**** BEGIN LOGGING AT Fri Jul 29 23:59:56 2005 Jul 30 00:02:28 i took the route suggested by the "RS232 connection with a kit Jul 30 00:02:34 section Jul 30 00:04:08 So theres really a MAX3232 in there? Jul 30 00:04:34 well it's totally encased in plastic, but the voltage levels are definetly correct (3.3V on the slug side) Jul 30 00:05:26 hmm Jul 30 00:06:22 i can cut one apart if we really need it, as I bought 5 of the siemens cables on a good deal Jul 30 00:06:26 have you tried rebooting with the cable attached? Jul 30 00:06:49 yeah, i've tried rebooting, no joy. no + or anything actually Jul 30 00:17:18 ok, i'm working on taking one of the cables apart now Jul 30 00:17:35 it's a major task, as it's basically a small plastic enclosure, cast into more plastic Jul 30 00:20:42 If youre sure about your wiring, I dunno what else it could be... Jul 30 00:21:38 I used a MAX232 in a pinch, and it didnt *really* work, but it didnt exhibit the symptoms you describe either. Jul 30 00:49:59 dyoung: does master make really take as long as it's taking Jul 30 00:50:25 i ran out of diskspace on vmware and had a nice time migrating to another virtual disk that is now 40gb instead of 4gb ;) Jul 30 01:47:44 Is it actually compiling now? Jul 30 01:48:07 I think a "from scratch" compile for me takes several hours. Jul 30 01:58:50 nudi does a from-scratch compile of unslung and openslug in 7 hours. Jul 30 02:12:08 03rwhitby * 10unslung/Makefile: Fixed a type on the packages target. Jul 30 02:47:19 03marceln * 10unslung/make/dovecot.mk: Jul 30 02:47:19 First version that compiles and makes a ipkg. Jul 30 02:47:19 TODO: Jul 30 02:47:19 - Cross compilation (configure error) Jul 30 02:47:19 - Start up scripts (inetd or daemonized?) Jul 30 02:47:20 - Testing Jul 30 02:49:52 03rwhitby 07org.nslu2-linux.dev * r74821799... 10/Makefile: Added a naive autobuild target Jul 30 03:09:23 03dyoung 07org.nslu2-linux.dev * rce6f9ade... 10/Makefile: change names from sync_foo to sync-foo Jul 30 03:48:37 I'm lost, is there a step by step guide somewhere on how to add a new kernel-module package? from enabling it in the kernel build to making sure there will be an ipk package after the build Jul 30 03:50:52 thats pretty much exactly what you need to do Jul 30 03:58:19 just enabling them in the config doesn't seem to be enough.. or I may be in the wrong config Jul 30 03:58:48 which "the config" are you talking about? Jul 30 04:00:03 nslu2-linux/packages/linux/openslug-kernel-2.6.12.2/defconfig Jul 30 04:03:15 that should do it Jul 30 04:10:28 well, I have a CONFIG_USB_HID=m line in that file, but searching for *hid* in the openslug source tree doesn't return anything other than .h and .c files Jul 30 04:11:10 did openslug-kernel rebuild? Jul 30 04:11:27 if it did, look in tmp/deploy/ipk for your kernel-module-foo file Jul 30 04:11:38 I added those lines to the clean source tree Jul 30 04:12:18 nope, no ipks there Jul 30 04:12:35 but I don't think it even built the module Jul 30 04:13:27 could there be a dependancy issue in the defconf file? I just added those lines by manually editing it Jul 30 04:13:44 if the kernel needs something else for those modules it probably fails? Jul 30 04:19:17 you may want to sanity check that defconfig then. Jul 30 04:19:19 um... Jul 30 04:19:34 make oldconfig or somethi nglike that Jul 30 04:19:47 from within tmp/work/openslug-kernel-foo Jul 30 04:21:42 k, I'll try that Jul 30 04:35:16 03rwhitby 07org.nslu2-linux.dev * rf32d41db... 10/ (2 files in 2 dirs): Removed the CVS_TARBALL_STASH from local.conf.template, cause it is now in nslu2.conf Jul 30 04:35:19 03rwhitby 07org.openembedded.nslu2-linux * r2bd677fe... 10/conf/machine/nslu2.conf: Added sources.nslu2-linux.org to nslu2.conf for CVS_TARBALL_STASH and MIRRORS_append. Jul 30 04:44:59 03rwhitby 07org.openembedded.nslu2-linux * r0850de9c... 10/ (5 files in 5 dirs): Jul 30 04:44:59 propagate from branch 'org.openembedded.dev' (head 0f025031c9581db625c09129982ac1a87eca3484) Jul 30 04:44:59 to branch 'org.openembedded.nslu2-linux' (head 2bd677fe4fe082554bae63ed747af0b3533bc7a7) Jul 30 04:50:06 03rwhitby 07org.openembedded.nslu2-linux * r009ad2eb... 10/ (classes/nslu2-mirrors.bbclass conf/machine/nslu2.conf): Changed nslu2 mirrors to use INHERIT and a .bbclass file instead. Jul 30 04:51:05 g'day DaKa2 Jul 30 04:51:24 :) Jul 30 04:51:51 dyoung: repo is now updated to use sources.nslu2-linux.org as the mirror of last resort and cvs stash tarball of choice Jul 30 04:53:50 hmm, make oldconfig doesn't recognise the ixp4xx so it becomes crap Jul 30 04:53:53 ;) Jul 30 04:54:39 Matta: what I usually do is diff the defconfig, and the .config file that results from removing and then rebuilding the openslug-kernel bb Jul 30 05:09:30 k, did that, recompiling Jul 30 05:11:05 you'll notice some "=n"'s in there which are required and which get changed to "is not set"'s. Jul 30 05:20:55 hmm, even though I've enabled stuff in defconfig, the tmp/work/.../linux-2.6.12.2/.config file has them as not set Jul 30 05:23:43 Matta: then you need to read the Kconfig files to see what is turning it off Jul 30 05:23:49 probably depend on something... Jul 