**** BEGIN LOGGING AT Sat Mar 04 10:59:57 2006 Mar 04 12:16:31 rwhitby: autobuild of optware is broken for 14 days now. Mar 04 12:17:06 oleo: which target? Mar 04 12:18:16 looking at autobuild-gastro-last.txt and also nudi. Mar 04 12:18:24 If a package is not compiling on a certain target, then add it to _BROKEN_PACKAGES in the Makefile and commit that change. Mar 04 12:19:00 the autobuilds for optware are fully under the control of the optware developers through the variables in the Makefile in CVS. Mar 04 12:19:16 I know! cvs.sourceforge.net: Connection timed out Mar 04 12:20:05 ok, so it looks like Gastro autobuilds is broken. Mar 04 12:21:50 But optware developers can fix optware on nudi. Mar 04 12:23:19 Yes. And I dont know why NUDI also does not work? Python should compile. But now with libuclibc++ /opt is also patched. Maybe there is a need for manual install ov libuclibc++-stage on build machine. Mar 04 12:24:13 surely the first package that depends on it will stage it. Mar 04 12:25:51 Hmm - dyoung will need to update the autobuild on gastro to push to limax Mar 04 12:27:18 Please please. Can you reslove this ASAP. Bzhou solved ldso problem on wl500g and now python should work and also all dependent packages. so there should not be any build problem when new py-package is added. Mar 04 12:27:45 oleo: if I knew how to resolve it, I would. Mar 04 12:28:37 So who has access to autobuild if not you? Mar 04 12:29:44 I have access to autobuild. You have not told me explicitly what needs to be done. You have said "maybe this needs to be done". If there is a problem which is not solvable by optware developers using the variables in the Makefile, then tell me what needs to be done on the build machine to fix that problem. Mar 04 12:30:59 Can you explain the differences in gastro, nudi? limax? Mar 04 12:31:00 (e.g. if you say "the build directory for package for wl500g on nudi needs to be deleted to force a rebuild", or something like that, then i can easily do that for you. Mar 04 12:31:10 nudi is the cross-compile build machine Mar 04 12:31:19 gastro is the native-compile build machine Mar 04 12:31:41 limax is the place where the built packages end up Mar 04 12:31:42 For sourceforge connection I would first test connectivity, then maybe SSH keys are broken Mar 04 12:32:26 Gastro autobuild is far more broken than just the ssh keys. dyoung will have to work on it. Are there native compiles that are holding things up? Mar 04 12:32:45 I thought you were talking about Python on nudi ... ? Mar 04 12:33:59 no. No native. But then nudi is the wlace to work with. Is there a command to push autobuild process on nudi? Is cron once a day? Mar 04 12:34:15 autobuild runs every 15 minutes. Mar 04 12:35:17 So if I remove all python staging reference and python itself, other packages will proceed? Mar 04 12:36:08 Looking at the logs, it seems that Python is broken on wl500g. You should simply add python to WL500G_BROKEN_PACKAGES and contact the maintainer of that package to get it fixed. Mar 04 12:36:20 OK will test this imediatelly? And see in half an hour. Mar 04 12:38:46 No this is not enough. We must mark broken all python dependent packages as they will stage it. Python is already marked broken! Mar 04 12:39:32 have fun! Mar 04 12:39:34 * rwhitby goes to bed Mar 04 12:39:59 OK. Thanks for info. Mar 04 12:42:26 rwhitby: But is /opt writable on NUDI? Mar 04 13:12:08 03oleon * 10unslung/Makefile: optware: Demote all python dependent packages on wl500g - Cleanup will follow when python is fuly functional Mar 04 16:07:43 hello, I have a short question to the OpenSlug build system Mar 04 16:07:52 I need to compile a kernel module for 2.7-beta Mar 04 16:08:32 If I set up the build system with 'make openslug-image' I get the most recent kernel tree, right? Mar 04 17:19:47 eno: How can python build have separate LDFLAGS for buildpython and crosscompiled stage? Is there a option in configure? Mar 04 17:20:18 i'll need