**** BEGIN LOGGING AT Wed Sep 07 02:59:57 2011 **** BEGIN LOGGING AT Wed Sep 07 04:07:21 2011 Sep 07 07:09:42 good morning Sep 07 09:15:45 ping JaMa|off Sep 07 09:15:48 ping JaMa|Off Sep 07 09:16:14 pong Sep 07 09:21:44 hi i tried your kexeboot 3.0 image . i booted the bzimage with kexeboot. but i ends up in a black screen with a dot in the bottom left corner Sep 07 09:25:57 pwgen: linux-kexecboot works for me fine Sep 07 09:26:17 pwgen: but linux-3.1.0-rc5 fails to boot Sep 07 09:26:46 maybe because io wrong cmdline parameters ( did not change boot.cfg ) Sep 07 09:27:36 do you have the kernel cmdline that u-boot gives to the kexecboot image Sep 07 09:27:49 I'll rebuild it with debug enabled, but cannot test it now Sep 07 09:27:58 I don't have u-boot on my spitz Sep 07 09:39:38 pwgen: try the one here http://projects.linuxtogo.org/frs/?group_id=55 Sep 07 09:40:16 this is ok on spitz, check the logs in System Menu Sep 07 10:04:10 ant_work . the linuxtogo image seems to boot ... Sep 07 10:06:47 this was 2.6.39.3 Sep 07 10:08:23 my 2.6.26 kexecboot coult boot the 2.6.39.3 image ( the jffs2 ist not quite well because i use a modified jffs layout ( only one root part ) Sep 07 10:11:19 on newer kernels there seems to be a problem with charging led . charger is connected but led is off, i have to reconnect the charger to get the led on Sep 07 10:22:38 yes, I've seen this. It is disconnetted during boot Sep 07 10:23:48 with 2.6.26 you will only miss CF boot for new kernels (/dev/hd*->/dev/sd*) Sep 07 10:24:48 about jffs2 layout, u-boot should just pass the right mtdparts to kexecboot Sep 07 11:18:01 is there a way to burn the newer kexecboot images to nand without the updater.sh magic ? Sep 07 11:27:35 pwgen: you do 3 line nandlogical magic from updater.sh by hand :) Sep 07 11:36:03 JaMa|Off: you have no u-boot ? how are you booting ? direct into kexecboot ? Sep 07 11:37:18 pwgen: default Angel bootloader Sep 07 11:37:39 pwgen: iirc you have a broken nand, firts sectors, isn't? Sep 07 11:37:50 .. i thing i have missed something. what is Angel Sep 07 11:38:07 this is on my c-1000 the c-3200 works fine . until now Sep 07 11:38:22 I see Sep 07 11:38:59 Angel is standard/old Zaurus bootloader Sep 07 14:45:50 lumag_work: fyi RP_ says please submit a patch for icecc.bbclass if you have a working version for oe-core :) Sep 07 15:06:11 hello people Sep 07 15:12:13 i need help Sep 07 15:12:32 i installed XML::Parser Sep 07 15:13:09 but when i compile my bitbake,it says 'xml module is missing' Sep 07 16:28:40 pwgen: looks like 3.1.0-rc5 booting is rather rootfs issue http://paste.pocoo.org/show/471895/ Sep 07 17:16:21 I'll submit a patch when I'm more or less confident with it. Let me play with it a little bit more. Sep 07 18:09:50 hi Sep 07 18:10:41 how can I avoid building *-native stuff ? Sep 07 18:11:00 ASSUME_PROVIDED += "*-native" ? Sep 07 18:14:23 btw, is it enforced DISTRO or it's OE's default ? Sep 07 18:29:54 oh .. I probaly have too many questions already, but my main question remains Sep 07 18:30:12 what makes OE better chois VS Gentoo portage ? Sep 07 18:54:17 hi, using oe-core with angstrom setup scripts, I compiled a qt4e-demo-image for qemuarm and got a workign result. Then I added a layer containing an armv5 machine (arm926ejs based as qemuarm) and launched the same build. I was expecting a very fast build as only the kernel is different, but in the end I see that all the tasks are executed again and the packages are compiled one more time. Is that an expected behaviour ? Is that because I a Sep 07 18:55:10 ericben: Well, it depends Sep 07 18:55:28 Use bitbake-diffsigs, iirc, to see what the signatures are of one or two of the things, before and after Sep 07 19:00:33 Tartarus: thanks, ok I'll try that once the build finish ... qt4-embedded is quite long