**** BEGIN LOGGING AT Mon Oct 01 02:59:56 2007 Oct 01 03:23:56 scruggs: pong Oct 01 03:25:54 sakoman: get it working? Oct 01 03:27:01 yes! Oct 01 03:27:30 awesome! Oct 01 03:27:30 trying to get it to autoload at boot now Oct 01 03:27:49 I might be able to help you there Oct 01 03:28:17 what is your approach now to get it to autoload? Oct 01 03:28:57 /etc/network/interfaces & /etc/modprobe.conf Oct 01 03:29:19 as well as specifying autoload in the conf Oct 01 03:29:51 I didn't realize using /etc/modprobe.conf would work - I'll have to give that a shot Oct 01 03:30:20 I'm specifying pcmcia to load in the machine.conf Oct 01 03:30:37 and then have an init script to load pxa2xx-cs Oct 01 03:31:08 I haven't tried loading all the way up to the mcf25 driver, but that's what I'm working on now Oct 01 03:31:10 just tried it -- sort of works, but does give an error message: Oct 01 03:31:16 BUG: scheduling while atomic: iwconfig/0x00000001/412 Oct 01 03:31:33 But the network works after that Oct 01 03:31:37 the current buildroot for gumstix does that too Oct 01 03:31:44 Ah, OK Oct 01 03:31:50 yep - still works after the message Oct 01 03:32:12 but for autoloading the modules you have to make sure pcmcia is loaded Oct 01 03:32:20 before loading pxa2xx-cs Oct 01 03:32:24 sure, I do that Oct 01 03:32:57 and adding both of them to do that in the machine.conf causes the stix to freeze during boot (at least it does for me) Oct 01 03:33:27 so I wrote the init script to run after udev starts and after all the other 'autoload' modules have been loaded Oct 01 03:33:56 I'm working off a custom machine and image, with angstrom as the distro Oct 01 03:34:05 with that I think even if you tell the mcf25 module to 'autoload' the wifi will still work once the pxa2xx-cs module has been loaded Oct 01 03:34:42 yeah, I've modified the machine confs for the .dev branch too, and am using a custom image Oct 01 03:35:43 Seemed to have no choice if I wanted to get the size down to match buildroot Oct 01 03:36:13 I'm starting to see that too Oct 01 03:37:04 but in my case I have attached cf as storage,so being able to tell an ipkg to install there makes size not as much of an issue for me Oct 01 03:37:05 I only autoload pcmcia and mcf25 (& some usb gadget stuff) and have no freeze issues Oct 01 03:37:31 then once you boot you load pxa2xx-cs with modprobe? Oct 01 03:37:46 Nope, don't have to do anything by hand Oct 01 03:38:24 just use the mods to modprobe.conf and network/interfaces Oct 01 03:38:45 so do you actually create the modprobe.conf file? Oct 01 03:40:11 yes, you should override modutils with the collection feature and put the modprobe.conf from buildroot in a machine directory in modutils/files Oct 01 03:40:37 I assume your image includes task-boot? Oct 01 03:40:37 ahhh... ok Oct 01 03:41:09 actually my image just iherits angstrom-minimal-image and adds to it Oct 01 03:41:11 That's where modutils-initscripts gets specified as a dependency Oct 01 03:42:05 I bet thats where you setup the interfaces file for netbase too - I notices a number of other distros do that Oct 01 03:42:29 I'm slowly learning how oe works :) Oct 01 03:42:34 Similar exercise for network/interfaces using the netbase package Oct 01 03:42:50 Same here Oct 01 03:43:52 not sure if you need it, but I tested the sound stuff and the gps on a gpsstix and it all worked just fine 1st time Oct 01 03:44:33 Cool! No sound for my device, so I haven't tried it Oct 01 03:45:08 I was amazed when it worked Oct 01 03:45:34 I'm so used to sound NOT working with buildroot that I haven't even tried it in months Oct 01 03:45:53 to have it work on the first try was great Oct 01 03:46:09 I'll be glad to get over this OE learning phase and get back to real work! Oct 01 03:46:33 same here my friend! Oct 01 03:47:15 I'm finally within about 1MB of the old buildroot image Oct 01 03:47:48 Need to figure out why gnuplot pulls in libstdc++ and I'll be just about at parity! Oct 01 03:47:56 1MB is still a lot of real estate when you're talking embedded Oct 01 03:48:18 I noticed that today too - first thing pulled was libstdc++ Oct 01 03:48:31 libstdc++ is 700+K Oct 01 03:48:40 I went to add gpsd and it had a dependency on X!! Oct 01 03:49:02 yeah, lots of stuff in my bitbake collection Oct 01 03:49:15 I can imagine :) Oct 01 03:49:31 A big part of the savings came from moving away from task-base Oct 01 03:49:48 I might bug you for edited bb files if I find some a bit too bloated for my liking Oct 01 03:50:17 I should dig into task-base a bit more Oct 01 03:50:17 Sure, no problem. Happy to share. Oct 01 03:50:24 thx Oct 01 03:50:52 After I clean up my "task-base-gumstix" I'll send you a copy Oct 01 03:51:15 that would be great Oct 01 03:55:24 scruggs: I think the gpsd dep on X11 is build time only... There are some X11 test prorams included.... Oct 01 03:57:09 hvontres|home: I'll have to check that out - I never actually followed through to see if it would require that X be installed - thx! Oct 01 04:00:09 scruggs: even if it isn't installed it may cause the config phase of other packages that you do need to detect libs that X required (in staging) Oct 01 04:00:43 And therefore configure themselves to use (and hence require) said libs Oct 01 04:01:06 Causing bloat :-( Oct 01 04:02:03 I was wondering if that would be the case Oct 01 04:02:20 at least using collections is pretty easy to get working Oct 01 04:03:04 Yeah, but a pain to have to override so much Oct 01 04:04:37 not sure that I want to know, but how many pkgs are you overriding to make the size cut? Oct 01 04:05:18 eleven at the moment Oct 01 04:05:50 what distro