30 05:23:50 what is it again? Jul 30 05:24:01 USB_HID Jul 30 05:25:04 CONFIG_USB_HID=m Jul 30 05:25:13 that's in the latest openslug kernel .config Jul 30 05:26:08 And /home/slug/openslug/tmp/deploy/ipk/kernel-module-usbhid_2.6.12.2-r4_nslu2.ipk gets created. Jul 30 05:26:34 so what's the problem? Jul 30 05:28:07 they don't work, adding the device results in symbols not found Jul 30 05:28:19 and I'm making a uClibc version... Jul 30 05:28:27 CONFIG_INPUT=m Jul 30 05:28:27 CONFIG_INPUT_EVDEV=m Jul 30 05:28:27 CONFIG_HW_RANDOM=m Jul 30 05:28:27 CONFIG_USB_HID=m Jul 30 05:28:27 CONFIG_USB_HIDINPUT=y Jul 30 05:28:28 CONFIG_USB_KBD=m Jul 30 05:28:36 that's my config I'm trying to create Jul 30 05:30:39 the interesting thing is that none of the above are in the generated kernel / .config Jul 30 05:30:59 so I'm having a hard time believing they all have dependancy issues Jul 30 05:32:12 USB_HID only depends on USB, which I'm quite sure is enabled ;) Jul 30 05:32:59 are you sure that ucblibc builds the 2.6.12.2 kernel and uses that defconfig - check the bitbake stderr output to make sure Jul 30 05:33:15 k Jul 30 05:38:15 yup, at one point it installs tmp/work/openslug-kernel-2.6.12.2-r3/defconfig as the .config file... the latter contains my modifications Jul 30 05:38:54 but I also tried to add them manually to the config file before the compile, bu do_compile apparently reverted back to the old .config Jul 30 05:42:00 Matta: I get all of those enabled in my .config except CONFIG_HW_RANDOM Jul 30 05:42:11 interesting Jul 30 05:42:33 which Kconfig does HW_RANDOM come from? Jul 30 05:43:09 drivers/char/Kconfig Jul 30 05:43:33 ah, it requires X86 or IA64 Jul 30 05:45:10 Matta: I'll check in my defconfig changes Jul 30 05:45:28 ok Jul 30 05:45:59 does glibc/uClibc affect the kernel & modules? Jul 30 05:46:13 with the new autobuild running, the glibc modules should hit the feeds within the hour Jul 30 05:46:19 dunno Jul 30 05:46:27 why are you using uclibc anyway? Jul 30 05:46:55 I have almost 2 Mb free space on the flash now, instead of 500k using glibc Jul 30 05:47:12 I'm trying to run this thing w/o external storage Jul 30 05:47:30 are the rest of your apps going to fit in that? Jul 30 05:47:55 2 Mb should be no problem, 500k would be tricky since I'd like to have libssl available Jul 30 05:48:43 ok, just realise that only jbowler can currently help you with uclibc openslug issues, cause I'm not aware of anyone else running ucslugc Jul 30 05:49:43 k, I'll have to reconsider whether I want uclibc or not Jul 30 05:49:57 so the new kernel mods will enter the feeds? Jul 30 05:50:28 03rwhitby 07org.openembedded.nslu2-linux * r3471c6c7... 10/packages/linux/ (nslu2-kernel_2.6.12.2.bb openslug-kernel-2.6.12.2/defconfig): Added INPUT_EVDEV and USB_KBD modules to OpenSlug kernel Jul 30 05:50:42 k =) Jul 30 05:51:22 will try again tomorrow... Jul 30 05:51:32 thanks for your help! l8r Jul 30 05:51:47 the feeds are updated automatically now, and the new kernel and modules will be there in about 15 minutes or less Jul 30 05:51:56 oh, nice Jul 30 05:53:18 well, unfotunately I have to go anyway but I'll see if I can check it out later tonight Jul 30 05:53:21 cya Jul 30 06:07:34 Feeds are updated. Jul 30 06:08:02 17 minutes from checkin to feeds updated. Jul 30 06:08:19 all automatic :-) Jul 30 06:08:57 brilliant Jul 30 06:08:59 great work :) Jul 30 06:09:04 ~hail rwhitby Jul 30 06:09:04 * jbot bows down to rwhitby and chants, "I'M NOT WORTHY!!" Jul 30 06:09:42 ~hail dyoung Jul 30 06:09:43 * jbot bows down to dyoung and chants, "I'M NOT WORTHY!!" Jul 30 06:20:31 ok smart people: I need a way of determining whether "make autobuild" is still running, and the means of detecting that cannot leave lock files around if the make has an error. Jul 30 06:21:00 I'm thinking something like "ps -ef | grep autobuild | grep -v grep" in an if statement Jul 30 06:21:10 but can't get the syntax right Jul 30 06:21:20 hmm? Jul 30 06:21:38 if ps -ef | grep "make autobuild" >& /dev/null ; then \ Jul 30 06:21:39 ${MAKE} update build upload ; \ Jul 30 06:21:39 fi Jul 30 06:21:46 something like that (but that one doesn't work) Jul 30 06:21:57 no, because that one will see the grep Jul 30 06:22:25 the if statement has to not execute the body if it sees autobuild in the process list Jul 30 06:23:07 NAiL: I can get the grep right - it's the if statement syntax and grep exit status detection which is stumping me Jul 30 06:23:14 hmm Jul 30 06:23:17 works on my end Jul 30 06:23:29 if ps -ef | grep "autobuild" | grep -v "grep" >& /dev/null ; then \ Jul 30 06:23:29 ${MAKE} update build upload ; \ Jul 30 06:23:29 fi Jul 30 06:23:43 if ps -ef | grep "lpd2"| grep -v grep >& /dev/null ; then echo "detected" ; fi Jul 30 06:23:46 for what values of "works" ? Jul 30 06:23:58 if lpd2 is running, it echoes detected, if not it doesn't Jul 30 06:24:07 I can get it to run the body easily. I can't get it to *not* run the body Jul 30 06:24:22 strange Jul 30 06:24:33 well. Jul 30 06:24:33 let me try again Jul 30 06:24:44 first of all, swap grep -v grep and grep autobuild Jul 30 06:24:53 you want the status of that last grep Jul 30 06:25:33 