to look Mar 04 17:38:48 oleo, we probably can change sources/python/configure.in.patch Mar 04 17:52:21 OK. Can you do this? different environment variables like PYTHON_BUILD_LDFLAGS and PYTHON_LDFLAGS Mar 04 17:58:43 yeah, testing that right now, one moment Mar 04 18:10:29 oleo, have to b away for a bit Mar 04 20:05:18 03bzhou * 10unslung/ (make/python.mk sources/python/configure.in.patch): HOSTPYTHON now have separate CPPFLAGS/LDFLAGS, added wl500g uclibc workaround Mar 04 20:27:22 eno, python now ends up with "/bin/sh: ./buildpython/python: No such file or directory" Mar 04 20:30:42 build time or run time? Mar 04 20:31:18 build time. Mar 04 20:32:06 any error when building buildpython/python ? Mar 04 20:35:12 i tested optware/nslu2 and optware/wl500g, both seem to build fine, no error in my build log Mar 04 20:35:26 I think that instead CC="" CXX="" AR="" RANLIB="" $absconfigcommand --prefix=$prefix you should have LDFLAGS=$PYTHON_HOSTPYTHON_LDFLAGS CC="" CXX="" AR="" RANLIB="" $absconfigcommand --prefix=$prefix Mar 04 20:37:40 that's for the CROSS_COMPILE=yes part, i don't think we want to introduce PYTHON_HOSTPYTHON_* in there Mar 04 20:39:22 I have just followed "cd buildpython" line. Mar 04 20:40:21 ok, i think you have a point Mar 04 20:42:58 And also builded python. :) Changet that line into CC="" CXX="" AR="" RANLIB="" LDFLAGS="$PYTHON_HOSTPYTHON_LDFLAGS" $absconfigcommand --prefix=$prefix and now it works. Will test the ipk on my machine. Znd maybe add -lz for zlib linking. Mar 04 20:44:42 And bump ipk version. Mar 04 20:46:02 although i still don't fully understand - after "Configuring build python" line, shouldn't $build and $host be equal? Mar 04 20:46:32 both should be i686-pc-linux-gnu Mar 04 20:47:12 thus, it should not take else (CROSS_COMPILE=no) branch Mar 04 20:47:47 it should take the else (CROSS_COMPILE=no) branch (disregard the previous line) Mar 04 20:49:19 I tought that you understand this? I am takinh naive approach, tho get the beast working. And of course ith bothers me because buildpython is geting built over and over in every stage. This should be a separate stage as I mentioned before. Mar 04 20:49:20 also, PYTHON_HOSTPYTHON_* is make variables, should not be visible to the configure scripts Mar 04 20:50:07 have you taken my corresponding python.mk change? Mar 04 20:51:31 yes. cvs -u up -dP Mar 04 20:51:47 s/-u/-q/ Mar 04 20:51:48 oleo meant: yes. cvs -q up -dP Mar 04 20:52:47 it does not bother me that much to build buildpython everytime python is built Mar 04 20:53:25 i'm not the original author of python cross compilation, to separate the buildpython requires more changes than i'm willing to take Mar 04 20:54:18 that's not something we need to address right now Mar 04 20:56:25 shure. Whath bothers me is that http://logs.nslu2-linux.org/buildlogs/autobuild-nudi-last.txt shows it is stuck. It looks like CVS update is not working. Mar 04 21:03:27 test_curses.py works, test_crypt.py works, test_gzip.py segfaults as there is no -lz, readline works of course. Is there a test for db? Mar 04 21:05:15 test_dbm.py Segmentation fault. But there is a way. Mar 04 21:10:25 "/home/slug/optware/wl500g/staging/opt/lib/libpython2.4.so: could not read symbols: Invalid operation" Mar 04 21:12:22 test_sundry.py works. I have removed all python references from Makefile for wl500g. It look like monotone is just not updating. Will ask rwithby for help. Again. Mar 04 21:13:19 sometimes monotone update takes quite a long while to finish Mar 04 21:13:54 8 hours? Mar 04 21:14:06 not that long Mar 04 21:14:38 It is stuck for 14 days. I think. Mar 04 21:16:04 If you look http://ipkg.nslu2-linux.org/feeds/unslung/wl500g/ Packages date. Mar 04 21:17:23 nudi-last.txt is definitely not using my latest check-in Mar 04 21:18:12 yeah, the builds are broken Mar 04 21:21:13 so we have two problems: 1) nudi autobuild broken 2) with my latest configure.in.patch, your ./buildpython/python build time