do you build on? Oct 01 04:06:00 angstrom-2007 Oct 01 04:06:20 I mean your build host machine ;) Oct 01 04:06:34 ah, Ubuntu 7.04 Oct 01 04:06:50 ok - I use gentoo Oct 01 04:07:10 one thing I like about gentoo is it has a concept of 'use' flags Oct 01 04:07:35 so if I say I never what to build an 'X' option into a package I can set that globally Oct 01 04:07:46 or on a package-by-package basis Oct 01 04:07:54 That's nice Oct 01 04:08:04 so you really never that to build anything that you don't want to build Oct 01 04:08:06 Too painful to switch now though :-) Oct 01 04:08:28 switching is not an option :) Oct 01 04:08:45 Yeah, that could easily kill a week! Oct 01 04:09:07 but coming from something like that it makes me wonder why a build system that targets embedded devices doesn't have a similar facility Oct 01 04:09:37 coming from gentoo that is Oct 01 04:09:39 Maybe it does and I'm just too much of a newbie to know about it! Oct 01 04:10:39 hmmm... good point - mabye someone here on the channel could confirm for a couple of newbies :) Oct 01 06:36:12 * Bernardo is back. Oct 01 07:12:39 moin Oct 01 07:12:51 having something like http://www.kegel.com/crosstool/crosstool-0.43/buildlogs/ on the OE website would be nice Oct 01 07:15:14 and could it be that i cant build uclibt 0.9.29 with gcc 3.3.4? Oct 01 07:26:59 morning Oct 01 07:28:23 good morning all Oct 01 07:29:00 good morning! Oct 01 07:37:47 yo koen, fixed networkmanager Oct 01 07:47:02 morning all Oct 01 07:47:58 03koen 07org.oe.dev * r9d27acb3... 10/ (1 conf/distro/include/sane-srcrevs.inc): sanesrcrevs: bump webkit dependants to catch up with gdk->gtk renaming Oct 01 08:03:08 hi, im using gcc preferred version 3.3.4 and binutils-cross 2.14.90 glibc 2.3.3 and uclibc 0.9.29 and get that error: http://pastebin.com/m5009dadb Oct 01 08:03:12 anybody saw this before? Oct 01 08:04:09 target ios arm ep93xx Oct 01 08:30:27 gm Oct 01 08:30:41 yo Oct 01 08:33:27 hm.. likewise - if i understand you right, you haven`t seen my bug with gcc-cross? Oct 01 08:36:12 Darth_Wader: I am not getting that bug, I did read your report though. I cannot understand what happens exactly. Oct 01 08:37:33 likewise: i think that it launches configure script for second time with incorrect enviropment or parameters Oct 01 08:39:29 and second thing, who ever will need fortran on devices like pda(most of which even without hw mmu unit :/) Oct 01 09:14:28 WTF happened to dbus-daemon Oct 01 09:39:21 can we fetch patches (single files) from a repo using SRC_URI? Oct 01 09:39:53 likewise: http://www.openembedded.org/repo/org.openembedded.dev/packages/linux/linux-ezx_2.6.21.bb Oct 01 09:40:29 fetching single files with svn is a bit icky, since you need a *very* recent svn to do that Oct 01 09:40:49 but fetching patches from http shouldn't be a problem Oct 01 09:42:58 koen: but from CVS seems impossible. Oct 01 09:44:54 koen: When moving patches out of the OE metadata (see discussion on the ml), we lose the hard link between versions and builds break more often (patches disappearing etc). Oct 01 09:44:58 which once again proves the point - stay away from CVS whenever possible ;] Oct 01 09:45:16 gilligan_: the main is SRC independent. Oct 01 09:45:19 gilligan_: the main problem is SRC independent. Oct 01 09:45:37 gilligan_: namely do we want patches to be part of the metadata or not? Oct 01 09:45:42 I like to have all patches we apply in .dev Oct 01 09:45:57 so we know what we apply and can track changes better Oct 01 09:45:57 koen: so, in the metadata Oct 01 09:46:08 koen: me too. Oct 01 09:46:11 but some people find that a too big a burden Oct 01 09:46:56 koen: well, I can see it blows up the metadata considerably, but I think it is the right way to keep track of things. Oct 01 09:47:13 disk space is cheap Oct 01 09:47:22 (on buildmachines) Oct 01 09:47:47 * XorA has plenty of disk space, lacks cpu power :-D Oct 01 09:48:17 XorA: dbus r7 should be regression free Oct 01 09:48:42 note to self: don't merge poky stuff before having coffee Oct 01 09:49:54 koen: cheers Oct 01 09:50:28 .oO( "coders are merely tools to convert coffee into code ) Oct 01 09:52:19 03koen 07org.oe.dev * r7902f4e2... 10/ (1 packages/dbus/dbus.inc packages/dbus/dbus_1.0.2.bb): dbus: tweak packaging, should help XorA's single core build :) Oct 01 09:52:50 that's the r6 commit Oct 01 09:52:57 r7 adds dbus-uuidgen Oct 01 09:53:13 gilligan_: s/tools/machines/ Oct 01 09:53:55 with that maybe NetworkManager will start Oct 01 09:55:58 koen: did you close #3091 Oct 01 10:08:36 hrm.. i am starting to lose orientation again.. where is the angstrom-minimal-image-with-mtd-tools or whatever the exact name was defined ? Oct 01 10:11:57 gilligan_: there's .bb with such name. find/grep -r are your friends Oct 01 10:13:20 is it possible to to a PREFERRED_PROVIDER_virtual/kernel_machine ? Oct 01 10:13:29 i want to specify the machine kernel in my distro conf Oct 01 10:13:32 ah.. in packages/images.. Oct 01 10:14:07 like PREFERRED_PROVIDERS_virtual/kernel_gumstix-connex Oct 01 11:35:02 hm... configure in uclibc 0.9.29-r7 just failed during a build of angstrom 2008.1 Oct 01 11:36:45 lxdialog/dialog.h : error: curses.h: No such file or directory Oct 01 11:49:42 Philippe: hey Oct 01 11:50:19 hi zecke Oct 01 11:53:10 * zecke gets slowly back to berlin, I'm in germany already :} Oct 01 12:05:34 who created the ep93xx machine configuration? Oct 01 12:13:37 hi Oct 01 12:14:11 I have a package that needs a special configure option when run on a Intel XScale CPU. Oct 01 12:15:17 how can I make the addition of this configure option automatic depending on the target