nope, still runs twice Jul 30 06:26:13 rwhitby didi you get it? Jul 30 06:26:28 ah okay Jul 30 06:26:39 get what? Jul 30 06:26:46 your greeping Jul 30 06:26:52 not yet Jul 30 06:26:55 I've been using pidof.. Jul 30 06:27:05 or ps -ef | grep... Jul 30 06:27:22 hmm Jul 30 06:27:56 if [ `pidof`.... Jul 30 06:27:56 <[cc]smart> i'm irritated: ERROR: Nothing provides cyrus-imapd Jul 30 06:29:38 pidof autobuild doesn't work, cause autobuild is the makefile target, not the executable Jul 30 06:30:34 if ["`ps -ef | grep autobuild | grep -v grep | cut -c10-15`" == "" ] ; then Jul 30 06:30:38 or somethinjg like that. Jul 30 06:31:15 I just made it work for slugbot last night, check dyoung@nudi:bin Jul 30 06:32:56 hello, sorry nobody answer on general : md5 of Unslung-5.5-beta-firmware.zip is ok ? I didn't get 7e0ec615a9f8fc614aae65ac484a93d8 as on web page Jul 30 06:52:02 <[cc]smart> NAiL: you have a hint for me why cyrus-imapd might be invisible ? Jul 30 06:52:19 [cc]smart: it's not in openslug-packages, I guess Jul 30 06:52:39 not quite sure how all that "no more oe-symlinks" stuff works Jul 30 06:53:01 <[cc]smart> no more oe-symlinks ? meaning the method has changed ? Jul 30 06:53:59 <[cc]smart> so i can dump the dir ? Jul 30 06:54:00 [cc]smart, bitbake -b cyrus-imapd Jul 30 06:54:31 that will add it to the list, which iirc is a filein your conf dir Jul 30 06:54:47 openslug-bbpackages.conf or something like that. Jul 30 06:54:56 bitbake -b packages/cyrus-imapd/xxxxxxx.bb Jul 30 06:55:01 then bitbake freeze Jul 30 06:55:06 yeah, that. ;-) Jul 30 06:55:06 03rwhitby 07org.nslu2-linux.dev * r7119e7e6... 10/Makefile: More merging in publish-openembedded, and updated autobuild (to be called from cron) Jul 30 06:55:24 or just edit conf/distro/openslug-packages.conf and packages/meta/openslug-packages.bb directly Jul 30 06:55:42 or run bb unfreeze and then bb cyrus-imapd Jul 30 06:55:54 doesnt the one in conf take precedence though? Jul 30 06:55:57 . Jul 30 06:55:57 . Jul 30 06:56:01 urgh Jul 30 06:56:06 I need a wireless keyboard Jul 30 06:56:06 <[cc]smart> me sayin Jul 30 06:56:42 dyoung: one sets BBFILES, the other sets what gets built automatically from that BBFILES Jul 30 06:57:01 and bb unfreeze unsets the restriction on BBFILES Jul 30 06:57:04 <[cc]smart> so which mthod am i actually meant to use if i'm preparing the package for endusers slowly Jul 30 06:57:35 whichever method you like :-) Jul 30 06:57:50 <[cc]smart> i like none of them. i liked the one before. Jul 30 06:58:03 <[cc]smart> which doesn't mean i'd vote for returning Jul 30 06:58:05 <[cc]smart> just for advice Jul 30 06:58:09 [cc]smart: I bet you didn't like it when svn.berlios.de was down ... Jul 30 06:58:11 you like dealing with berlios going down every other weekend? Jul 30 06:59:01 [cc]smart: the only difference now is that you add a line to openslug-packages.conf instead of adding a symlink in oe-symlinks. no big deal. Jul 30 06:59:06 <[cc]smart> a) as said, i'm not in any of these really. i just would like clean advice from those who are in the know. b) i'm not sure if berlios is the only possibility the get oe-symlinks out there Jul 30 06:59:25 <[cc]smart> rwhitby:, so i should add it in this file Jul 30 06:59:26 [cc]smart: since youre a hardcore dev, why even bother using the symlinks? Jul 30 06:59:36 <[cc]smart> me hardcore dev ? Jul 30 06:59:38 <[cc]smart> now way .) Jul 30 07:00:04 anyone who builds something which is not already in openslug-packages is a hardcore dev :-) Jul 30 07:00:13 <[cc]smart> i'm on frontend rather. like refining distro related thingd Jul 30 07:00:37 <[cc]smart> configfiles etc. Jul 30 07:00:53 <[cc]smart> no clue about bitbake stuff really Jul 30 07:01:05 that makes five of us ... Jul 30 07:01:22 [cc]smart: if the package works, feel free to add it to openslug-packages.foo Jul 30 07:02:12 <[cc]smart> the package compiles, but it's not really ready yet, since you would have to add a lot of stuff manually which is no option for typical users. Jul 30 07:02:33 <[cc]smart> but i think since it's the dev repository i can dare that Jul 30 07:02:53 <[cc]smart> so i'll go openslug-packages. it's in meta, right ? Jul 30 07:02:58 yep Jul 30 07:03:06 <[cc]smart> thx Jul 30 07:03:07 Okay, then the fast way for you is to bitbake -b it Jul 30 07:03:20 if it goes in there, then you need to add it to the .conf too Jul 30 07:03:35 <[cc]smart> once i sync, the openslug-packages change is ready Jul 30 07:03:51 (just like you had to add a symlink before) Jul 30 07:04:03 ay yes, since this is unstable feed, endusers should expect it to be.. well.. unstable. ;-) Jul 30 07:04:23 in fact, we should ensure that it is unstable Jul 30 07:05:22 <[cc]smart> now that i look at this file... thougt i had added cyrus stuff there before.... Jul 30 07:05:48 ppierre: we're not ignoring you, we know about the md5 being incorrect; just noone has dealt with the issue yet. Jul 30 07:06:02 03nail 07org.openembedded.nslu2-linux * r5a8e09b9... 