**** BEGIN LOGGING AT Wed Sep 02 02:59:58 2009 Sep 02 03:12:12 hi Sep 02 03:12:29 how can i remove beagle logo at angstrom boot seq Sep 02 03:12:34 fb logo Sep 02 03:13:19 is it in the top corner of the screen during boot? Sep 02 03:13:59 yes Sep 02 03:14:11 like baner Sep 02 03:14:15 banner Sep 02 03:14:23 that would likely either be in a patch, or in a path given through the defconfig Sep 02 03:14:29 of the beagleboard kernel Sep 02 03:15:12 need recompile? Sep 02 03:15:46 openembedded/recipes/linux/linux-omap/beagleboard/logo_linux_clut224.ppm Sep 02 03:15:51 yes its compiled into the kernel Sep 02 03:16:09 you could change that .ppm to something else, or give it a different path in defconfig Sep 02 03:17:21 for angstrom? Sep 02 03:18:11 if you defined your board/machine type as 'beagleboard', i'm assuming it uses the linux-omap kernel Sep 02 03:18:55 thanks alot Sep 02 03:21:03 yea np Sep 02 03:28:17 m4t , is it Sep 02 03:29:20 CONFIG_LOGO? Sep 02 03:36:22 can we change it without recompile Sep 02 03:36:38 just by renaming some file Sep 02 03:45:24 yes that sounds about right, let me check Sep 02 03:45:44 you must recompile the kernel Sep 02 03:47:51 also ./linux-omap_2.6.29.bb:SRC_URI_append_beagleboard = " file://logo_linux_clut224.ppm \ Sep 02 04:59:05 hello Sep 02 04:59:13 m trying to build maemo on OE Sep 02 04:59:39 but while trying to build it i'm getting an error in relation to C compiler Sep 02 05:00:04 it is not able to build most of the recipes Sep 02 05:00:17 like glibc Sep 02 05:00:22 can neone help me? Sep 02 05:02:34 which error ? Sep 02 05:16:41 03Angus Ainslie  07shr/import * ra16905f062 10openembedded.git/recipes/python/pyring_1.1.9.bb: pyring : new version 1.1.9 Sep 02 05:16:52 03Angus Ainslie  07shr/import * r2216e7c832 10openembedded.git/recipes/bt-configure/bt-configure_git.bb: bt-configure : fix SRCPV Sep 02 05:17:28 the error is that "C preprocessor "arm-angstrom-linux-gnueabi-gcc -E" fails sanity check' Sep 02 05:17:34 @crux Sep 02 05:17:52 and it is saying failed with 256 Sep 02 05:35:04 * * OE Bug 5317 has been created by kjomara(AT)scienceus.org Sep 02 05:35:06 * * Shell for CPAN module won't load, MakeMaker Config.PL not in @INC Sep 02 05:35:08 * * http://bugs.openembedded.org/show_bug.cgi?id=5317 Sep 02 05:40:39 [10:29] hello [10:29] m trying to build maemo on OE [10:29] but while trying to build it i'm getting an error in relation to C compiler [10:30] it is not able to build most of the recipes [10:30] like glibc [10:30] can neone help me? [10:32] which error ? [10:33] == jeaneus [n=orb@pool-173-71-43-116.dllstx.fios.verizon.net] has left #oe [] [10:46] Angus Ainslie m trying to build maemo on OE [10:29] but while trying to build it i'm getting an error in relation to C compiler [10:30] it is not able to build most of the recipes Sep 02 05:47:15 can neone help me Sep 02 05:47:25 the error is that "C preprocessor "arm-angstrom-linux-gnueabi-gcc -E" fails sanity check' Sep 02 05:47:31 and it is saying failed with 256 Sep 02 05:57:10 texas: which application is it compiling Sep 02 05:57:26 glibc Sep 02 05:57:28 texas: paste the config.log output of that app Sep 02 05:57:31 ah ok Sep 02 05:58:00 and after some 5000 tasks out of 9000 odd tasks the build is stopping Sep 02 05:59:08 ok sure post the config.log I have few mins Sep 02 06:02:28 khem it