CPU? Oct 01 12:15:36 (the package is kaffe 1.1.8 and I want to sent my updates for inclusion) Oct 01 12:16:25 rschuster: hmm, we have a TARGET_ARCH but no TARGET_CPU Oct 01 12:16:29 ah the configure option should not be there if the target CPU is some other arm-variant (no iwmmxt) Oct 01 12:16:54 rschuster: check the mplayer recipes Oct 01 12:17:03 those have iwmmxt overrides Oct 01 12:17:03 koen: thanks. will do so Oct 01 12:17:11 koen: when are you going to come to berlin? Oct 01 12:17:12 not perfect, but good enough Oct 01 12:17:17 zecke: the 5th Oct 01 12:17:27 I arrive at 13:18 at hbf Oct 01 12:17:45 and will wait a bit for XorA to get there and then head in your direction Oct 01 12:18:58 koen: basically all devices which are known to contain xscale CPUs are handled specially in the BB right? Oct 01 12:19:16 * RP suspects he should have been arriving earlier for OEDEM :/ Oct 01 12:19:16 rschuster: yes Oct 01 12:19:37 RP: we will wait for you. When are you going to join us? Oct 01 12:19:44 rschuster: in the near future I want to have something like MACHINE_FEATURES = "iwmmxt" Oct 01 12:19:50 zecke: Sunday evening Oct 01 12:20:45 koen: ok. would OE accept a machine conf even it is not possible to build a real operating system image for it (I am talking about the irex iliad ereader which has proprietary kernel modules) Oct 01 12:21:10 rschuster: the iliad is already in :) Oct 01 12:22:25 rschuster: er0100.conf is the iliad Oct 01 12:22:42 (that's how irex calls it in their OE snapshot) Oct 01 12:23:17 koen: know who build the ep93xx machine configuration? Oct 01 12:23:27 nik0n: I did Oct 01 12:23:50 koen: hm, thats odd. I thought they name it ereader. they /etc/ipkg/arch.conf suggests that to me ... Oct 01 12:23:57 koen: If I'm bored I'm going to pick you up Oct 01 12:24:15 rschuster: well, I used the files matthijs gave us Oct 01 12:24:21 and he's the irex OE dude :) Oct 01 12:25:06 koen: ok then I will integrate that with this description :) Oct 01 12:26:21 koen: from where u took the eraseblocksize? i get tons of this errormsg Oct 01 12:26:22 jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x0031ff7c: 0x0689 instead Oct 01 12:26:25 jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x0031ff80: 0xb174 instead Oct 01 12:26:35 when using ep93xx as machine and flash the image to my 9302 Oct 01 12:26:57 I took the erasesize for my glomation board Oct 01 12:27:06 could be otherboards need something different Oct 01 12:27:16 which on eyou got? Oct 01 12:27:32 or where u took that size? flash chip data sheet? Oct 01 12:29:12 zecke: have you looked at cython for the bitbake c parser? Oct 01 12:29:42 koen: I just updated my org.oe.dev branch and there is no conf/machine/er100.conf Oct 01 12:29:59 koen: no, do you what is missing for the parser? locking me into a room 48hours and giving me enough mate... (hint there is a mate dealer next to our office) Oct 01 12:30:44 oh and my mtn2git tool is done :) Oct 01 12:32:39 rschuster: ah, heh Oct 01 12:32:48 it seems I never actually committed it.... Oct 01 12:32:55 lol Oct 01 12:33:03 well, it's in now :) Oct 01 12:33:34 gilligan_: install libncurses5-dev -- uclibc build is still a little broken Oct 01 12:33:37 lol, looks I have new friends at IBM GES :} Oct 01 12:33:45 * cbrake checks if there is an open bug on this ... Oct 01 12:34:01 cbrake: dooesn't OE build ncurses-native? Oct 01 12:34:13 koen: thanks Oct 01 12:34:32 koen: I think it does, but if I recall, uclibc must still be looking in the host headers Oct 01 12:35:08 ./me arrives Tegel at 10AM on Oct 5 Oct 01 12:35:28 Friday will be spent staying up as long as I can Oct 01 12:38:35 bbl Oct 01 12:38:48 koen: http://bugs.openembedded.org/show_bug.cgi?id=2545 Oct 01 12:39:42 cbrake: ah, right Oct 01 12:40:52 rschuster: I tried your 0.95 classpath recipe from the bug tracker last week -- I hope to work on it a little more this week. Oct 01 12:43:04 cbrake: I am sorry. I have not updated the recipes in the bugtracker for a while. but I have newer ones locally. Oct 01 12:43:51 cbrake: only problem is that my approach conflicts with the one taken by henning. and to fix the real issue (java toolchain in oe) it would take much longer ... :( Oct 01 12:43:54 rschuster: I would be interested in testing and pushing into mtn if you have time to upload them. Oct 01 12:44:06 rschuster: ahh, IC Oct 01 12:44:18 cbrake: I will upload them Oct 01 12:44:55 rschuster: does any of the current java stuff in OE -- your stuff was the first time I had any success running anything Oct 01 12:45:08 rschuster: *stuff in OE work ... Oct 01 12:49:04 cbrake: yep. it works Oct 01 12:49:29 cbrake: I have this successfully running on the irex ereader (arm xscale) and neo1973 Oct 01 12:49:59 rschuster: excellent -- was this with your recipies or the ones in OE? Oct 01 12:51:07 XorA, rschuster: what do you think of this: http://ewi546.ewi.utwente.nl/tmp/mplayer-iwmmxt.diff.txt ? Oct 01 12:52:40 cbrake: this was done with the recipes I have locally Oct 01 12:52:58 cbrake: I am using them via an overlay (don't beat me to much for this) Oct 01 12:53:10 cbrake: they are available through a public SVN repo, too Oct 01 12:53:25 cbrake: along with updated recipes for cacao, java-gnome etc Oct 01 12:53:44 cbrake: I plan to merge this in OE. but I have very little time recently :( Oct 01 12:54:53 cbrake: if you have a neo or iliad you can even install working binary packages (http://jalimo.org/wiki/doku.php?id=packages) Oct 01 12:55:33 cbrake: current building instructions for OE are here: http://jalimo.org/wiki/doku.php?id=doc:howto:buildingforopenembedded Oct 01 12:58:22 