10/packages/at76c503a/ (at76c503a-modules/makefile.cc.patch at76c503a-modules_cvs.bb): Fix at76c503a crosscompile. Thanks to wrichter Jul 30 07:07:53 <[cc]smart> hmm... so what should this look like now : BBFILES := "${OEROOT}/oe-symlinks/packages/*/*.bb" Jul 30 07:08:15 <[cc]smart> local.conf that is Jul 30 07:08:18 check the template. Jul 30 07:08:35 <[cc]smart> i can just use that ? amending for my environment for sure. Jul 30 07:15:57 the openslug-packages.conf file overrides your local.conf BBFILES now Jul 30 07:19:58 Announcement: "make autobuild" is now run on the official cross build machine every 15 minutes. It will update the feeds at the end of a successful build. Jul 30 07:20:39 (make autobuild will be improved in the future so that a failure in, say, openslug, doesn't stop a successfull optware build and feed upload) Jul 30 07:26:15 ok thanks dyoung Jul 30 07:59:39 03rwhitby 07org.nslu2-linux.dev * rf40f69c1... 10/Makefile: Fixed the sync-packages_clean script name (should be sync-packages-clean) Jul 30 08:10:33 03daka 07org.openembedded.nslu2-linux * r0ea2fa31... 10/packages/perl/perl_5.8.7.bb: On openslug create a package perl-modules depending on the 840+ other packages. Jul 30 08:36:05 03jbowler 07org.openembedded.nslu2-linux * r51692032... 10/packages/ (9 files in 5 dirs): Jul 30 08:36:05 propagate from branch 'org.openembedded.dev' (head 3e8a1aa38ecbd1a04d0cc182d8d1809d71b51079) Jul 30 08:36:05 to branch 'org.openembedded.nslu2-linux' (head 0ea2fa311bb3dc015275b544c3a9a6a3ed97e5ec) Jul 30 08:38:47 rwhitby, i've updated the http://www.nslu2-linux.org/wiki/HowTo/FrogLoggingSlug page considerably. The most significant thing found is a USB combo microphone/soundcard doodad for $59. If it performs, it would be an inexpensive solution for the audio recording part. Jul 30 08:41:02 geojeff: nice Jul 30 08:42:01 i bought a slug on ebay and have it on the way (my first ebay bidding experience) Jul 30 08:43:03 how much? Jul 30 08:44:15 $51 Jul 30 08:44:36 he said he used it for only 2 hours :) Jul 30 08:52:22 Building openslug: build from scratch will work again now. If you have a from-scratch build from the last few hours, delete the bitbake cache and all the tmp/*/ipkg* directories/files to get it working again. Jul 30 09:37:38 i'm always getting: Jul 30 09:37:40 NOTE: package postfix-2.0.20-r0: task do_package: started Jul 30 09:37:41 ERROR: function do_install failed Jul 30 09:37:54 even after i "make update" Jul 30 09:38:22 is there some cache i need to clean? Jul 30 09:38:33 bb -cclean postfix should do it Jul 30 09:38:45 thx Jul 30 09:38:48 That's probably the /usr/lib/libpcre.so bug isn't it? Jul 30 09:38:55 ... or is that something different Jul 30 09:49:20 did a "bb -cclean postfix" under openslug/, then tried "make update openslug" at the top, still the same error Jul 30 09:50:54 ok, what's the actual error? Jul 30 09:50:58 It's a known problem. Jul 30 09:52:45 ctory Jul 30 09:52:53 /home/slug/openslug/tmp/staging/i686-linux/bin/postconf: error while loading shared libraries: libpcre.so.0: cannot open shared object file: No such file or dire Jul 30 09:52:53 ctory Jul 30 09:53:26 LD_LIBRARY_PATH I think. Jul 30 09:53:37 any work around? Jul 30 09:53:51 I think so, several other people have encountered it. Jul 30 09:54:28 Funny, there is there gpl-source-package for an upcoming linksys device, that isn't even officially annouced. ftp://ftp.linksys.com/opensourcecode/RTP300/1.00.37/rtp300_cyt_1_00_37_gpl.tgz It looks quite promising, as it has fxos for voip... Jul 30 09:54:39 IIRC [g2] encountered it Jul 30 09:55:51 eno-away: try running postconf at the command line, see if you get the same error Jul 30 09:57:31 yeah, i got the same error invoking /home/slug/openslug/tmp/staging/i686-linux/bin/postconf on the command line Jul 30 09:58:14 It's something to do with the link command line. Jul 30 09:59:16 Yep, libpcre is there in the staging directory. Jul 30 09:59:17 something like setting rpath? Jul 30 10:00:05 Ok, you need /usr/lib/libpcre.so installed on your build system - it's a bug in the .bb file, it should be using the one in staging, but it isn't Jul 30 10:00:14 Can you slugbug this? Jul 30 10:00:34 ok, let me raise a slugbug on this Jul 30 10:01:26 The workround is to install libpcre, but that's a bad work round since libpcre is in staging and that is what the executable should be using. Jul 30 10:02:58 I can't see who the postfix maintainer is, but I know several NSLU2 people were using it... Jul 30 10:03:14 just apt-get installed libpcre3-devel Jul 30 10:04:49 jbowler-away, file the bug against "openslug fw" or "openslug packages"? Jul 30 10:05:42 installing libpcre3-devel wasn't enough on my system, as it requires the so.0 file... adding a symlink to .so.3 works however Jul 30 10:06:10 i just found that, thx Matta Jul 30 10:06:12 just remember to run ldconfig after adding it Jul 30 10:06:14 k Jul 30 10:06:27 eno-away: openslug packages Jul 30 10:07:26 smart me Jul 30 10:07:28 ;) Jul 30 10:10:16 "NOTE: package postfix-2.0.20: completed" :-) Jul 30 10:25:17 <[cc]smart> jbowler-away, eno-away: i'm on postfix stuff Jul 30 10:25:21 03daka 07org.openembedded.nslu2-linux * r652a4835... 