will take some time Sep 02 06:02:39 as i have run bitbake again Sep 02 06:02:51 are you there for sometime here? Sep 02 06:03:15 I will be off soon Sep 02 06:05:32 I am sure someone else will be there to help u later Sep 02 06:05:41 * khem ->sleeep() Sep 02 06:05:44 ohk Sep 02 06:44:05 mornk Sep 02 08:54:00 good morning Sep 02 08:56:29 can neone help me with the maemo build in oe Sep 02 08:57:04 m getting a prob relate to some C complier preprocesor Sep 02 09:02:05 florian: good morning Sep 02 09:11:45 ti_: oh... that's a little bit tricky :) you know the latest development in thei area? Sep 02 09:12:01 no Sep 02 09:12:05 @florian Sep 02 09:12:21 is it due to some toolchain problem Sep 02 09:12:21 ? Sep 02 09:23:58 ti_: the maemo stuff in oe is far from complete and quite old. but there was a gsoc project to get an up to date maemo into oe a short time ago Sep 02 09:49:01 @ florian : i'm using rkitri maemo_angstrom to build Sep 02 09:49:36 ti_: ok, very good... do you use it as an overlay? Sep 02 09:50:40 yaa Sep 02 09:50:43 @florian Sep 02 10:07:04 ru there....florian? Sep 02 10:18:09 ti_: yes... somewhat :) Sep 02 10:18:42 ti_: what's your target device and which component fails? Sep 02 10:20:45 target device is omap3430 Sep 02 10:20:59 glibc and gtk+ failing Sep 02 10:23:30 who needs glibc anyway.. Sep 02 10:28:07 good morning Sep 02 10:28:19 hey kirti Sep 02 10:28:51 hi ti_ Sep 02 10:28:54 there is still a problem with the build Sep 02 10:29:07 ti_: Anish ? Sep 02 10:29:13 hello florian Sep 02 10:29:15 yup Sep 02 10:29:17 ti_: hmm... strange, what's the error message you get? Sep 02 10:30:29 ti_: looks like I entered while you were in the middle of describing your problem - where are you stuck at ? Sep 02 10:30:40 should i paste my local.conf here so that you can review and tell me whether it is fine? Sep 02 10:31:29 because the way you told me to rebuild gcc cross and gcc initial Sep 02 10:31:36 clean and rebuild Sep 02 10:31:40 it is not working out Sep 02 10:32:01 because the problem wqas Sep 02 10:32:10 cC Compiler fails sanity check Sep 02 10:32:32 C preprocessor "arm-angstrom-linux-gnueabi-gcc -E" fails sanity check Sep 02 10:32:42 ah Sep 02 10:32:49 so can there be a problem with toolchain Sep 02 10:32:57 should i paste my local.conf here Sep 02 10:33:04 no Sep 02 10:33:07 ok Sep 02 10:33:20 then? Sep 02 10:33:27 ~pastebin Sep 02 10:33:42 in fact its not necessary to paste is Sep 02 10:33:52 ohk Sep 02 10:34:00 that's a problem recompiling glibc - no idea why this pappens Sep 02 10:34:12 ohk Sep 02 10:34:13 config.log? Sep 02 10:34:20 git log -M -C --find-copies-harder recipes/ti doesn't work :-\ Sep 02 10:34:24 any suggestions? Sep 02 10:35:05 if you clean glibc, glibc-initial, gcc-cross-intermediate it will build again Sep 02 10:35:25 ti_: I remember the stuff about you have to clean gcc-cross , gcc-cross-initial and so on - I gave you the same list which florian gave me - iirc, that had one thing missing. It worked for me when I added gcc-cross-initial Sep 02 10:35:35 florian: ^^ Sep 02 10:35:56 ah right Sep 02 10:36:03 forgot about this :-} Sep 02 10:36:13 :-) Sep 02 10:36:30 tardyp: iirc it was complainign about something missing but i'm not sure any more. Sep 02 10:36:48 but yes we should all move to eglibc or fix this issue some time Sep 02 10:37:49 florian: may I