to build :-( Sep 07 19:00:56 yes :( Sep 07 19:01:01 qt4-* really Sep 07 19:01:22 hi Sep 07 19:01:41 especially when it comes to linking webkit where 7 cores are sleeping while one is eating as much memory as possible :-) Sep 07 19:01:43 what was said about qt4? Sep 07 19:01:46 hi GNUtoo|laptop Sep 07 19:02:44 GNUtoo|laptop: I'm trying ot understand why swithing from qemuarm to our arm926ejs machine leads to a nw compilation of most packages (including the longest to build qt4) Sep 07 19:03:06 ah ok Sep 07 19:03:29 GNUtoo|laptop: maybe we have one bad setting somewhere which would explain why your first build was not working on the target Sep 07 19:03:51 possible Sep 07 19:04:34 Tartarus: I saw patches from you on u-boot ML : have you switched from Mentor to TI ? Sep 07 19:04:45 ericben: yeah Sep 07 19:05:29 good, you are working on omap sw support ? Sep 07 19:10:05 I'm in one of the Linux groups, yes :) Sep 07 19:10:19 It's a little more complicated that just "omap", I've learned once I've gotten on the inside Sep 07 19:10:50 * Tartarus is over with chase and near koen/denix, roughly Sep 07 19:12:08 if you can help TI to update DM3730 dsp support to newer kernel that would be a good thing ;-) Sep 07 19:15:01 Tartarus: ok I tried bitbake-diffsigs and find differences for example of flex Sep 07 19:15:44 OK, pastebin the diffs? Sep 07 19:15:47 It might be a bug Sep 07 19:15:58 And I'm not on the DM side of things, sorry ;) Sep 07 19:16:25 Tartarus: was a bad example. I met a dependency problezm on flex so added bison as a DEPEND (will provide a patch for this soon for coments) Sep 07 19:16:58 but bison is a good example Sep 07 19:17:01 also try a diffsigs between with and without the extra layer Sep 07 19:17:10 To see what it didn't reuse what you expectedc Sep 07 19:18:10 Tartarus: http://pastebin.com/st5td74e Sep 07 19:18:26 it seems only the machine changed Sep 07 19:19:42 Hi all! I'm working with a demo-image recipe and get lots of "NOTE: multiple providers are available for runtime kernel-module-" when I bitbake it. What am I missing? Sep 07 19:20:02 The demo image is from technexion and not in the official repos Sep 07 19:21:26 http://pastebin.com/qK8R8Bxw <-- technexion-demo-image Sep 07 19:22:11 what's diffsigs? Sep 07 19:22:13 a new tool? Sep 07 19:22:24 GNUtoo|laptop: https://wiki.yoctoproject.org/wiki/Enable_sstate_cache Sep 07 19:22:47 ok thanks Sep 07 19:23:58 I would like to define linux-omap-2.6 as PREFERRED_PROVIDER for all my modules. How do I do that? Sep 07 19:24:18 modules+kernel of course Sep 07 19:24:19 risca, prefer that kernel Sep 07 19:24:33 look in your machine config Sep 07 19:25:05 In my machine config I have PREFERRED_PROVIDER_virtual/kernel = "linux-omap-psp" Sep 07 19:25:18 Isn't that enough? Sep 07 19:26:09 that prefer the psp kernel Sep 07 19:26:18 prefer the omap one Sep 07 19:26:34 and be sure it is compatible with your machine in the kernel recipe Sep 07 19:26:55 ericben: that seems odd, go post over in #yocto and bug RP or sgw :) Sep 07 19:27:19 Tartarus: ok will do Sep 07 19:27:22 thanks Sep 07 19:28:01 What is the difference between regular omap and omap-psp? Sep 07 19:28:20 psp as in playstation portable? Sep 07 19:28:55 lol Sep 07 19:28:55 I don't think so Sep 07 19:29:02 Not me either ^^ Sep 07 19:29:20 the portable playstation is mips I think Sep 07 19:29:22 not sure tough Sep 07 19:29:28 and lack a touchscreen Sep 07 19:29:31 anyway Sep 07 19:29:36 I believe omap-psp is the correct kernel Sep 07 19:29:51 I think the psp version has everything, including the DSP Sep 07 19:29:52 The machine configuration is downloaded from technexion directly Sep 07 19:30:46 Technexion supplied a complete openembedded tree, and I'm extracting the important part to be used in an updated tree Sep 07 19:31:01 s/part/parts/ Sep 07 19:31:38 