koen: the ep9302 board needs eraseblocksize 0x20000 jfyi Oct 01 12:58:32 not your 40000 Oct 01 13:00:14 rschuster: thanks for the link -- looks very good. I'll try to help out on the OE integration where I can ... Oct 01 13:00:51 cbrake: updated my recipes at http://bugs.openembedded.org/show_bug.cgi?id=2730 Oct 01 13:01:00 rschuster: thanks! Oct 01 13:01:36 cbrake: thanks for the help. If you have any question about the recipes or something just ping me :) Oct 01 13:02:07 rschuster: will do Oct 01 13:02:37 rschuster: amazing all the things that are going on -- this is the first I've heard of Jalimo :-) Oct 01 13:03:42 cbrake: I am sorry for that. tarent is already using this for customer projects and I am involved in that. so time for the OE merging (which is also my part) is low :| Oct 01 13:04:17 koen: good for me, saves adding a bazilion machines Oct 01 13:04:46 cbrake: but it is good to know that someone else apart from me is now reviewing these recipes :) Oct 01 13:04:53 rschuster: yeah, I understand -- I'm overbooked myself right now -- are there any maillists or irc channels you recommend for keeping up with mobile java developments Oct 01 13:06:11 cbrake: project page: http://evolvis.org/projects/jalimo/, http://evolvis.org/mail/?group_id=11 Oct 01 13:06:47 cbrake: but please note that we are only using the free runtimes and class libraries. so it is not the 'official' java :) Oct 01 13:07:59 cbrake: openjdk integration is done by the various runtime people. there is high flux on that. We plan to integrate that once it is easy to build and work Oct 01 13:08:13 rschuster: yes, that makes sense Oct 01 13:08:44 cbrake: will you be at OEDEM? Oct 01 13:10:06 rschuster: no, I don't plan to be there this year -- maybe next year if I do some planning ahead Oct 01 13:12:35 koen: the ereader machine conf you put in will not generate ARCH=arm packages by default. would you mind setting PACKAGE_ARCH to "arm"? that way the created packages would be architecture compatible to the original OS distribution from irex. or am I supposed to do this through local.conf? Oct 01 13:14:06 rschuster: it has TARGET_ARCH = "arm", which is all it needs Oct 01 13:15:08 rschuster: any munging to PACKAGE_ARM will need to be done in the ereader distro (which is still in bugzilla) Oct 01 13:16:57 koen: ok. there is still a small glitch. the machine conf references tune-xscale.*conf* but it is .inc instead Oct 01 13:33:54 how do you do X forwarding using dropbear ? Oct 01 14:07:05 morning Oct 01 14:10:53 koen: Is there a package I can use to insert a custom modprobe.conf or modprobe.d into my image? Oct 01 14:23:35 having to wait up to a minute to get a reply from 'mtn status' = pain in the a*** ... Oct 01 14:28:37 curse nokia! Oct 01 14:29:01 they are buying Navteq at no real premium to the share price Oct 01 14:29:21 * mwester looks at the build log for Yet Another Failed OE Build, curses people who resist working on branches, syncs up, and settles down to wait. Again. Oct 01 14:29:33 mwester, what failed? Oct 01 14:29:37 Webkit Oct 01 14:29:46 hmm Oct 01 14:29:55 lots of monekying around there over the weekend Oct 01 14:30:00 Yeah. Oct 01 14:30:09 That's why I delayed the full build until the last minute. Oct 01 14:30:12 mwester: branches??? this concept is so advanced that a lot of people fail to understand it ;) Oct 01 14:30:20 Grrrrr..... Oct 01 14:30:39 If wasn't starting to lose my hair, I'd be pulling it out! ;) Oct 01 14:32:42 Crofton: Do you know if there a package I can use to insert a custom modprobe.conf or modprobe.d into my image Oct 01 14:32:55 What comes around, goes around - or whatever the phrase is. I'm sure that I'll commit something that will break OE, and then everyone can jump all over me -- especially given my recent rants about OE, stability, and branches. :) Oct 01 14:33:22 sakoman, not sure Oct 01 14:33:30 mwester: your best bet is to break gcc Oct 01 14:33:34 what are you trying to do? Oct 01 14:34:08 * Crofton|home is in 401K management mode ... Oct 01 14:35:05 XorA: The pango-directfb was another way to really mess things up. Oct 01 14:35:26 XorA: but those have been done already - I need something truely spectacular. Oct 01 14:35:28 Crofton: I'm trying to autostart wifistix Oct 01 14:35:34 ah Oct 01 14:35:47 and the autload line does not work? Oct 01 14:35:51 mwester: webkit builds fine on the autobuilder Oct 01 14:35:54 or there are conflicts? Oct 01 14:36:09 mwester: well the good thing about screwing up gcc is you cant go backwards, and so everyone needs to rebuild twice Oct 01 14:36:17 I have it working, just trying to find the "OE way" to put modprobe.conf or a file in modprobe.d Oct 01 14:37:06 Crofton: it needs a combo of /etc/network/interfaces and /etc/modprobe.conf in addition to the autoload line Oct 01 14:37:30 sakoman: any package should be able to add a file to modprobe.d Oct 01 14:37:39 sakoman, will it work if you add a gumstix-connex dir in packages/modutils/files and have it in there? Oct 01 14:37:44 module_autoload_ohci-hcd_omap5912osk = "ohci-hcd" Oct 01 14:37:50 in linux.inc I think Oct 01 14:37:54 scruggs: no Oct 01 14:38:04 koen: | checking for WEBKITGTK... configure: error: Package requirements (WebKitGdk) were not met: Oct 01 14:38:04 * Crofton|home thinks he had some issues with this process though ... Oct 01 14:38:08 thought that might be too easy Oct 01 14:38:19 koen: | No package 'WebKitGdk' found Oct 01 14:38:32 mwester: that's not a problem with webkit Oct 01 14:38:55 mwester: that's a problem with your app depending on an outdated version of webkit Oct 01 14:39:18 * mwester scratches his