ask why eglibc ? Sep 02 10:39:34 rkirti: it should be more suitable for embedded and mobile... in fact the was not meant to be really serious - its a bug that needs to be fixed. Sep 02 10:41:33 eglibc == glibc for embedded crap ;-) Sep 02 10:41:59 haha Sep 02 10:50:36 morning Sep 02 10:50:51 tardyp: all Debian boxes with testing or unstable runs eglibc Sep 02 10:51:34 florian: will you release xserver-common 1.32 with all my patches applied? Sep 02 10:54:00 hrw: oh yes i should do this Sep 02 10:54:20 hrw: are days in poland as short as in germany? ;) Sep 02 10:55:25 florian: yes Sep 02 10:57:07 Unknown Event: ['bb.event.NoProvider', {'_runtime': False, '_data': None, '_item': 'virtual/arm-angstrom-linux-gnueabi-gcc-4.3.3-csl-arm-2007q3-53', 'pid': 0, 'data': None}] ERROR: Required build target 'maemo-angstrom-image' has no buildable providers. Missing or unbuildable dependency chain was: ['maemo-angstrom-image', 'task-base', 'task-boot', 'virtual/kernel', 'virtual/arm-angstrom-linux-gnueabi-gcc-4.3.3-csl-arm-2007q3-53'] Sep 02 10:57:18 how to solve this problem? Sep 02 10:59:16 looks like you try to use a cross gcc which is not available Sep 02 11:16:48 florian: ping ? Sep 02 11:17:18 rkirti: yep Sep 02 11:17:46 florian: http://www.fell.it/2009/06/20/bitbake-and-proxy/ Sep 02 11:18:02 florian: this seems like a REALLY hacky soln Sep 02 11:18:11 but its also probably the only way out Sep 02 11:18:26 does bitbake really remove env variables ? Sep 02 11:19:07 There is a .wgetrc... Sep 02 11:19:43 yes it is Sep 02 11:19:55 wgetrc should help Sep 02 11:20:06 it might be a good idea to add these to the whitelist Sep 02 11:20:20 iirc this is in bitbake.conf Sep 02 11:20:21 yes, I have wgetrc set to use my proxy Sep 02 11:20:23 but.. Sep 02 11:20:33 I have no clue why the auth fails Sep 02 11:20:47 I have cntlm running with the right auth settings Sep 02 11:21:06 (I am also tunnelling - no clue if that affects thing) Sep 02 11:21:08 *things Sep 02 11:23:08 if you run it yourself it works? Sep 02 11:23:45 right now, no. same 407 error Sep 02 11:31:39 nasty... unluckily my knowledge of proxy authentication is quite limited. Sep 02 11:40:56 MVL6 is now available Sep 02 11:49:06 yah, kergoth mentioned a few days ago that it was shipping Sep 02 11:55:41 I wonder will there be discounts for us Sep 02 12:11:08 hrw, consult for people using mvl6 Sep 02 12:42:01 mickeyl: good morning Sep 02 12:42:07 hi pb__ Sep 02 12:43:00 heh, oh dear Sep 02 12:43:52 apparantly Strato (one of the biggest german hosting providers) changed some DNS one week to early :D Sep 02 12:44:03 doh Sep 02 12:44:18 ;( Sep 02 12:44:25 how would this impact us? Sep 02 12:44:37 Crofton|work: linuxtogo.org machine is strato Sep 02 12:49:38 * XorA assumes florian will just fallback ltg to the other server :-) Sep 02 12:51:08 I'm trying... unluckily I don't have the account information yet. Sep 02 12:51:53 nilsf did the contract stuff and he is in a plane to .uk right now. Sep 02 12:51:53 Crofton|work: all our mail is going through ltg Sep 02 12:52:31 but oe.org ppoints at a number, not ip Sep 02 12:52:33 does someone have the old ltg ip around? Sep 02 12:53:12 true, we are lucky since we use another domain Sep 02 12:53:20 :) Sep 02 12:53:23 florian, looking Sep 02 12:53:36 all the other ltg projects are doomed though :) Sep 02 