#emc-devel | Logs for 2010-09-30

[06:54:02] <NTU> mozmck, I'm going to bed right now but I have more info regarding EMC not working with custom make install directory (not /usr) along with backtrace :) http://pastebin.com/k84W6XfC Good night all!
[06:55:24] <NTU> also milltask[24527]: segfault at bf1b1ffc ip b77d3d78 sp bf1b2000 error 6 in libposemath.so.0[b77c0000+16000]
[07:04:36] <NTU> alex_joni I remember sending you backtrace before as well, but now I found out the reason. when you download emc source, configure to /usr, then compile, then install, it works fine. However, when using configuring to /usr, compile, and then make install DESTDIR=${pkgdir} EMC doesn't work. When building arch packages, you have to "install" all files in $pkgdir (in my case its /home/ntu/emc2/pkg) and whatever is in the "pkg" folder when running p
[07:04:36] <NTU> acman (Arch Linux package manager) is installed to the actual system. To see all my EMC build script, go to http://aur.archlinux.org/packages/emc2/emc2/PKGBUILD
[07:05:58] <psha> NTU: seem to be make install bug
[07:06:13] <psha> but deb packages are built same way i think...
[07:07:12] <psha> (cd src; export DESTDIR=`pwd`/../debian/tmp; $(MAKE) $@)
[07:07:18] <psha> from debian/rules.in
[07:08:34] <NTU> well i also did this ;) http://pastebin.com/bw1WANXX
[07:09:06] <psha> hint: grep -R '/home/ntu' in installed dir
[07:09:29] <psha> so if make install corrupts some library paths with DESTDIR you'll find it
[07:10:18] <NTU> installed dir?
[07:10:24] <NTU> pkg?
[07:10:26] <psha> $pkgdir
[07:10:29] <NTU> ah
[07:11:19] <psha> when is your morning by the way? ) i may try to build deb package on debian and check if such errors may be found in debian
[07:13:13] <NTU> I'll probably be back up in 7 or 8 hours.
[07:14:10] <NTU> when building for debian, using checkinstall won't help my situation out though ;) we already know that works
[07:15:18] <NTU> its make install DESTDIR=<whatever debian uses here> || return 1 which makes it fail. and arch doesnt use check install for building packages.
[07:17:22] <psha> wait
[07:17:24] <NTU> oh before I go, make install PREFIX=${pkgdir} for some reason requires root access, which shouldn't happen.
[07:17:33] <psha> it's ok!
[07:17:37] <psha> run it under fakeroot
[07:18:04] <psha> you need root access (or faked one) to set proper permissions/groups etc
[07:18:17] <psha> in debian you always use it
[07:18:26] <psha> i mean fakeroot
[07:18:30] <psha> no real of caus
[07:18:31] <psha> e
[07:18:35] <NTU> fakeroot make install?
[07:18:38] <psha> yes
[07:18:42] <NTU> ok :)
[07:18:44] <psha> try it )
[07:19:08] <psha> debian runs smth like 'fakeroot debian/rules binary'
[07:19:16] <psha> and whole process is wokring under faked root
[07:19:54] <NTU> hmm i thought makepkg is always done under fakeroot anyway.. trying it anyway
[07:21:32] <NTU> fakeroot: nested operation not yet supported :( "This error occurs if an attempt is made to run fakeroot from within a fakeroot environment."
[07:23:44] <psha> so don't needed
[07:24:07] <psha> fakeroot is used probably in every packaging system
[09:09:27] <psha> SWPadnos: micges_work: I've tried to generate type-list from halui structure and it seem not-so-bad solution
[09:09:42] <psha> since patch to halui will be Huge
[09:09:49] <psha> but extremly simple
[09:33:19] <psha> http://psha.org.ru/git?p=psha/emc2.git;a=shortlog;h=refs/heads/halui-ptr
[09:33:26] <psha> git://psha.org.ru/psha/emc2.git
[09:33:28] <psha> branch halui-ptr
[16:39:31] <psha> jepler: i've send patch for subsequent mdi commands to mailing list
[16:39:34] <psha> have you seen it?
[16:40:20] <jepler> psha: I haven't had any time to evaluate it. Thanks for working to address my concerns, though.
[16:41:58] <psha> sad :) any chance for review on this week?