ok Sep 07 19:33:07 Well, the build starts without me setting PREFERRED_PROVIDER, but I don't know if it will work Sep 07 19:33:22 On the machine that is Sep 07 19:33:26 I sure hope so Sep 07 19:34:36 The original technexion-demo-image basically only contained ANGSTROM_EXTRA_INSTALL := "lots of programs". Sep 07 19:35:36 I used task-beagleboard-demo.bb as a template and added a task-technexion-demo.bb, and rewrote the technexion-demo-image recipe Sep 07 19:35:55 I hope it works ^^ Sep 07 20:08:13 Tartarus: could having PREFERRED_PROVIDER_virtual/kernel = "linux" in our machine and PREFERRED_PROVIDER_virtual/kernel ?= "linux-yocto" in qemuarm explain my problem ? Sep 07 20:11:44 bluelightning, RP_, regarding icecc. I want to test it a little bit more and then I'll push it to oe-core (probably at the weekend). Currently icecc + icecc-create-env is accessible at github.com/lumag/meta-lumag Sep 07 20:18:28 hi gnutoo Sep 07 20:18:42 hi Sep 07 20:18:48 https://codeshite.wordpress.com/2011/08/15/how-to-become-a-great-software-developer/ is amusing Sep 07 20:23:50 bye Sep 07 20:23:50 kergoth, unfortunately it seems that #12 is applicable more and more to softdevs Sep 07 20:23:57 heh, indee Sep 07 20:23:58 d Sep 07 20:26:17 he I am always ask uncle kergoth Sep 07 20:29:55 lumag: ok, sounds good, thanks Sep 07 20:30:51 Damn. do_fetch for udev fails, as git.kernel.org isn't working :( Sep 07 20:36:29 hi all Sep 07 20:37:08 lumag: yeah, util-linux too. Sep 07 20:37:18 hi pb_ Sep 07 20:37:33 I found a pair of udev mirrors, verifying tags now. Sep 07 20:37:36 bit of a shame that the kernel h4x0rs are taking so long to get their machines working again. Sep 07 20:39:49 kergoth: is srctree.bbclass expected to work with a recent oe-core? I tried pulling the one from oe.dev into my tree and had some annoying issues to do with stamp file generation. Sep 07 20:40:09 also some slightly irritating things to do with NoneType not being iterable, but I don't think that's an oe-core issue specifically :-} Sep 07 20:40:36 for git I've found two mirrors: http://mirrors.arsc.edu/kernel/scm/linux/hotplug/udev.git/ and http://ftp.twaren.net/Unix/Kernel/scm/linux/hotplug/udev.git/ . Hope this helps Sep 07 20:44:48 ericben: I would hope different kernels wouldn't invalidate the whole cache... Sep 07 20:48:07 pb_: Hmm, don't think I've tested it out there. Will do so. Of course, the main thing it does is combines the main tasks to avoid interleaving amongst target/native/cross/etc variants. If there aren't any variants, setting S to the path and emptying SRC_URI is sufficient. alternatively, if separating objdir with B works for that recipe, that works to avoid the issue as well. I'll take a look though Sep 07 20:48:36 should probably have a variable control the task mangling behavior, to opt in or opt out of it Sep 07 22:18:44 I had been using Gentoo for quite a long time now Sep 07 22:19:04 and bitbake looks quite streight forward Sep 07 22:19:16 though I am still unsure of a few things Sep 07 22:19:32 I see there is quite different approach in OE/bitbake Sep 07 22:19:54 how packages are staged ..etc Sep 07 22:20:09 but what do you call the USE flags ? Sep 07 22:20:52 or there would be a different bbfile for each different flavour ? Sep 07 22:25:10 errordeveloper: we don't have a direct equivalent of USE, but there are some parallels for a few aspects; DISTRO_FEATURES is one example Sep 07 22:25:10 is line like 'PACKAGES += "${PN}-symlinks"' doing the job of USE flags essentially ? Sep 07 22:25:28 ok .. Sep 07 22:25:52 yes, we do tend to err on the side of enabling lots of features and then splitting the result into packages that let you avoid the feature at runtime Sep 07 22:26:10 (by not installing the appropriate package) Sep 07 22:26:23 ok .. I sort of see Sep 07 22:26:43 ok, I'm