failure Mar 04 21:23:03 2) is solved solved by me now and waiting CI after zlib abd db regression tests. Mar 04 21:24:09 For 1) rwhitby in the morning. Maybe. Mar 04 21:25:40 And before checkin I want to solve this ipkg: Cannot create symlink from ./usr/bin/python to '/opt/bin/python': Read-only file system Mar 04 21:25:54 rwhitby: Any idea on the dependency-problem (nothing provides libc6) that breaks the builds on nudi? Mar 04 21:26:15 Is it time to clean out the whole dir and start over? Mar 04 21:26:23 that line is problematic for wl500g Mar 04 21:27:01 but we want the symlink for other platforms Mar 04 21:27:43 oleo: there is no way that an autobuild should be writing to /opt on the build system. Mar 04 21:27:56 Why this link is needed? I saw that rwhitby set this on nslu2. But we do not have /usr/bin/ writable. Just /usr/local/bin. Mar 04 21:28:43 oleo: cause some scripts assume /usr/bin/python, just like /bin/sh or /usr/bin/perl. Mar 04 21:28:59 just do a conditional in make/python.mk around line 247 Mar 04 21:30:16 OK for that link. But rwhitby, libuclibc++ must be installed over existing toolchain. I do not see any other simple way. Mar 04 21:30:46 oleo: by existing toolchain what exactly do you mean? Mar 04 21:31:36 03bzhou * 10unslung/make/python.mk: /usr/bin is read-only on wl500g Mar 04 21:32:36 seems there was a problem with monotone update on nudi due to the OE problems in the last week. I have now cleared that. Mar 04 21:32:52 I mean oleg toolchain /opt/brcm/hndtools-mipsel-uclibc-3.2.3 is updated with C++ libraries. and g++ is then wrapped. Mar 04 21:33:22 oleo: ok, someone has to provide a tarball to install replacing the existing installation. Mar 04 21:33:48 (i.e. a complete replacement, not an addition, cause we need to keep the tarball for disaster recovery) Mar 04 21:34:03 (and also so others can download the tarball to replicate the build environment) Mar 04 21:34:10 It is just to enter make libuclibc++-stage or make libuclibc++-install Mar 04 21:34:42 um, foo-stage should never write outside of the staging dir Mar 04 21:35:08 but if /opt/brcm/hndtools-mipsel-uclibc is writable is enough. Mar 04 21:37:26 This stage should be there for the case of uClibc++ upgrade. Mar 04 21:38:11 oleo, is there a way to have the wrapper be under optware/wl500g/toolchain ? Mar 04 21:38:12 why isn't uClibC++ staged into staging, and then found in there with a -L switch ? Mar 04 21:38:13 I can add writability check if you suggest so? Mar 04 21:41:23 This was my first try. But all configure and autoconf tools assumes hard paths. At start I've had mipsel-uclibc-g++-uc compiler wraper. But then most of the packages failed to find it as there was g++. "Use the force luke!" was not enough. Mar 04 21:44:25 Or we can just say: "Before using Oleg toolchain, One must issue make libuclibc++-install". Oleg says "make uclibc" must be issued before any other. Mar 04 21:47:15 oleo: here's the thing. We are *not* going to give write permission to /opt/... to the build account. That would be a *huge* security hole, let alone a maintenance and recovery management nightmare. Mar 04 21:47:40 The toolchain in /opt comes from a tarball. If it's to be updated, it's gonna be by another tarball. Mar 04 21:48:18 To put it simply, a .mk file is simply not allowed to write outside of it's build and staging areas. Mar 04 21:51:37 So let me know when you have a replacement tarball for /opt/brcm/hndtools-*, and I'll be happy to install that. We can even put it up on nslu.sf.net/downloads so other people can download it too. Mar 04 21:52:11 I understand that security hole. I will remove this staging and then will ask you to issue make libuclibc++ install. Mar 04 21:53:16 s/++ i/++-install/ Mar 04 21:53:17 oleo meant: I understand that security hole. I will remove this staging and then will ask you to issue make libuclibc++-installnstall. Mar 04 21:53:36 oleo, maybe you can provide a tarball containing all the touched files