[17:26:18] <jepler> maybe this weekend, not sure yet.
[18:07:29] <psha> i'll send halui patchset to mailing list
[18:07:58] <psha> it would be great if someone'll take a look
[18:30:29] <psha> logger_dev: bookmark
[18:30:29] <psha> Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emcdevel/2010-09-30.txt
[18:42:59] <psha> http://thread.gmane.org/gmane.linux.distributions.emc.devel/3552
[18:43:23] <psha> SWPadnos & micges: I've sent halui patches to mailing list
[18:43:46] <psha> when you'll have a piece of spare time please consider reviewing them
[18:44:10] <andypugh> Have you checked the alternative syntax (with it all on one line)?
[18:44:26] <SWPadnos> how it's connected is irrelevant
[18:44:39] <SWPadnos> or rather, how the connections are made is irrelevant
[18:45:15] <psha> andypugh: multiline syntax is only for clearness of what is connected
[18:46:46] <psha> I've tested it a bit but external review is always needed
[20:27:03] <psha> bb
[21:19:57] <NTU> Hey guys! I'm going about this EMC install issue I have by taking the .deb package posted here: http://www.linuxcnc.org/emc2/dists/lucid/emc2.4/binary-i386/emc2_2.4.4_i386.deb and converting it / using it in arch package, but was wondering what files to edit within the .deb to make it run on something other than the "2.6.32-122-rtai" kernel
[21:38:59] <jepler> NTU: because emc2 uses kernel modules, it only works with the specific version of the kernel it was compiled for.
[21:40:08] <NTU> so even if i get all the files edited correctly, it wouldn't work, because the kernel version is different?
[21:40:55] <NTU> hmm.. maybe if i replace the kernel modules with my kernel modules...
[21:43:45] <cpresser> emc needs to be linked against the running realtime-kernel
[21:43:51] <NTU> right
[21:44:46] <jepler> bbl
[21:44:52] <cpresser> if you are familiar with debian packages and compiling; it should be no big deal to build it all by yourself
[21:45:01] <cpresser> I did so for debian
[21:45:19] <NTU> building it isnt an issue
[21:45:39] <NTU> its getting it to work when make install DESTDIR doesnt =/usr
[21:46:13] <cpresser> "dpkg-buildpackage" worked fine here
[21:46:15] <NTU> but if i compile from hand, it works fine
[21:46:37] <NTU> arch doesnt use dpkg ;)
[21:47:29] <NTU> i have to figure out how to get EMC working on other distros that aren't ubuntu and debian but EMC always seg faults when make install DESTDIR=$pkgdir
[21:47:37] <andypugh> Why not just use a dedicated CNC machine, and put up with Ubuntu?
[21:47:55] <NTU> not going down that road ;)
[21:48:12] <NTU> i rather not bash a distro in front of a bunch of people that use it...
[21:48:44] <cpresser> * cpresser would help basing ubuntu :D
[21:48:45] <andypugh> We use it for our CNC machines. Some might use it for other stuff, but I certainly don't.
[21:49:13] <cpresser> try to do a backtrace and figure out, why it does segfault
[21:49:17] <NTU> i did
[21:49:38] <cpresser> no clues found?
[21:49:39] <NTU> http://pastebin.com/k84W6XfC
[21:49:52] <andypugh> I think you might be making the same error as the Windows/Mach3 folks, thinking that it makes sense to use the CNC controller as a general purpose PC. It makes a lot more sense to use something cheap and effective on the CNC and treat it as an embedded appliance.
[21:50:43] <cpresser> mmh.. hard to say what really happens without debugging symbols
[21:50:55] <NTU> i compiled EMC with -O0 -g3...
[21:51:02] <NTU> so.. :/
[21:51:31] <cpresser> the makefile most likely strips debuggin symbols
[21:51:43] <NTU> ugh.. ok let me edit that
[22:03:23] <NTU> so difficult editing source by hand with makepkg because it decompress tarballs after each pass..
[22:03:41] <NTU> ill get it :)
[22:10:12] <NTU> it works?
[22:10:24] <NTU> ...?? Yay! My PKGBUILD works for once!
[22:11:03] <NTU> no segfault! :D
[22:25:14] <andypugh> For your next challenge, build me a Darwin/OSX version :-)
[22:25:45] <NTU> i dont have any mac software
[22:26:26] <andypugh> A trivial problem, I am sure, to a man of your skills.