looking into adding openpgm.bb Sep 07 22:27:17 and include a feature in zeromq.bb to compile with openpgm Sep 07 22:27:32 what's best to look at in this case as an example ? Sep 07 22:27:39 presumably it would be openpgm_5.1.115.bb Sep 07 22:27:57 or, which ever version you choose to provide the recipe for :) Sep 07 22:28:17 damn Sep 07 22:28:20 well, there are lots of examples; pick something simple Sep 07 22:28:22 I'm on the wrong repo Sep 07 22:28:36 openembedded.git is what I had been looking at for now Sep 07 22:29:06 yeah, that's OE, but it's sort of deprecated for new items Sep 07 22:29:23 it's a bit messy if all the repos, how on earth I'm meant to figure what's already avail Sep 07 22:29:49 I know, it's something we want to address Sep 07 22:30:04 http://www.openembedded.org/index.php/OpenEmbedded-Core Sep 07 22:30:15 ok Sep 07 22:30:17 this gives a basic explanation of the future direction Sep 07 22:30:56 though .. I still see that gentoo has quite a bit of ponetial, but it's missing just a few things that oe has Sep 07 22:31:15 that's what bugging me to be honest Sep 07 22:31:39 of course, look at freebsd and they have quite a bit potential also Sep 07 22:32:01 but let's not touch them for now, lol :) Sep 07 22:33:21 well, as you might be aware OE has its roots within Gentoo Sep 07 22:33:32 indexing would be great to have .. where did you pull openpgm_5.1.115.bb from ? Sep 07 22:33:34 although it is quite different in a lot of aspects Sep 07 22:33:48 errordeveloper: I didn't, that's an example filename that includes the version Sep 07 22:33:54 bluelightning: yeah, I kind hear that .. suppose those were early days Sep 07 22:34:21 now I do 'emerge -j10' and that tries to do up to 10 compiles if it can Sep 07 22:34:37 yeah, you can do parallel builds with OE as well Sep 07 22:34:40 and takes care of load avg too Sep 07 22:34:44 (or rather, bitbake does that work) Sep 07 22:35:01 it seems to be doing a few things, yeah .. :) Sep 07 22:35:33 it's not self-adjusting like that though, but newer kernels do a better job of keeping the machine usable even with aggressive bitbake parallelisation settings I find :) Sep 07 22:35:40 sure, on the other hand gentoo would need to be forked to get the things right for the embedded world - Sep 07 22:35:55 their policies and spirit overal are a bit dusty Sep 07 22:36:11 or whatever you wanna call it, just trying to be not too rude Sep 07 22:36:14 yeah, I know gentoo embedded has its adherents but I'm not convinced it's been especially successful on a large scale Sep 07 22:36:37 bugtracker is quite a nasty place to go for example Sep 07 22:37:15 yeah, crossdev is great attempt, but it's not quite going forward as I see Sep 07 22:37:40 hmm... well in OE we deprecated our bugtracker because there aren't the resources to manage it... but issues tend to be dealt with in a reasonably efficient manner via the mailing list Sep 07 22:38:00 and if you don't receive a reply in a day or two it's OK to ping :) Sep 07 22:38:11 yeh, who can be responsible for bugs in oe anyway Sep 07 22:38:17 well, maybe 2-3 days Sep 07 22:38:45 though gentoo tries to deal with nvidia/xorg compatibilities and similar things Sep 07 22:39:07 I'm a little biased but I'd like to think that the arrival of yocto has brought some focus on upgrading/fixing recipes that would otherwise go unmaintained Sep 07 22:39:36 s/yocto/the yocto project/ Sep 07 22:39:43 hm .. I was trying to figure out what exactly yocto guys are doing Sep 07 22:40:02 so seems like they trying to keep some core stuff tidy and provide Sep 07 22:40:08 a bunch of easy to use scripts Sep 07 22:40:31 I presume they send pull requests also Sep 07 22:40:35 yep, and some additional tools plus up-to-date documentation Sep 07 22:41:09 yes, all of us who work on the yocto project go through