12:54:14 85.214.40.226 it seems Sep 02 12:54:32 yes this looks familiar :) Sep 02 12:54:36 many thanks Sep 02 12:54:38 85.214.40.226 Sep 02 12:54:46 pb_, I concur :) Sep 02 12:54:53 that is where our mail goes Sep 02 12:55:09 hello all: has any one ran into gst-plugins-good_0.10.15.bb failed from .dev branch? Sep 02 12:55:44 Crofton|work: in this case... there must be a way to fix this now Sep 02 12:55:54 i need to a minimal image with busybox, scp, which packages do i need to select Sep 02 12:56:14 the machine is still there of course Sep 02 12:56:26 actually, openembedded.org is still pointing to that old ip for me, though linuxtogo.org has changed. Sep 02 12:57:58 what could be the smallest foot print image one can have using oe Sep 02 12:58:15 rsv: that rather depends on what you want the image to do Sep 02 12:59:11 busybox and ssh/scp Sep 02 12:59:53 i try to select task-base ; but apparently it puts other stuffs too Sep 02 12:59:55 probably about 500k, maybe a bit less with some effort. Sep 02 13:00:39 pb_: 500K? Sep 02 13:00:40 I'm waiting for Nils to land (sent him mail and sms) and initiating a big sync so that I can get up mail on the new machine as fast as possible. Sep 02 13:01:06 florian, let me know when that happens so I can switch oe.org dns Sep 02 13:01:12 incredible, i need to find that out Sep 02 13:01:15 rsv: right, something like that. Sep 02 13:01:23 Crofton: will do so Sep 02 13:01:34 is there a uclibc based micro-image? Sep 02 13:01:51 that's a compressed size though, admittedly. probably closer to 800k before compression. Sep 02 13:02:37 Crofton: if oe mx still points to the old ip... shouldn't we be able to deliver mail without a valid ltg dns entry? Sep 02 13:03:22 we shoudl Sep 02 13:03:32 I think :) Sep 02 13:03:54 florian: yes, I would have thought so. oe.org doesn't actually have an mx record (just an A) but I still can't think of any particular reason why it would fail. Sep 02 13:03:57 mail is something i only have a very limited experience with Sep 02 13:04:04 heh Sep 02 13:04:08 of course, some other MXs will refuse to talk to serenity now because its forward and reverse DNS entries won't match Sep 02 13:04:15 oe.org is set up to use an A record to the ip Sep 02 13:04:18 not via MX Sep 02 13:04:29 ah right, that might be a problem Sep 02 13:04:32 we need to fix that one day Sep 02 13:04:57 it doesn't really matter, an A record is okay. Sep 02 13:05:16 it is annoying becasue we have to redirect the web traffic though Sep 02 13:05:36 ah, I see. Sep 02 13:05:51 well, you could go ahead and add an mx record with the same address now, it won't hurt anything. Sep 02 13:08:40 for some reason having an mx record didn't work Sep 02 13:08:49 we can look again after the move :) Sep 02 13:10:07 what did you put in the mx? you can't actually put an ip literal there directly, it needs to be a hostname. Sep 02 13:10:38 and I think the pointed-to thing might need to be an A record, not a CNAME, though I'm not sure about that. Sep 02 13:10:44 If necessary we can have have an oe mirror and additinal services on the new machien then... it is way more powerful Sep 02 13:12:00 ah yeah, cname not permitted in mx data. Sep 02 13:12:17 florian: fwiw, I did just receive a mail from oe-devel, so the list does appear to be working at the moment Sep 02 13:12:41 * florian watches