10/packages/perl/perl_5.8.7.bb: Jul 30 10:25:21 Remove previuos change and create package perl-modules that recommends all the Jul 30 10:25:21 other 859 packages. Unfourtanly also recommends 86 packages packages that Jul 30 10:25:21 doesn't get created, so some warning messages gets displayed on ipkg install. Jul 30 10:29:53 [cc]smart: my theory was that it's just a matter of setting LD_LIBRARY_PATH somewhere. I'll assign the bug to you, it's 227 Jul 30 10:31:45 <[cc]smart> wait a sec, i'll read it first so i know if i'm able to do sth. about it Jul 30 10:35:10 <[cc]smart> hmm. i think i understand it, but i've got no real clue about how bitbake works and gave up frustrated about missing documentation more than once, so that i'm in the state of "don't bother me on bitbake". so i don't really feel inclined about htis one. i'm more on issues about how the distro operates than on how the build system operates. Jul 30 10:36:30 <[cc]smart> see, i don't even get this somple one resolved, gettin me the creeps: ERROR: Nothing provides cyrus-imapd Jul 30 10:36:56 <[cc]smart> i' ve been bitbaking this all the time, now what gives ? Jul 30 10:38:09 <[cc]smart> lost so much time about this stupid thing now. Jul 30 10:42:52 [cc]smart, as optware developer, i agree with you that the openembedded+bitbake thing seems quite complicated, but OTOH, it's a way of sharing packages with other distributions Jul 30 10:46:12 <[cc]smart> i've got nothing against OE/bitbake, to the contrary, i think it's quite good. just that i've never seen sufficient docu to make it worthwhile for me to look into it. Jul 30 10:46:40 <[cc]smart> got no time to scan source to answer questions. Jul 30 10:47:37 I find it brilliant. But docs are indeed lacking. Jul 30 10:47:45 <[cc]smart> right now e.g. i would like to know why i can't make bitbake find cyrus-sasl and cyrus-imapd anymore Jul 30 10:47:55 <[cc]smart> which it always did Jul 30 10:51:13 03nail 07org.openembedded.nslu2-linux * r5f2c5f5b... 10/packages/base-files/ (base-files/profile base-files_3.0.14.bb): Make /etc/profile use TZ variable if /etc/localtime not present. link in comments explain usage. Bump base-files rev. Jul 30 10:53:03 i have reboot problems again..... Jul 30 10:53:11 might this be a clue? Jul 30 10:53:13 <[cc]smart> dammit... somebody tell me what it takes to make bitbake find cyrus again Jul 30 10:53:14 # ls -l /share/hdd/data/etc/rc.d Jul 30 10:53:14 ---------- 1 root root 0 Jul 30 18:55 /share/hdd/data/etc/rc.d Jul 30 10:53:26 [cc]smart: gimme a sec or 200 Jul 30 10:53:31 <[cc]smart> i added it to meta/openslug-packages Jul 30 10:54:02 Try bb freeze Jul 30 10:54:21 Or add it into conf/distro/openslug-packages.conf explicitly Jul 30 10:55:16 <[cc]smart> then, what's meta/openslug-packages ? Jul 30 10:55:47 The list of packages you want built. Jul 30 10:55:53 03nail 07org.openembedded.nslu2-linux * r249b4cfe... 10/packages/ (7 files in 7 dirs): Jul 30 10:55:53 propagate from branch 'org.openembedded.dev' (head 72702dc0f20d493c1f7ba7870f55f950538bb364) Jul 30 10:55:53 to branch 'org.openembedded.nslu2-linux' (head 8d508fe9e9f9819ed89e6d7642eb19594323e771) Jul 30 10:56:10 <[cc]smart> er, then we have this 2 times ? Jul 30 10:56:34 openslug-packages.conf specify which .bb's are scanned at all Jul 30 10:56:35 <[cc]smart> like first you add it to meta then you move that to conf/distro by bb freeze ? Jul 30 10:56:35 No, openslug-packages.conf is the list of package directories. Jul 30 10:56:47 meta/openslug-packages specify which packages are actually built Jul 30 10:56:50 <[cc]smart> NAiL: we've got two files of this name as i see now Jul 30 10:57:26 openslug-packages.conf is a file containing the same information as the directory oe-symlinks (which is no longer used). Jul 30 10:57:32 <[cc]smart> ah one is bb th eother is conf Jul 30 10:58:01 <[cc]smart> so i should never have put it to .bb only to .conf Jul 30 10:58:25 It will disappear if you only put it in the .conf Jul 30 10:58:33 NOTE: Handling BitBake files: \ (0166/0191) [86 %]ERROR: [Errno 2] No such file or directory: '/home/repvik/unstable/glibc/openslug/openembedded/packages/rsync/rsync_2.6.4.bb' while parsing /home/repvik/unstable/glibc/openslug/openembedded/packages/rsync/rsync_2.6.4.bb <-- huh? Jul 30 10:59:12 <[cc]smart> so bb freeze reads meta/.bb and creates .conf ? Jul 30 10:59:21 NAiL: It's in your cache - clean the bitbake cache Jul 30 10:59:37 jbowler-away: uh, ok? How? Jul 30 10:59:55 jbowler-away: Is there any docs at all describing this fancy new way of handling bbfiles? Jul 30 11:00:05 Presumably someone removed rsync_2.6.4.bb Jul 30 11:00:17 NAiL: no Jul 30 11:00:21 yeah, two upstream upgrades in a rather short time, so I'm not surprised Jul 30 11:01:07 how do I clear the cache? Jul 30 11:01:20 rm -rf ${TMPDIR}/cache Jul 30 11:01:24 aha Jul 30 11:01:32 BTW: that error message is harmless Jul 30 11:01:55 yeah, just wondering Jul 30 11:02:05 I.e. it won't stop anything which can be built from building. Jul 30 11:02:27 [cc]smart: read packages/freeze/freeze.bb Jul 30 11:04:10 <[cc]smart> so hoiw should this help bitbake find cyrus when i wasn't able to built it ? Jul 30 11:06:20 I don't know. Jul 30 11:06:27 <[cc]smart> now, does openslug-packages.conf override local.conf ? Jul 30 11:06:46 Yes - see the include order at the bottom of bitbake.conf Jul 30 11:06:58 <[cc]smart> er, well THIS i would call a bug Jul 30 11:07:46 You have no choice now but to use the oe-symlinks Jul 30 11:08:00 <[cc]smart> oe-symlinks ? i was told this is gone Jul 30 11:08:20 openslug-packages.conf replaces it - but, unlike before, you have to use it. Jul 30 11:08:27 <[cc]smart> and since it didn't seem to have any effect i removed them Jul 30 11:08:56 Ah well, then just do bb unfreeze. Jul 30 11:09:23 If you don't care about the time to parse 2500 bb files it's easy. Jul 30 11:11:12 03nail 07org.openembedded.nslu2-linux * rff41830a... 10/packages/meta/openslug-packages.bb: Added nano Jul 30 11:11:39 <[cc]smart> ok, i'm getting closer to understand what's going on here... a) why is my config frozen right now ? i didn't fo it. b) overriding local.conf IS a bug., since local.conf is what one uses to customize on the front. if i get this right, freeze should wirte into local.conf, not into openembedded. Jul 30 11:11:42 <[cc]smart> am i getting closer ? Jul 30 11:15:16 It's a chicken and egg problem - local.conf can't override DISTRO.conf because it defines DISTRO. Jul 30 11:15:56 03nail 07org.openembedded.nslu2-linux * rebeee292... 10/conf/distro/openslug-packages.conf: Added nano and cyrus-imapd Jul 30 11:16:23 As a matter of policy we want "OpenSlug" to be strictly the result of building openslug.conf/nslu2.conf/local.conf.template Jul 30 11:16:31 <[cc]smart> in any case, it's not ok when one defines BBFILES on the front end configuration that beck end overrides it. Jul 30 11:16:38 <[cc]smart> s/beck/back/ Jul 30 11:16:41 The problem is that this pre-empts the user use of local.conf Jul 30 11:17:27 The BBFILES overide is annoying, but I couldn't find another way of doing it. Anyway, this stuff is only half done. Jul 30 11:17:49 <[cc]smart> then a possibility might be to add a new configfile next to local.conf which only lists these Jul 30 11:17:50 I think I will probably add in something like OPENSLUG_EXTRA_BBFILES. Jul 30 11:18:18 <[cc]smart> that is maybe moving out openslug-packages conf into oeroot/conf Jul 30 11:18:22 MACHINE/DISTRO should be in auto.conf. The path names (including BBFILES) should be in site.conf Jul 30 11:18:32 But the path names depend on the *distro*! Oops. Jul 30 11:19:17 In the OE view of the world (as I understand it) BBFILES is just a global - all the BBFILES available on the system - but we've moved from that because of the parse time issues. Jul 30 11:19:20 <[cc]smart> move openslug-packages.conf out of openembedded/conf into oeroot/conf and remove BBFILES from local.conf Jul 30 11:19:54 ? openslug-packages.conf has to be source controlled - like oe-symlinks Jul 30 11:20:10 <[cc]smart> what ? Jul 30 11:20:34 <[cc]smart> then why should it be overwritten when i apply changes to it ? Jul 30 11:20:35 The list of bbfiles has to be under source control. Jul 30 11:21:02 It isn't if you make the changes in the correct way. Jul 30 11:21:43 But if you do that, you can't include updates from the souce controlled list. Another problem... Jul 30 11:22:21 BTW python in .conf files, executed at the time the conf files are read, does not work. Yet another problem... Jul 30 11:23:03 <[cc]smart> so openslug-packages.conf is the distro default, correct ? Jul 30 11:23:27 What we have at present is semi-broken. In particular it is simply unworkable to have configuration information in local.conf when that configuration information may need to be changed from the source tree. Jul 30 11:24:13 So I started to try to work out how to fix that - it's fairly obvious - but I hit a brick wall, not because it can't be done, but because I can't motivate myself to do it. Jul 30 11:25:00 Since I want ucslugc.conf to work, and that's a whole lot of work (a new distro), I don't think I'll end up doing anything on the master makefile stuff for a while. Jul 30 11:25:21 <[cc]smart> i'm cluerupted Jul 30 11:25:55 <[cc]smart> but what seems to be realtively easy to me is separating local.conf in sth. like core.conf and distro.conf Jul 30 11:26:23 <[cc]smart> core.conf containing things like MACHINE and DISTRO Jul 30 11:26:23 Correct. Jul 30 11:26:31 Well, no - that's auto.conf Jul 30 11:26:34 <[cc]smart> and distro.conf containg e.g. BBFILES Jul 30 11:27:06 <[cc]smart> so that chicken and egg are eaten by hawk Jul 30 11:27:11 The stuff in local.conf (i.e. the one generated from the local.conf.template by the master makefile) needs to be moved to three places - auto.conf, site.conf and openslug.conf Jul 30 11:27:22 local.conf should not exist. Jul 30 11:27:53 <[cc]smart> and all these three are frontend configfiles.... Jul 30 11:27:55 <[cc]smart> ? Jul 30 11:28:05 See bitbake.conf Jul 30 11:28:07 <[cc]smart> so user configs Jul 30 11:28:19 <[cc]smart> taking a bit time... Jul 30 11:28:37 Now, auto.conf could define BBFILES, then local.conf could add/hack it. Jul 30 11:29:12 Hum. Ok - this is why my brain started to turn to jelly. Getting it right is pretty difficult. Jul 30 11:29:18 <[cc]smart> i thought conf/openslug-packages.conf would be the deafult for openslug.... Jul 30 11:29:37 Well, look