the same post-stuff-to-mailing-list procedure that everyone else does Sep 07 22:43:39 would be nice to have a simple search engine to find whether a bbfile exists for something specific Sep 07 22:44:14 I was gonna have a bit of a play with node.js .. so let's see if something can be done for that :) Sep 07 22:46:51 yeah, I guess there would be two parts, a backend that queries the appropriate repos and writes to a database, and the frontend that allows you to query the db Sep 07 22:49:56 you would need to have only .git dirs .. and I am assuming there must be something out there already anyway Sep 07 22:52:17 well, true, but I figure building an index of just the recipes would be quicker, and you could also extract the metadata from them and cache that in the db as well Sep 07 22:52:27 (description, homepage, etc.) Sep 07 22:53:12 there's nothing web-based like you describe, although I was discussing the very same idea just a few days ago on this channel :) Sep 07 22:54:51 well, see I thought to make something minimal witch will give you file listing and a the clone url Sep 07 22:55:18 yeah, and the description Sep 07 22:55:19 well, there is the angstrom web package querying system, but that's more from the package end rather than the recipe end Sep 07 22:56:39 so you would do 'find-me-a-bb regexp` Sep 07 22:57:19 with options like --print-clone-urls, --print-description, --just-get-single-file` Sep 07 22:57:39 and someone else could write the front with nice web ui Sep 07 22:59:11 so then you could also pipe --print-clone-urls and clone them to look at what's going on with it Sep 07 22:59:17 that might be a little simplistic with our recipe structure (e.g. bb files that refer to a common .inc), but yeah that would be the basic idea Sep 07 23:00:05 yeah, but then we can decide whether --what-it-depends-on is really needed Sep 07 23:00:10 and so on .. Sep 07 23:00:13 although I think it really has to be web-based to be of any value, otherwise you're expecting the user to have all the repos checked out Sep 07 23:00:52 true, though will need to find a person with decent ui skills and Sep 07 23:01:21 have host it somewhere too Sep 07 23:01:24 yes, that is what we lack right now, someone with the time and skills in web design & web programming Sep 07 23:01:35 hosting wouldn't be too much of an issue I wouldn't think Sep 07 23:01:51 if it's just a simple server, it can be hosted on a simple place Sep 07 23:02:28 oh, but may be just have a git repo which will have listing in it Sep 07 23:02:59 we're doing reasonably well within OE for hosting... could always use a few more boxes, but they are administered within the team at least Sep 07 23:03:02 and one could pulll that and run a script on it Sep 07 23:03:24 that may be silly though Sep 07 23:04:03 it's not a totally crazy idea :) Sep 07 23:04:32 so each repo would have a file .. and that's it Sep 07 23:05:10 yeah .. 100% crazy ideas don't exist Sep 07 23:05:26 well I don't think you could do it that way, it would have to be two scripts - one to gather the info and another that the user of the index repo could use to search it Sep 07 23:06:11 I'm convinced the web approach is the best one though, assuming there's someone available to implement it... Sep 07 23:08:05 maybe I could at least start the backend indexing code in the hope that someone could do the frontend :) Sep 07 23:08:36 I'm more of back-end-thinking person also Sep 07 23:12:17 bitbake can only do it as far as '-s' goes, right ? Sep 07 23:16:03 errordeveloper: yeah Sep 07 23:16:43 with some closer integration though there's the possibility to speed things up and update the index directly Sep 07 23:18:37 * bluelightning -> sleep Sep 07 23:18:41 cya Sep 07 23:18:58 errordeveloper: good talking with you :) **** ENDING LOGGING AT Thu Sep 08 02:59:57 2011