of 'libuclibc++' Mar 04 21:54:10 Yes no problem on that. Mar 04 21:54:19 s/libuclibc++/libuclibc++ install/ Mar 04 21:54:20 eno meant: oleo, maybe you can provide a tarball containing all the touched files of 'libuclibc++ install' Mar 04 21:55:16 stupid bot. Mar 04 21:58:06 oleo, does the following look reasonable (just started another make python-ipk)? Mar 04 21:58:13 # workaround for uclibc bug, see http://www.geocities.com/robm351/uclibc/index-8.html?20063#sec:ldso-python Mar 04 21:58:13 ifeq ($(OPTWARE_TARGET),wl500g) Mar 04 21:58:13 -PYTHON_LDFLAGS=-lncurses -lreadline -lcrypt -lssl Mar 04 21:58:13 +PYTHON_LDFLAGS=-lcrypt -ldb-$(LIBDB_LIB_VERSION) -lncurses -lreadline -lssl -lz Mar 04 21:58:13 else Mar 04 22:01:23 please provide a tarball with the complete /opt/brcm/hndtools-... directory. We want to have a single tarball that we (and other people) can install. "make libuclibc++" is not a means to configuration manage the /opt directory. Mar 04 22:02:03 (as the results depend on when you type that command, if the .mk file changes, or upstream changes) Mar 04 22:05:01 03oleon * 10unslung/ (Makefile make/libuclibc++.mk): libuclibc++: toolchain on wl500g requires manual install with 'make libuclibc++-install' or separate toolchain tarball Mar 04 22:06:45 03bzhou * 10unslung/make/python.mk: exclude /usr/bin for wl500g ipk Mar 04 22:08:17 oleo: is Oleg going to provide uclibc++ in the next release of his toolchain? Mar 04 22:11:26 C++ programs are not of his interest, I presume. So there is not a big interest. I've begged him to upgrade uclibc to 0.9.28, without success. Mar 04 22:14:09 03oleon * 10unslung/sources/python/configure.in.patch: python: add PYTHON_HOSTPYTHON_LDFLAGS LDFLAGS to buildpython stage Mar 04 22:14:28 I have a short question to http://www.nslu2-linux.org/wiki/DebianSlug/Bootstrap: touch /.recovery doesn't work, it boots into debian Mar 04 22:14:55 * NAiL whistles Mar 04 22:14:56 at least I don't see any difference to Debian Mar 04 22:15:05 no, .recovery doesn't work Mar 04 22:15:13 just unplug the disk Mar 04 22:15:26 should I correct in Wiki? Mar 04 22:15:35 yeah Mar 04 22:17:32 NAiL: the rest is correct? ;) Mar 04 22:17:55 I do believe so, lemme check Mar 04 22:18:12 Of course I made a backup, but it's more comfortable if I don't loose data :) Mar 04 22:18:52 well, "rm /mnt/debinst/.recovery" isn't required anymore then Mar 04 22:19:31 yes, of course, my mistake Mar 04 22:20:11 np ;) Mar 04 22:22:15 NAiL: Maybe I should add that login account openslug to the Wiki? Mar 04 22:22:36 hmm? Mar 04 22:22:59 root Mar 04 22:23:00 of course Mar 04 22:23:05 opeNSLUg passwor Mar 04 22:23:09 ah, yeah Mar 04 22:25:02 rwhitby: how should this toolchain be called? Original name is hndtools-mipsel-uclibc-3.2.3-full.tar.bz2 at http://wl500g.dyndns.org/uClibc-0.9.19 but I dont know if this is the latest as I am building this toolchan with make uclibc which patches /opt/brcm :) Mar 04 22:29:23 NAiL: could it be that the 'turnup ram' is necessary after the reboot, too? Mar 04 22:29:40 at least I got an error Mar 04 22:30:24 bwalle: Yeah, the turnup ram should be done Mar 04 22:31:10 then the reboot is not needed, right? Mar 04 22:31:25 to reflash, no Mar 04 22:31:55 NAiL: halt; unplug; turn on nslu; turnup ram; plug disk; mount; reflash Mar 04 22:32:01 right? Mar 04 22:33:04 turnup ram; reboot; mount; reflash Mar 04 22:33:22 that's in the wiki but that doesn't work Mar 04 22:33:27 hmm Mar 04 22:33:33 where does it error out? Mar 04 22:33:49 It didn't boot from this, so the 'turnup ram' was successful. The disk was plugged in. Mar 04 22:33:55 root@villabacho:/# reflash -i mnt/debinst/home/debianslug-nslu2-20060304193441.flashdisk.img Mar 04 22:33:55 /sbin/reflash: /dev/mtdblock4 is mounted on /, use turnup ram Mar 04 22:34:05 hmm Mar 04 22:34:12 gotta go, can look at it later Mar 04 22:35:07 oleo: give it the version number from the uclibc you are building from, and add -uclibc++ to the end