head. Oct 01 14:39:23 sakoman, I am off to school, once I see what disasters are happening, I will try and put in some gumstix time this week Oct 01 14:39:24 XorA: I know I can add it myself in a custom recipe, I just wanted to check to see if there was a "standard" way ala base-files Oct 01 14:39:35 Crofton: I ran into an issue with the pxa2xx-cs module when using the module_autoload.. method Oct 01 14:39:40 koen: Isn't that what sane-srcrevs is supposed to fix, then? Oct 01 14:39:46 bbiab Oct 01 14:39:48 sakoman: I think the preferred way would be to add it from your modules package Oct 01 14:39:59 mwester: fix what? Oct 01 14:40:03 Crofton: OK, have fun :-) Oct 01 14:40:16 mwester: sane-srcrevs only locks down srcrevs Oct 01 14:40:29 it has nothing todo with upstream being stale or buggy Oct 01 14:40:34 mwester: WebkitGdk is now WebkitGtk, OE was version bumped this morning Oct 01 14:40:56 Upstream is fine, koen -- the problem is that somewhere a change to it leaked through without being tested. Oct 01 14:41:08 XorA: I'm confused! Which modules package are you referring to? Oct 01 14:41:18 sakoman: those wifi modules Oct 01 14:41:56 the two dependants of webkit I know off (feedreader and browser) both build fine over here and on the autobuilder Oct 01 14:42:32 koen: then presumably a recent commit (in the past 8 hours) fixed it. Oct 01 14:42:32 XorA: thanks for the clarification! Oct 01 14:42:54 sakoman: I assume they are an out of kernel build? Oct 01 14:42:55 mwester: so you're complaining that your outdated checkout is outdated? Oct 01 14:43:12 sakoman: if so add the modprobe.d/wifimodules.conf file into that package Oct 01 14:43:19 XorA: yes a separate package, so it should be easy Oct 01 14:43:28 koen: No, I'm complaining that it wasn't fixed when someone first committed the change that broke it, of course. Oct 01 14:43:59 XorA: thanks! Oct 01 14:44:56 sakoman: let me know if that works out for you - that would be a lot easier than what I am doing now for autoloading Oct 01 14:47:02 scruggs: I know it works when the files are added by hand. Every boot afterwards has a functioning network from the get go Oct 01 14:48:11 scruggs: got an eye exam now, so it will be a while before I make the changes to the wifistix recipe Oct 01 14:48:27 I'll let you know when it is complete & functioning Oct 01 14:48:59 sakoman: thanks for the info. I won't be able to work on it for a bit too, but do let me know your progress Oct 01 14:49:32 scruggs: will do -- ttyl Oct 01 14:52:24 sakoman: thx - later Oct 01 15:10:16 Anyone have any particular interest in the orinoco tools in OE? Oct 01 15:13:01 mwester: yeah, symbol spectrum24 wifi card here Oct 01 15:17:18 Ok. The recipe for the firmware download is sub-optimal, and seems it gets built with OM builds. If someone is actively using and developing that, I'll leave it alone. Otherwise I'll put it on my list of things to look into. Oct 01 15:28:49 lrg: another mail on the buildroot list about broken alsalibs :) Oct 01 15:30:12 koen: funnily enough, I'm just speaking to someone in #alsa-soc now about using OE for his builds ;) Oct 01 15:30:31 koen: alsa-lib issues... Oct 01 15:30:58 lrg: the only problem in OE with alsa libs is missing atomic ops support for avr32 Oct 01 15:32:43 koen, XorA: still no fins of blackness yet :( Oct 01 15:40:03 hello all. wondering if someone can explain something weird I am seeing with configure. Oct 01 15:40:12 koen: wow that email about GPRS really works Oct 01 15:40:22 If I do a rebuild of gqview (which builds fine) it shows pkg-config using the one for arm as it should. Oct 01 15:40:26 koen: now how to build that into udev Oct 01 15:40:38 When I do a rebuild of rhythmbox, it uses pkg-config from i686 instead of arm. Oct 01 15:40:38 ~lart buildroot Oct 01 15:40:39 * ibot whips out his power stapler and staples buildroot's foot to the floor Oct 01 15:40:55 ~lart gumstix for not pushing buildroot patches upstream Oct 01 15:40:55 * ibot steals gumstix's mojo for not pushing buildroot patches upstream Oct 01 15:41:07 can anyone explain that? anything that may not be setup correctly or things to check? Oct 01 15:42:50 the same thing happens with a custom recipe where I am trying to build gtkpod using libgpod. even though there is a libgpod-1.0.pc file, configure doens't see it when gtkpod is built Oct 01 15:42:58 and I don't understand how to tell it which one to use Oct 01 15:43:23 inherit pkgconfig Oct 01 15:43:54 * Crofton forgets again if that is to build pkgconfig pkgs or for packages that use pkgconfig Oct 01 15:45:09 Crofton: I recall it places pkgconfig data in staging, but there may be benefits for packages that use pkgconfig Oct 01 15:45:38 http://www.openembedded.org/user-manual&dpage=pkgconfig_class Oct 01 15:45:55 ah thannks Oct 01 15:46:29 hmm, is anyone inclined to push my patch to make termcap smaller? http://bugs.openembedded.org/show_bug.cgi?id=3075 Oct 01 15:47:26 Henryk: on my TODO for tomorrow Oct 01 15:47:51 ah, fine, thanks Oct 01 15:50:19 * cbrake ponders why mtd-utils is not generating any packages in my glibc build, but ucblic build works fine -- same OE metadata ... Oct 01 15:52:57 cbrake: crofton: well, it uses pkgconfig, it just uses the wrong one. so, by inheriting it, will it force it to use the arm version or just use it...I'll give it a shot, but it's not like it isn't using pkgconfig...just using a different one Oct 01 15:56:13 03koen 07org.oe.dev * re7fcaa38... 10/ (1 conf/machine/er0100.conf): er0100: commit it, this has been sitting in my repo for too long Oct 01 15:56:19 03koen 07org.oe.dev * rf68cd62e... 10/ (1 packages/dbus/dbus.inc packages/dbus/dbus_1.0.2.bb): dbus: include uidgen as well Oct 01 15:56:23 03koen 07org.oe.dev * r1401e622... 