at openslug.conf Jul 30 11:30:16 What matters is where 'freeze.conf' gets included. It has to be in a file generated by the build system (either source controlled or from the makefile). Jul 30 11:30:30 jp30, if you read this log, read my comments in bug #224, your suggestion seems to work for me ;) Jul 30 11:30:38 if i wanted to compile apex... is there somewhere i should put the source? Jul 30 11:30:42 an environment that needs to be setup? Jul 30 11:31:02 i have completed build-openslug with my master makefile. Jul 30 11:31:09 landslide: iirc, there's a "howto" of somesort on the wiki Jul 30 11:31:14 auto.conf is fine I think - simply MACHINE="nslu2" DISTRO="openslug" include conf/distro/freeze.conf Jul 30 11:31:45 ok. Jul 30 11:32:17 Well, ok, I might try this some time, but it's unrewarding work (cause people just whine about the result). Jul 30 11:32:56 nail: i've read it but they use crosstool. i was wondering if it was possible to setup and 'environment' from the master makefile. Jul 30 11:33:26 landslide: aha. I think that is possible. Don't know how to though. beewoolie and [g2] are the apex guys Jul 30 11:33:31 <[cc]smart> jbowler-away: if i get this righzt, freeze.conf should not exist as a separate thing. freeze should be a tool that modifies a frontend user configuration file. Jul 30 11:33:42 landslide: but they're not here at the moment Jul 30 11:33:55 [cc]smart: I don't think so Jul 30 11:34:07 <[cc]smart> and auto.conf should be renamed, since the name is meaningless. Jul 30 11:34:10 Well, the first part - the second is the way it is at present. Jul 30 11:34:36 <[cc]smart> or maybe not Jul 30 11:34:39 <[cc]smart> will have to read it first Jul 30 11:34:40 landslide: there is a target setup-apex in this mastermakefile, but dunno if it works Jul 30 11:35:18 auto seems fine to me. I believe they meant it as 'automatically configure for different builds'. Jul 30 11:35:19 <[cc]smart> Q: in bitbake.conf it's always conf/sth. how to tell which conf dir this is, in OEROOT or in openembedded ? Jul 30 11:35:29 BBPATH Jul 30 11:36:08 <[cc]smart> e.g. it says: include conf/distro/${DISTRO}.conf Jul 30 11:36:13 <[cc]smart> which is in openembedded dir Jul 30 11:36:25 <[cc]smart> and it says: include conf/local.conf Jul 30 11:36:30 <[cc]smart> which is in OEROOT Jul 30 11:37:42 and both are on BBPATH Jul 30 11:38:07 (It's a search path, hence the name) Jul 30 11:38:08 <[cc]smart> in any case, i don't see a way to override DISTRO.conf since it's preety much the last Jul 30 11:38:31 That's the conclusion I reached. Jul 30 11:38:36 <[cc]smart> so how is this again that i would be able to make cyrus come back ? Jul 30 11:38:44 <[cc]smart> i don't see a way Jul 30 11:39:05 Do any of the things I said above. Jul 30 11:39:35 (Well, I'm unclear as to what the problem is - I was assuming it's file-not-found) Jul 30 11:40:13 <[cc]smart> freeze won't fo it Jul 30 11:40:26 <[cc]smart> since cyrus isn't baked yet Jul 30 11:40:54 <[cc]smart> oe-symlinks afaik are out already. i had them and there where the cyrus packages linked. didn't do anythig Jul 30 11:41:01 <[cc]smart> so what's left ? Jul 30 11:41:28 <[cc]smart> adding to openslug-packages.conf only i guess Jul 30 11:41:59 I don't know - I don't understand what the problem is. Jul 30 11:42:18 <[cc]smart> hmm.. can you bitbake cyrus-imapd on your copy ? Jul 30 11:42:44 <[cc]smart> for me the problem seems clear though Jul 30 11:42:53 * NAiL is bitbaking cyrus-sasl right now Jul 30 11:42:56 I can't do anything because ipkg* were all broken so I don't have a build. Jul 30 11:43:26 <[cc]smart> the last deifnition for BBFILES is in openslug-packages.conf and it overrides all before Jul 30 11:43:52 <[cc]smart> so now cyrus anymore wherever you might have included it Jul 30 11:44:06 <[cc]smart> ha Jul 30 11:44:24 <[cc]smart> so there's maybe a simple improvement possible Jul 30 11:44:56 Yes, for me bb cyrus-imapd builds openssl (first) Jul 30 11:45:11 ... because I've only build openslug-image so far today Jul 30 11:45:29 [cc]smart: what is the problem? Jul 30 11:45:30 <[cc]smart> modifyinf openmbedded/conf/distro/openslug-packages.conf (and the others) replacing BBFILES := with BBFILES += Jul 30 11:45:33 NOTE: package cyrus-imapd-2.2.12: started Jul 30 11:46:30 On a system after a bb unfreeze? Jul 30 11:46:40 <[cc]smart> for me, and i recently updated my working copy i get: ERROR: Nothing provides cyrus-imapd Jul 30 11:46:55 <[cc]smart> all the day Jul 30 11:46:55 [cc]smart: wait five mins, do make update, retry. Jul 30 11:47:06 <[cc]smart> whatever i try Jul 30 11:47:07 NOTE: package cyrus-imapd-2.2.12: completed Jul 30 11:47:20 Ah, ok - you added it to openslug-packages.bb... Jul 30 11:47:27 <[cc]smart> and removed it again Jul 30 11:47:34 <[cc]smart> tried all kind of things Jul 30 11:47:43 No, NAiL added it. Jul 30 11:47:53 yes Jul 30 11:47:59 missed cyrus-sasl though Jul 30 11:48:03 Added that just now ;-) Jul 30 11:48:04 <[cc]smart> i did this alread too but it means nothing Jul 30 11:48:45 NAiL: bb freeze will generate the correct openslug-packages.conf, then copy it over conf/distro/openslug-packages.conf and do an