> thats a new one to me
[13:59:19] <mhaberler> note John added rpath commands to the linker flags to resolve the flavor-dependent path
[14:00:09] <jepler> apparently this is a deliberate change in linking .. the first ref I found is about fedora, though. http://fedoraproject.org/wiki/UnderstandingDSOLinkChange
[14:00:31] <mhaberler> hm
[14:01:15] <mhaberler> well I liked the idea that we have a library API to userland RTAPI now without HAL ontop
[14:01:46] <mhaberler> I hoped to get away without fudging all the places where liblinuxcnchal.so is referenced
[14:02:21] <mhaberler> sure I can drop the link of hallib against ulapilib and add ulapilib as needed
[14:02:38] <mhaberler> thats probably more robust in the face of this change
[14:05:24] <jepler> I'd suggest to leave it as one library for now and tackle it another day
[14:05:42] <mhaberler> ok
[14:08:05] <jepler> huh only about 24kB of code in liblinuxcnchal.so here, that's positively tiny
[14:26:32] <seb_kuzminsky> oh man i love seeing the linuxcncrsh tests not fail
[14:36:18] <jepler> me too
[14:39:14] <cradek> good so far (the first deb- builds have the right version and say is-release)
[14:39:54] <cradek> iirc it was always lucid-rt-binary-i386 that got it wrong - that one's not started yet
[14:42:43] <cradek> yayyy, it looks right
[14:43:23] <cradek> Get:2 http://buildmaster.local/ lucid/v2.5_branch-rt linuxcnc 1:2.5.3 (tar) [17.8MB]
[14:45:10] <mhaberler> jepler: sledgehammer applied, it's 'resolved'
[14:47:08] <seb_kuzminsky> cradek: yeah, that all looks right :-)
[14:47:33] <cradek> getting it in only 3 tries is not bad at all
[14:48:00] <cradek> (and we won't include the tries those first few days in our count)
[14:49:36] <seb_kuzminsky> you mean we should ignore commits like 161abe630c371c2238a33eac157d825361154cb8? ;-)
[14:49:52] <cradek> hahaha shhhhhh
[14:50:07] <jepler> I wonder if anyone ever considered that git commits should have shorter names
[14:50:12] <jepler> nah forget it
[14:50:28] <cradek> it was March - we were probably cold and our brains were numb
[14:50:30] <seb_kuzminsky> the mercurial guy did, and look where it got him
[14:51:02] <seb_kuzminsky> he's got short integers for commit ids, but they're horribly ambiguous because they're per repo
[14:51:19] <cradek> er no, '12 was the weirdo early summer
[14:51:27] <cradek> oh well I'm sure there was some reason
[14:51:36] <seb_kuzminsky> btw, "mercurial" is a way better name than "git", and "hg" is a sweet name for the tool
[14:55:58] <cradek> I understand hg is smarter than diff/patch, too
[14:56:15] <cradek> it may be the superior solution that nevertheless lost
[14:58:22] <jepler> yeah apparently cp a b; hg add b is different than hg copy a b
[14:58:58] <jepler> I leave it to the peanut gallery to debate whether this should be the case
[14:59:36] <seb_kuzminsky> one shortcoming of git helped introduce a problem in the ja3 branch
[14:59:51] <seb_kuzminsky> a bunch of work was done in ja3 to update .ini files in configs/sim
[15:00:01] <seb_kuzminsky> while in master those files got moved
[15:00:21] <seb_kuzminsky> the merge of master into ja3 caused conflicts and the resolution dropped the modified files from ja3 :-(
[15:00:33] <jepler> it didn't ask you for each file whether to use the removed or the modified file?
[15:00:41] <jepler> (that's what git mergetool says to me in those cases)
[15:00:47] <seb_kuzminsky> i didnt do that merge, micges did
[15:01:14] <cradek> well it's still almost always a mistake to rename or move files in vc :-/
[15:01:18] <seb_kuzminsky> and yes, it does squawk about those kinds of conflicts like you say
[15:01:53] <seb_kuzminsky> in bzr, files have a unique id that remains unchanged when the file is renamed/moved, and merges like that automatically apply the patch to the new location
[15:01:56] <seb_kuzminsky> it's pretty cool
[15:02:10] <seb_kuzminsky> (bzr sucks in other ways tho, so i think over all git is still a better solution for us)
[15:03:42] <seb_kuzminsky> oh, btw please dont fix that in ja3, i've fixed it in ja4 and i'll push it any week now
[15:05:57] <cradek> there was no danger of that
[15:08:16] <jepler> I guess that at some level I like the simplicity of tracking trees and nothing but trees
[15:09:10] <jepler> as opposed to whatever mercurial and bzr do that is supposed to support rename/modify conflicts better
[15:09:55] <seb_kuzminsky> yeah the simplicity of git is hugely appealing
[15:10:01] <seb_kuzminsky> (despite the very un-simple ui...)
[15:10:04] <jepler> git has detection of renames during merges, but many real-world trees have so many files that the feature is turned off. there's a tunable merge.renameLimit the numeric meaning of which is unclear
[15:10:16] <jepler> git's badness is no worse than general unix badness
[15:10:34] <seb_kuzminsky> did you see the git koans that went around a few weeks ago?
[15:10:52] <jepler> yes I think so
[15:13:20] <seb_kuzminsky> precise sim i386 debs are up in release/ and look good
[15:13:40] <seb_kuzminsky> and lucid sim i386
[15:15:35] <seb_kuzminsky> (i haven't tried installing them, but the filenames & sizes look right
[15:16:49] <seb_kuzminsky> and precise sim amd64
[15:17:02] <seb_kuzminsky> ok i'm going to stop cheering for every builder that doesnt croak now
[15:17:34] <cradek> but it was endearing
[15:18:27] <seb_kuzminsky> i wonder why the size of the french docs is only 80% of the english
[15:27:38] <seb_kuzminsky> and why are the rt docs smaller than sim?
[15:36:54] <seb_kuzminsky> finished!
[15:36:59] <seb_kuzminsky> easiest release ever?
[15:56:21] <cradek> woohoo
[17:20:24] <alex_joni> cradek: yay
[21:01:46] <skunkworks> dad bought a acroloc as2240
[21:04:12] <skunkworks> looks like simens servo drives
[21:06:31] <CaptHindsight> there was one for sale up in your neck of the woods a few weeks ago
[21:09:46] <cmorley1> cradek: No problem. Just putting on some meat for our meeting :)
[21:16:19] <cradek> I feel a little weird about it - an agenda item that means "we decide that a particular person should do a thing" - I guess I'd prefer if you'd just poke me and remind me you're waiting for it (and I'll have it done before Saturday, haha)
[21:19:52] <cmorley1> Ok - that's what I'll do next time. I sometimes forget I'm in the 'in ' crowd lol
[21:24:49] <cradek> hah, you know the secret handshake and everything, right?
[21:31:02] <cmorley1> I do !
[21:32:56] <cradek> I may actually get it pushed tonight. it's all built already.
[21:36:19] <cmorley1> Now I feel back for seemingly making a bid deal of it. But I'm glad it being done. Thanks Chris!
[21:38:22] <cradek> not a big deal at all, thanks for the kick in the pants to get started.
[21:43:28] <skunkworks> CaptHindsight: yes - that is probably the one - on Craigs list...
[22:04:01] <pcw_home> Acroloc seems not to be universally loved
[22:14:35] <skunkworks> price was right... :)
[22:18:49] <CaptHindsight> skunkworks: I almost bought that one but I didn't have the time to move it