10/ (10 files in 3 dirs): iwmmxt machines: add 'iwmmxt' as machine feature Oct 01 15:56:29 03koen 07org.oe.dev * rec037fe4... 10/ (1 packages/mplayer/mplayer_svn.bb): Oct 01 15:56:30 mplayer svn: use iwmmxt when appropriate MACHINE_FEATURE has been set Oct 01 15:56:32 [15:04] XorA: koen: good for me, saves adding a bazilion machines Oct 01 15:56:36 03koen 07org.oe.dev * rff780297... 10/ (1 conf/machine/er0100.conf): er0100: catch up with include renaming Oct 01 15:56:39 03koen 07org.oe.dev * r5adbd160... 10/ (1 packages/gstreamer/gst-meta-base_0.10.bb): gst-meta-base: gst-plugins-bad is needed for ivorbis Oct 01 15:56:55 christopher: /classes/pkgconfig.bbclass Oct 01 15:58:43 christopher: looks like about all it does is copy *.pc files to staging for packages provide pkgconfig data Oct 01 15:59:18 cbrake: ok. well, it did get further and got through configure for gtkpod so that is a step in the right direction! I think I'm still missing some dependencies Oct 01 17:03:41 cbrake: ok, I have a working mono recipe, it's kind of ugly though. I don't like they way I'm doing dependencies Oct 01 17:04:05 Henryk: excellent -- well, its a start and better than what is in there now :-) Oct 01 17:07:05 Maybe I'll ask for some kind of policy decision her: What I have is a lot (I'm talking in the region of 50's) of packages which each contain (provider) some number of assemblies and reference (depend on) some other assemblies. For example the package libmono-accessibility1.0-cil provides "Accessibility 1.0.5000.0" and depends on "mscorlib 1.0.5000.0" (which would be contained in libmono-corlib1.0-cil) Oct 01 17:08:15 I'd like to use virtual dependencies to accurately convey this information: E.g. libmono-accessibility1.0-cil would RPROVIDES "virtual/cli-Accessibility-1.0.5000.0" and RDEPENDS on "vritual/cli-mscorlib-1.0.5000.0" Oct 01 17:08:21 Would that be acceptable? Oct 01 17:09:01 s/her/here/, s/provider/provides/ Oct 01 17:10:37 are virtuals really necessasry? Oct 01 17:10:52 are there going to be more providers for virtual/cli-Accessibility-1.0.5000.0 ? Oct 01 17:11:17 and if so, how is that going to be handled at runtime? Oct 01 17:11:34 koen: no, but there's no really good way to map the assembly name "Accessibility 1.0.5000.0" to the package libmono-accessibility1.0-cil Oct 01 17:12:10 (e.g. "System.Drawing.Design 1.0.5000.0" is in libmono-winforms1.0-cil Oct 01 17:12:38 why does it need to get mapped? Oct 01 17:13:30 Henryk: nice to see that including dotnet is even more fun than supporting java. I am planning a full java toolchain for OE and have similar issue than what you are presenting here. Oct 01 17:13:35 koen: because I want to (need to) express a ipkg dependency, so that the necessary packages get installed Oct 01 17:14:48 so why again do you need virtuals? Oct 01 17:15:26 Henryk: you can use "cli-Accessibility-1.0.5000.0" as RPROVIDES, too Oct 01 17:15:45 koen: I don't. I just thought that would be the most appropriate way. E.g. RPROVIDES "mono-cli/Accessibility-1.0.5000.0" or something would be fine with me too Oct 01 17:15:46 RDEPENDS_libmono-accessibility1.0-cil = "libmono-corlib1.0-cil" whould solve your problems as well Oct 01 17:16:32 koen: except that I only get the information on which package provides and depends on what other package only at compile time Oct 01 17:16:37 of you *know* that libmomo-access needs corlib, why do you need to invent rprovides around that? Oct 01 17:17:19 Henryk: do_split_packages has the same problem and is able to solve it Oct 01 17:17:58 no need to pollute OE and ipkg namespaces when all you need is a bit of python Oct 01 17:18:12 koen: it's more complicated than that Oct 01 17:18:20 not from your description Oct 01 17:18:36 koen: there are two points I need to add dependencies: While compiling the mono recipe (which will create a huge number of packages each containing one .dll) and then setting the dependencies on these. That I can do (and have done) with Python. Oct 01 17:19:19 s/compiling/do_package/ Oct 01 17:19:20 The other point is when compiling external packages (e.g. gtk-sharp) outside of the main mono recipe. These also will depend on other assemblies (which I only know by their assembly name, not by their package name) Oct 01 17:20:15 So you could just dump that info in staging and have dependant recipes use that, right? Oct 01 17:20:24 OE does that for shlibs already Oct 01 17:20:36 koen: that would be another way, yes Oct 01 17:20:49 without need virtual/libfoo.so everywhere Oct 01 17:20:55 needing* Oct 01 17:21:53 I considered that for a short moment but then thought it would not be clean enough. But if you say that's the way I should be going then that's all the better Oct 01 17:22:34 using virtuals and/or RPROVIDES means something is either *very* wrong or that something was *very, very* wrong in the past and we need to provide bugcompatibilty Oct 01 17:22:50 (from long OE experience) Oct 01 17:24:09 Henryk: virtuals seems very attractive (especially to gentoo people missing USE flags), but it is a nightmare at runtime when installing stuff Oct 01 17:24:52 <-- gentoo user :-) Oct 01 17:25:15 koen: okay, I'll look into how shlibs does things. Oct 01 17:26:27 Henryk: it's possible that using virtuals *is* the way to go for mono stuff, but let's look at other options first :) Oct 01 17:33:48 dbus-glib_0.74 fail to compile --> configure: error: cannot run test program while cross compiling Oct 01 17:38:17 gremlin[it]: uclibc? Oct 01 17:40:03 yes Oct 01 17:40:17 previous version compiled fine ... Oct 01 17:40:32 the commit I just made should solve it Oct 01 17:41:32 okey koen, i go to dinner and try after ... :) Oct 01 17:41:41 I applied for a degree and crashed the internet in our building Oct 01 17:57:26 03koen 07org.oe.dev * ra2075ec1... 