instead of -full. Mar 04 22:37:29 OK. Will update wiki with link. And then you should install this. Right now I am compiling latests firmware and then will patch toolchainm just to verify it is OK. Mar 04 22:38:16 ok Mar 04 22:38:37 ping me when it is ready to be installed on nudi Mar 04 22:40:00 uclibc++, neat **** BEGIN LOGGING AT Sat Mar 04 22:47:34 2006 Mar 04 22:48:18 oleo, i don't think your last sources/python/configure.in.patch should make any difference Mar 04 22:50:52 just do a before/after comparison of builds/python/buildpython/Makefile **** BEGIN LOGGING AT Sat Mar 04 23:04:08 2006 **** BEGIN LOGGING AT Sat Mar 04 23:09:36 2006 Mar 04 23:22:03 03bzhou * 10unslung/ (make/python.mk sources/python/configure.in.patch): removed PYTHON_HOSTPYTHON_*, simply setting LDFLAGS="" before buildpython configure Mar 04 23:29:11 NAiL: after another reboot, it worked. Don't ask me why. Mar 04 23:29:18 The new version runs fine :) Mar 04 23:29:29 good :) Mar 04 23:29:33 Thanks for your help! Mar 04 23:29:48 np Mar 05 00:41:40 join nslu2-general Mar 05 01:15:41 eno-away : python test_zlib.py works, test_crypt.py works, test_gzip.py works, test_tarfile.py segraults on missing bz2 library. test_sundry.py works, test_userstring.py works. test_zipfile.py works, test_zipimport.py works, test_zlib.py works. Python is thus missing libbz2-stage and -lbz2. All regression tests should now work on wl500g. Mar 05 01:19:54 ok, let me add bz2 Mar 05 01:20:00 rwhitby, I have updated http://www.nslu2-linux.org/wiki/FAQ/OptwareWl500gBuild and now I am waiting for nudi to be upgraded. I have tested install of the toolchain as described and I can say that there is no problem with it. Mar 05 01:20:13 and i'll bump IPK_VERSION Mar 05 01:21:28 yeah. Q.E.D. Mar 05 01:26:17 oleo, python.mk has --enable-unicode=ucs4, do you want to explicitly disable unicode for wl500g? Mar 05 01:35:39 03bzhou * 10unslung/make/python.mk: added the missing bzip2-stage, bumped IPK_VERSION Mar 05 01:38:51 eno-away : there is no wchar support in wl500g. So disabling ucs4 seems reasonable. But then, maybe Unicode is handled internally and not relyiny on wchar support. Is there any unicode test out there? Mar 05 01:41:01 /opt/lib/python2.4/test/test_unicode.py ? Mar 05 01:44:22 OK then. But python.mk should have bzip2 depend then. Or not. How about moving some libs into PYTHON_SUGGESTS as described in template.mk Mar 05 01:49:45 python test_unicode.py OK, test_unicode_file.py No Unicode filesystem semantics on this platform. As suspected. test_unicodedata.py OK. I suggest that unicode ramains supported. Mar 05 01:50:27 PYTHON_DEPENDS=libstdc++, readline, ncurses Mar 05 01:50:27 PYTHON_SUGGESTS=bzip2, openssl, libdb, zlib Mar 05 01:50:54 although debian python2.4 package has everything as dependencies Mar 05 01:51:39 oleo: downloading now Mar 05 01:52:13 let's follow debian, both nslu2 and wl500g not lacking disk space Mar 05 01:52:36 Yes, you are right. It python is so big pavkage that this little libs means nothing. But I always wonder why geting the monkey if you wanted just bananna. Mar 05 01:55:01 03bzhou * 10unslung/make/python.mk: added bzip2 to DEPENDS, bumped IPK_VERSION Mar 05 02:00:32 you have to admire debian, they have python2.4-minimal depends on nothing but libc6 and zlib1g Mar 05 02:01:41 OTOH, python has this "batteries included" philosophy Mar 05 02:07:25 oleo: installing now Mar 05 02:09:32 oleo: next optware autobuild will use the new toolchain Mar 05 02:09:56 Its 3AM here, but I'll wait for breakout. Mar 05 02:10:57 But how about this CVS update. It seems like not working at all. Mar 05 02:11:19 on nudi? Mar 05 02:13:11 yes. Today I've removed all python references but then realised no fresh copy is coming onto nudi. See http://ipkg.nslu2-linux.org/feeds/unslung/wl500g/ Packages is 10 days old. Mar 05 02:15:40 that was due to the monotone problem causing make update to fail before cvs was called Mar 05 02:17:24 OK then. A little