[22:33:38] <NTU> next step, arch linux emc live cd!
[22:33:39] <micges> hi seb
[22:38:08] <micges> oh
[22:44:42] <qq-> wget http://sourceforge.net/projects/emc/files/emc2/2.4.x/emc2_2.4.4/emc2_2.4.4.tar.gz
[22:45:55] <NTU> wrong window? :P
[22:46:22] <qq-> no
[22:46:35] <qq-> NTU , was for you ..
[22:47:49] <NTU> Oh. Yes I already have the source downloaded, and working as an arch package, by modifying one of the CFLAGS :D
[22:48:57] <qq-> the right place where to get last emc2 source to compile , usually compile w/o any CFLAGS changes
[22:49:24] <NTU> the CFLAGS changes had to be done, otherwise would segfault
[22:49:45] <qq-> if yes , your distribution is some weird
[22:50:02] <NTU> no its not weird. Its just not Ubuntu or Debian
[22:50:36] <NTU> Anyway, my issue has been solved and already pushed online. :)
[22:51:00] <qq-> emc2 was not 'made' debian/ubu specifics
[22:51:24] <NTU> you dont understand qq-.. I was building a package. not compiling by hand
[22:52:17] <qq-> NTU, sure , its same thing
[22:52:23] <NTU> no it isn't
[22:52:39] <NTU> EMC seg faults when built as a package on Arch if stock CFLAGS are used.
[22:53:19] <qq-> use a real distribution than ..
[22:53:41] <NTU> ...
[22:54:56] <NTU> Go say that in #archlinux :)
[22:55:22] <NTU> This isnt the place for distro arguments.
[22:55:35] <NTU> maybe #defocus might suit you better
[22:58:21] <cpresser> NTU: what CFLAGS worked? which not?
[22:58:41] <qq-> NTU, you assumed emc2's CFLAGS are wrong because don't works on archlinux ?
[22:58:47] <NTU> qq- no..
[22:59:07] <qq-> NTU, that's what you said all long
[22:59:13] <NTU> I didnt assume anything. It actually doesn't work.
[22:59:23] <NTU> when building a package for arch linux
[22:59:56] <NTU> cpresser: DEBUG := -g3 -Wall $(call cc-option,-Wno-write-strings) CFLAGS := $(INCLUDE) -O0 $(DEBUG) -DULAPI $(call cc-option, -Wdeclaration-after-statement) $(BITOPS_DEFINE) CXXFLAGS := $(INCLUDE) -O0 $(DEBUG) -DULAPI $(BITOPS_DEFINE)
[23:00:48] <NTU> that worked in Makefile
[23:00:55] <cpresser> which did not?
[23:01:00] <NTU> stock Makefile settings would cause sincos errors
[23:02:07] <NTU> DEBUG := -g -Wall $(call cc-option,-Wno-write-strings) CFLAGS := $(INCLUDE) $(OPT) $(DEBUG) -DULAPI $(call cc-option, -Wdeclaration-after-statement) $(BITOPS_DEFINE) CXXFLAGS := $(INCLUDE) $(OPT) $(DEBUG) -DULAPI $(BITOPS_DEFINE)
[23:02:50] <cpresser> so it may be a to strong optimation? or what else changed?
[23:03:13] <NTU> apparantly -Os is used in EMC stock build
[23:03:36] <NTU> -Os = optimize for size
[23:05:44] <NTU> -O3 -funroll-loops might fail... i haven't tried it though..
[23:06:07] <cpresser> what gcc are you using?
[23:06:24] <NTU> 4.5.1 (latest stable)
[23:06:58] <NTU> failed on i686 and x86_64 builds.
[23:08:29] <NTU> RTAI also had problems with -Os... it would cause kernel panic on start-up. I had to compile kernel with -O2.
[23:08:39] <cpresser> it may be possible, that newer gcc or non i386 builds produce an error. that may need further investigation
[23:08:49] <cpresser> i am no expert at all :)
[23:08:56] <NTU> heh neither am I
[23:09:06] <NTU> i just try stuff and fiddle with things till it works
[23:12:47] <NTU> brb
[23:30:11] <qq-> hm , gcc 4.5.1 , on 4.4.5 compile ok