10/ (5 files in 2 dirs): site: commit some local changes Oct 01 18:02:19 Crofton|home, gotta love that campus IT (-; Oct 01 18:40:25 koen, dbus perfect now! Oct 01 18:42:41 cbrake: Right now, I am past the configure stage, but compilation of gtkpod fails when it can't find /libxml/xmlversion.h and I have libxml2 as a dependency but I don't see the includes on the compile line. I also added in --with-libxml2=yes to the configure line but still can't find it. Oct 01 18:42:57 is there a better way to do that? I've looked at other bb files including xml and they don't do much in there Oct 01 18:43:05 some do configure statements, some don't Oct 01 18:43:16 but the -I include statement is not there for the header files Oct 01 18:47:27 christopher: it is difficult to say how to fix this without debugging the package build -- it is likely something wrong with the gtkpod build files Oct 01 18:48:18 christopher: adding a BB dependency does not do anything but make sure the package is built first. It is up to the actuall sw being build to set up include paths, etc. Oct 01 18:49:19 any opinions if we should take TARGET_OS out of local.conf since it is typically set by a DISTRO.conf? Oct 01 18:49:35 cbrake: ok. I was thinking that you could pass info into configure to force the build to know to use xml or whatever. I need to be able to pass in the directory where those includes are. I'll look for a previous example Oct 01 18:50:10 christopher: it is probably wrapped up in autotools magic -- this is where things really get fun :-\ Oct 01 18:52:57 christopher, have you had a look at the output of ./configure --help ? you might be able to specify a specific path for libxml there ? Oct 01 18:56:58 gilligan__ : I did. it only has something about if you don't want the package or if you do, you can specify something like --with-libxml=yes Oct 01 18:57:13 and I have tried that, but it still can't find xmlversion.h (and a couple of other ones Oct 01 18:57:47 christopher, but you are sure that they have already been installed at that point? Oct 01 18:59:11 well, I did do the bitbake libxml2 to be sure it had been built. it is also part of the depends statement Oct 01 18:59:19 is there something else that I would need to do ? Oct 01 19:00:07 maybe look in the configure.ac file of gtkpod and look for the line where it checks for libxml .. most probably via pkg_config Oct 01 19:00:19 oh, or.. hold on.. does the ./configure fail at all ? Oct 01 19:01:29 or does it fail during the actual compilation? Oct 01 19:01:31 gilligan__ : it passes the configure. Oct 01 19:01:37 just fails the actual compilation Oct 01 19:01:46 because it can't find libxml/xmlversion.h Oct 01 19:02:09 it was failing configure before not finding libgpod but that works now with inheritting pkgconfig Oct 01 19:04:17 I don't know how the bootstrapping/fakeroot/chroot/whatever is set up .. but maybe someone from the channel who knows can tell you how to check if the file is actually installed or not Oct 01 19:05:26 or have a look at the configure.ac of gtkpod and then fgrep -R through the packages dir to see if there are other packages requiring libxml2 and how they deal with it Oct 01 19:06:21 dinner time Oct 01 19:06:21 What was the PATH_MAX fix for procps? rebuilding procps alone doesn't do it Oct 01 19:08:18 hi Oct 01 19:12:25 gilligan__: yeah, I was trying that and trying to see if any other package did anything special but I didn't see it Oct 01 19:25:25 wtf is libmudflap? Oct 01 19:25:53 what's wtf? Oct 01 19:26:00 um Oct 01 19:26:11 what the f*ck :) Oct 01 19:26:19 xD Oct 01 19:26:35 mud flap is something you have on a big pick up truck Oct 01 19:27:09 hi all Oct 01 19:27:17 hi flo_lap Oct 01 19:31:06 XorA, I wonder if you and vivijim are on the same flight Oct 01 19:35:08 XorA: Crofton my flight is from Frankfurt (11:40) to Berlin (12:45) Oct 01 19:35:15 vivijim: yes thats mine Oct 01 19:35:25 XorA: cool! Oct 01 19:35:43 * Crofton makes a note in case he misses his flight from Frankfurt to Berlin Oct 01 19:36:02 what time do you guys arrive in Frankfurt? Oct 01 19:36:30 I will arrive at Frankfurt at 10:50 Oct 01 19:36:33 from lisboa Oct 01 19:37:36 I arrive at 0600 and leave at 0855 Oct 01 19:40:29 weather forecast is good in Philadelphia for Thursday Oct 01 19:45:47 next year we can OEDEM in the US, by then us usd based people won't be able to afford to leave the country .... Oct 01 19:49:03 Crofton hehe ;) Oct 01 19:54:33 polyonymous: hmm, the procps patch itself doesn't solve the issue with PATH_MAX Oct 01 19:54:50 (with me anyway) Oct 01 19:54:55 I'm not sure how procps is supposed to actually compile Oct 01 19:55:21 PATH_MAX is not in limits.h, only linux/limits.h Oct 01 19:55:35 and limits.h does not include the linux one Oct 01 19:57:34 likewise, meaning it still fails for you? Oct 01 19:58:21 polyonymous|note: well, I now notice that your patch isn't applied during bitbake patch, let me investigate Oct 01 19:58:37 likewise, perhaps you need -c rebuild? Oct 01 19:59:08 likewise, I didn't bump PR, because compile fixes definitely do not justify reinstall of the package. Oct 01 19:59:09 polyonymous|note: just did pull, up, -c clean procps and procps Oct 01 19:59:15 hmm... Oct 01 19:59:20 03cbrake 07org.oe.dev * r70743353... 