snack before bed, and fingers crossed. Mar 05 02:17:54 autobuilds re-enabled Mar 05 02:18:05 make update works Mar 05 02:18:17 * rwhitby goes out Mar 05 02:44:01 rwhitby , from broken log of wl500g I see that libuclibc++.mk was not the latest. Also python is now 2.4.2-6 but log states 2.4.2-4 was used! Mar 05 04:01:09 jbot, test Mar 05 04:01:11 Test Failed! Mar 05 07:52:05 03bzhou * 10unslung/make/ruby.mk: mv sitedir to /opt/local/lib/ruby/site_ruby Mar 05 08:13:22 oleo: py-mx-base failed to build Mar 05 08:15:02 /opt/brcm/hndtools-mipsel-uclibc-3.2.3-full/include/wchar.h:32:2: #error Attempted to include wchar.h when uClibc built without wide char support. Mar 05 08:15:03 mx/DateTime/mxDateTime/mxDateTime.c:3187: warning: static declaration for `mxDateTimeDelta_FromDaysEx' follows non-static Mar 05 08:15:03 mx/DateTime/mxDateTime/mxDateTime.c:3205: warning: static declaration for `mxDateTimeDelta_FromSeconds' follows non-static Mar 05 08:15:03 error: command '/opt/brcm/hndtools-mipsel-uclibc/bin/mipsel-uclibc-gcc' failed with exit status 1 Mar 05 08:15:04 make[1]: *** [/home/slug/optware/wl500g/builds/py-mx-base/.built] Error 1 Mar 05 08:15:06 make[1]: Leaving directory `/home/slug/optware/wl500g' Mar 05 08:15:08 make: *** [optware-wl500g] Error 2 Mar 05 08:17:54 rwhitby: the 10 hour thing i was talking about last week has not happened again :) guess openntpd (not reinstalled) is to blame Mar 05 08:20:05 good Mar 05 08:45:59 out of curiosity, is there any way to find changelogs for recent modifications? wondering what has changes in the kernel (and related things) in the last week Mar 05 09:14:36 pepijn: cia.navi.cx - openembedded and nslu2-linux projects Mar 05 09:14:39 rwhitby : Just looking at this wchar problem which is not libuclibc++ related, but rather uClibc. Mar 05 09:17:44 found that link, but find the info there quite limited :( Mar 05 09:18:19 [09:21] *** now talking in #openslug Mar 05 09:18:20 [09:21] *** topic is Developer (not end-user) discussion about the OpenSlug custom firmware for the Linksys NSLU2 device. Mar 05 09:18:20 [09:21] *** set by ChanServ on Mon Aug 08 08:56:40 2005 Mar 05 09:18:20 [09:21] *** channel #openslug mode is +tn Mar 05 09:18:20 [09:21] *** channel created at Sun Feb 27 05:55:01 2005 Mar 05 09:18:21 [09:21] *** now talking in #openslug Mar 05 09:18:23 [09:21] *** topic is Developer (not end-user) discussion about the OpenSlug custom firmware for the Linksys NSLU2 device. Mar 05 09:18:26 [09:21] *** set by ChanServ on Mon Aug 08 08:56:40 2005 Mar 05 09:18:28 [09:21] *** channel #openslug mode is +tn Mar 05 09:18:30 [09:21] *** channel created at Sun Feb 27 05:55:01 2005 Mar 05 09:18:33 oh ffs. Mar 05 09:18:44 Quick question - can I use the openslug devkit (i.e. bitbake) to cross-compile for opendebianslug (aka debian/armeb)? Mar 05 09:19:27 dunno. kernel modules work, dunno about libc dependencies or other things for userland apps. Mar 05 09:20:03 there's some floating point difference which jacques knows about Mar 05 09:20:09 i'm trying to port the latest eciadsl snapshot to my slug - atm it's running the 'debianised' version of 0.10, which is insanely buggy. Mar 05 09:21:01 it uses the gnu auto* stuff, so it's easy to port, but my only two linux boxes are running fedora and i want to cross-compile (saves me installing gcc on the slug) Mar 05 09:22:17 in which case, is there a "debianopenslug crosscompile howto" kicking around anywhere? Mar 05 09:30:30 philpem: lots of examples in openembedded Mar 05 10:11:07 03oleon * 10unslung/ (Makefile make/py-mx-base.mk): py-mx-base: add missing python depend Mar 05 10:14:18 ok, this [http://www.nslu2-linux.org/wiki/DebianSlug/CrossCompiling] looks interesting, but i can't see how you'd go about doing that on a fedora system... Mar 05 10:15:01 i've got vmware, so its not a big deal to create a debian vm and use that, but i'd rather compile natively on fedora if possible. **** ENDING LOGGING AT Sun Mar 05 10:59:56 2006