mt diff to make sure no cruft gets included. Jul 30 11:48:59 <[cc]smart> and i bet if i add to openslug-packages.conf it will do it. Jul 30 11:49:32 [cc]smart: either use bb unfreeze or use the -b argument to bitbake or edit openslug-packages.conf or create your own. Jul 30 11:49:36 [cc]smart: "make update" Jul 30 11:49:45 <[cc]smart> jbowler-away: if i underestood correctly, this will ONLY be a good idea if NAiL bwas bitbaking all openslug-packages before Jul 30 11:49:45 then bb cyrus-imapd. Jul 30 11:50:18 <[cc]smart> i do not bake openslug-packages. this will be WAY more timeconsuming than parsing all bb files for what i need Jul 30 11:51:01 Then you want option 4 Jul 30 11:51:14 <[cc]smart> option 4 ? Jul 30 11:51:26 03nail 07org.openembedded.nslu2-linux * rabe78818... 10/conf/distro/ (openslug-packages.conf openslug.conf): Added cyrus-sasl, needed for cyrus-imapd Jul 30 11:51:28 the fourth option I gave Jul 30 11:51:44 *argh* Jul 30 11:51:51 bad commit Jul 30 11:52:16 The openslug.conf? Jul 30 11:52:33 yes Jul 30 11:52:38 just feed stuff Jul 30 11:52:49 two #'s in the wrong place :P Jul 30 11:53:19 How do you (NAiL) feel about all those ?= assignments in openslug.conf becoming = ? Jul 30 11:53:20 <[cc]smart> sorry, didn't count the options, would oyu mind naming it instead of numbering ? Jul 30 11:53:32 I.e. remove the ability to hack the openslug firmware image. Jul 30 11:54:12 jbowler-away: I care incredibly little about changing the firmware image itself Jul 30 11:54:19 [cc]smart: (1) bb unfreeze (2) use the -b argument to bitbake (3) edit openslug-packages.conf (4) create your own [openslug-packages.conf] Jul 30 11:54:23 actually, I haven't even tried Jul 30 11:54:32 oh, yes, I have... udev. Jul 30 11:54:46 Yes - but udev is something which would probably be checked in. Jul 30 11:54:58 I.e. you would be happy to directly hack on openslug.conf (I think) Jul 30 11:54:58 <[cc]smart> i will try 5) soonish (once update completed) Jul 30 11:55:06 sooner or later yes. But I really don't care much about not being able to hack the firmware image without hassle Jul 30 11:55:12 [cc]smart: what is (5) Jul 30 11:55:27 (5) is just "bb cyrus-imapd" :-P Jul 30 11:55:35 :) Jul 30 11:55:43 <[cc]smart> 5) consits of hacking openslug-packages.conf to go replacing BBFILES := with BBFILES += Jul 30 11:55:47 aha Jul 30 11:55:56 <[cc]smart> and emptying local.conf BBFILES definition to "" Jul 30 11:55:58 [cc]smart: just so long as you never check the result in. Jul 30 11:56:05 <[cc]smart> no, wont Jul 30 11:56:21 <[cc]smart> but if it works, i would say this is a good solution to go with Jul 30 11:56:32 It doesn't work. Jul 30 11:58:07 [cc]smart: with the current repo, cyrus-imapd builds just fine on my end Jul 30 11:58:22 (with the changes i checked in a while back) Jul 30 11:58:45 <[cc]smart> NAiL: yes, just saw that your update adds the two to openslug-packages.conf, but this is cheating about the issue we discuss :) Jul 30 11:59:19 cheating? Jul 30 12:00:07 [cc]smart: look. Do "bb -b ../openembedded/packages//_.bb, wait until completes, then bb freeze" Jul 30 12:00:23 s/,/",/ Jul 30 12:00:26 03nail 07org.openembedded.nslu2-linux * ra06f8ca7... 10/conf/distro/openslug.conf: Set the feeds correctly again Jul 30 12:00:29 03nail 07org.openembedded.nslu2-linux * r42d68bf4... 10/packages/meta/openslug-packages.bb: Added cyrus-imapd back to packages Jul 30 12:00:30 <[cc]smart> yes, since what we discuss is how to go about deviating from the default given by "official" feed Jul 30 12:00:40 <[cc]smart> this is what you normally do with local.conf Jul 30 12:00:46 <[cc]smart> frontend configuration Jul 30 12:01:05 If you are making a package for the official feed the correct and only way to do it is option (3) - edit the checked in openslug-packages.conf Jul 30 12:01:20 <[cc]smart> yes, that's true Jul 30 12:01:37 <[cc]smart> and when i'm working on sth. for myself only ? Jul 30 12:01:52 It's like with oe-symlinks, except now any package developer with write access can update the 'symlinks', therefore now every package developer is expected to do it! Jul 30 12:02:29 ah, finally Jul 30 12:02:33 <[cc]smart> with oe-symlinks i was able to add sth. for myself without the risk to - overwrite it with next update- or unintentionally influence others Jul 30 12:02:38 * NAiL is set up to do changes to stable Jul 30 12:03:40 If you are doing private packages then you can use any of the other three options, depending on how much you want to track the core changes. Jul 30 12:03:48 [cc]smart: you still are, as long as you don't check in the change... Jul 30 12:04:18 if you accidentally do, you can revert the commit Jul 30 12:04:32 Well, option (4) (which is actually equivalent to option (1) or (2) followed by bb freeze) avoids that accident. Jul 30 12:05:08 In fact the same accident could occur with oe-symlinks (very easily) Jul 30 12:05:18 <[cc]smart> bb freeze fixes me on what i have, and i don't want to each time i want to change/add a single thing go freeze and unfreeze Jul 30 12:05:45 <[cc]smart> if i didn't bake all openslug-packages before i freeze i make all "official" invisible Jul 30 12:05:47