10/ (1 packages/mtd/mtd-utils-tests_1.0.0+git.bb): Oct 01 19:59:20 mtd-utils-tests: add package Oct 01 19:59:20 It may make sense to add this to the main package, but because Oct 01 19:59:20 so few people will use it I did not want to risk breakage at this Oct 01 19:59:20 time. Oct 01 19:59:27 03crofton 07org.oe.dev * r7b3eb3f5... 10/ (7 files in 3 dirs): linux : Update patches for gumstix from gumstix buildroot. Oct 01 19:59:29 polyonymous|note: let me just take a minute at this Oct 01 19:59:36 Weird. It did apply for me, Crofton and dcordes. Oct 01 19:59:39 likewise, sure, take your time. Oct 01 20:04:51 polyonymous|note: never mind, local problem. (BBFILES still pointed at a different checkout) Oct 01 20:05:04 ah ok :) Oct 01 20:07:14 polyonymous|note: sigh, that means yesterdays head build wasn't really that Oct 01 20:07:16 :-) Oct 01 20:07:36 Yeah, just thought how many other changes you missed :) Oct 01 20:09:12 polyonymous|note: I even fixed bugs against the old checkout :-) Glad someone else didn't check them in yet :-) Oct 01 20:09:25 heh Oct 01 20:09:30 heh, that too :) Oct 01 20:09:40 * Crofton still wonders how procps builds Oct 01 20:10:02 Crofton, how did it build before the patch... that's interesting, indeed :) Oct 01 20:11:09 one day I will try on my desktop Oct 01 20:11:47 I doubt I will, even though I'm curious too. Let me know if you find out :) Oct 01 20:11:51 vivijim: were you at FOSDEM? Oct 01 20:14:48 XorA: unfortunately no... :( Oct 01 20:15:22 vivijim: ah, so not much chance of recognition at Frankfurt then :-) Oct 01 20:15:29 vivijim, are you Nokia? Oct 01 20:19:40 XorA: :( it will be difficult... Oct 01 20:20:02 vivijim: Im 5'11'' blond hair, will be weating full length leather coat and New Rock boots Oct 01 20:20:21 vivijim: and all dressed in black Oct 01 20:20:51 Crofton I'm from Nokia's Institute (INdT) that is a researcher center with founds of Brazilian government and Nokia... Oct 01 20:20:55 Crofton: procps may have build previously because of a different version of the C lib headers were installed? Oct 01 20:21:06 maybe Oct 01 20:21:26 mwester, have had libc change recently? Oct 01 20:21:28 vivijim, as long as you had nothing to do with Nokia buying Navteq :) Oct 01 20:21:38 mwester, I haven't checked, but people say no? Oct 01 20:22:27 XorA: I'll be using a black t-shirt too, nike shoes and jeans pens pants... Oct 01 20:22:40 vivijim: I might be more obvious and holding a Neo1973 as well Oct 01 20:23:48 Crofton navteq is nice, but I don't believe that it works on Brasil :( Oct 01 20:23:50 http://www.flickr.com/photos/32615155@N00/405228464/in/set-72157594561002629/ Oct 01 20:23:56 polyonymous|note: I thought I saw something about a new set of kernel glibc headers or some such on one of the recent OE syncs... Oct 01 20:23:57 xora is on the right Oct 01 20:24:45 mwester, well, that would justify it... Oct 01 20:25:03 Crofton now I got: "all in black" of XorA :) Oct 01 20:25:20 vivijim: heh heh, yeah thats an accurate picture Oct 01 20:25:34 anyway offski for a bit Oct 01 20:25:48 :) Oct 01 20:30:49 I'll probably shine by not being there. Oct 01 20:31:13 (last guy on the photo - do not walk on red) Oct 01 20:57:41 http://www.flickr.com/photos/32615155@N00/405228464/in/set-72157594561002629/ now has a few more notes :) Oct 01 20:58:13 angstrom@serenity:~/website/repo$ php sort.php Oct 01 20:58:13 Totals: 10156 files seen, 5744 files already present in feeds, 4327 need to get added Oct 01 21:09:03 I've seen that angstrom-2008.1 seems not to choose kdrive as PREFERRED_PROVIDER_xserver. Is this intentional? Oct 01 21:13:33 sirfred: somewhere along the line xserver became virtual/xserver and the machine.conf files didnt get updated Oct 01 21:13:45 sirfred: as far as I can see anyway Oct 01 21:13:57 sirfred: I could never quite decide which was right Oct 01 21:15:22 XorA: PREFERRED_PROVIDER_virtual/xserver is not set to kdrive neither Oct 01 21:15:32 XorA: How about your new job? Oct 01 21:15:41 sirfred: I set it locally in zaurus-2.6.inc Oct 01 21:15:55 sirfred: so far so good :-) OpenMoko BTW Oct 01 21:16:08 XorA: Great Oct 01 21:16:25 PREFERRED_PROVIDER_virtual/xserver = "xserver-kdrive" Oct 01 21:16:49 XorA: That's not what the .dev repo file has, true? Oct 01 21:17:01 sirfred: no, thats what I locally have Oct 01 21:17:05 XorA: OK. Oct 01 21:17:24 sirfred: I havent had time to figure out by looking at changelogs what actually happened Oct 01 21:18:11 sirfred: also pH5 told me you need to add -mouse tslib for kdrive 1.4 to get ts working Oct 01 21:18:12 I noticed it only because xserver-xorg failed to compile. I still didn't figure out the exact reason. It seems that the XID typedef is not included while compiling some render sturff Oct 01 21:19:01 XorA: Good tip. Thanks. Oct 01 21:19:02 sirfred: same way I noticed it Oct 01 21:20:12 XorA: My first thought is that perhaps the XID is defined in some arch-dependant fashion, and remained undefined, but while thinking about it, I noticed that what I was compiling was xorg-xserver and not kdrive. Oct 01 21:20:55 sirfred: all hurdles to pass, then they will release 1.5 with different breakages :-D Oct 01 21:21:03 anyway time for sleep Oct 01 21:21:12 XorA|gone: night Oct 01 21:25:28 gn Oct 01 22:00:06 Good night Oct 02 01:55:35 Hmm, what's the best way to insert myself into PACKAGEFUNCS? Oct 02 01:57:02 E.g. I have a mono.bbclass and a mono_do_clilibs (which does something similar to package_do_shlibs) and want to have that called for all recipes inheriting mono after populate_packages but before read_shlibdeps Oct 02 01:57:22 Is it ok for me to just change package.bbclass? **** ENDING LOGGING AT Tue Oct 02 02:59:57 2007