data moving from Berlin to Nürnberg Sep 02 13:12:57 pb__: that's good news... Sep 02 13:15:22 pb_, I'll be out of town until next week, so if it works, I do not want to fiddle with things until I am back :) Sep 02 13:26:49 morning Sep 02 13:28:01 hi all Sep 02 13:28:49 is there anyone who uses qt-embedded with touchscreen? Sep 02 14:50:41 anyone else seeing this: Packaged contents of base-files-doc into /build/oe/oe-testing/build/micro-uclibc/tmp/deploy/ipk/cm-x270/base-files-doc_3.0.14-r88_cm-x270.ipk Sep 02 14:50:44 *** Error: CONTROL/conffiles mentions conffile /etc/fstab which does not exist Sep 02 14:56:12 cbrake: that sounds like you are somehow not getting "micro" in your OVERRIDES. Sep 02 14:56:48 does your micro-uclibc.conf contain something along the lines of: Sep 02 14:56:54 # Use overrides for micro as well Sep 02 14:56:54 OVERRIDES .= ":micro" Sep 02 14:56:54 ? Sep 02 14:58:25 checking ... Sep 02 14:58:48 yes, that is there Sep 02 14:59:39 hm. in that case, you should be getting CONFFILES set to the empty string by: Sep 02 14:59:39 I'm puzzled by this line: NOTE: Unpacking ../../openembedded/recipes/base-files/base-files/licenses/GPL-3 to tmp/work/cm-x270-oe-linux-uclibcgnueabi/base-files-3.0.14-r88/ Sep 02 14:59:45 CONFFILES_${PN}_micro = "" Sep 02 15:00:25 ... and hence it shouldn't be possible for /etc/fstab to be showing up in conffiles at all. Obviously some part of that mechanism isn't working for you but I can't immediately think what it would be. Sep 02 15:00:47 might be worth trying with "bitbake -e" and looking to see what value CONFFILES_${PN} is ending up with. Sep 02 15:00:47 yes, I'll look at what variables I have set, etc Sep 02 15:01:51 # CONFFILES_base-files_micro= Sep 02 15:01:54 hmm Sep 02 15:02:17 that looks good. what do you see for CONFFILES_base-files itself? Sep 02 15:02:49 might be a good idea to check OVERRIDES as well, make sure that its value is intact. Sep 02 15:02:58 CONFFILES_base-files="/etc/fstab /etc/hostname" Sep 02 15:03:17 hm, that's not so good. obviously it should be getting overwritten with the value from CONFFILES_base-files_micro. Sep 02 15:03:27 OVERRIDES="local:cm-x270:${MACHINE_CLASS}:micro-uclibc:linux-uclibcgnueabi:arm:build-linux:fail-fast:pn-base-files" Sep 02 15:03:40 oh ho, no micro in there. Sep 02 15:03:45 I wonder how that happened. Sep 02 15:04:16 yes, that is strange Sep 02 15:04:39 using latest dev: micro-uclibc.conf has Sep 02 15:04:40 # Use overrides for micro as well Sep 02 15:04:40 OVERRIDES .= ":micro" Sep 02 15:04:59 right, that looks like it should be doing the right thing. let me pull to make sure I am looking at the same files as you. Sep 02 15:06:13 ahh, OVERRIDES is getting overwritten in micro.conf Sep 02 15:06:42 doh, yes, so it is. that must have gotten broken when khem reorganised those files recently. Sep 02 15:06:51 if you move the OVERRIDES .= to the bottom of micro-uclibc.conf then that should fix it. Sep 02 15:07:06 g'day kergoth Sep 02 15:07:42 which bb file in images directory gives the low foot print image Sep 02 15:07:43 hey pb__ Sep 02 15:09:56 hrw: please remind me to apply the patches this evening :) Sep 02 15:11:51 03Cliff Brake  07org.openembedded.dev * rea7e5a05c4 10openembedded.git/conf/distro/micro-uclibc.conf: Sep 02 15:11:51 micro-uclibc.conf: move OVERRIDES to bottom of file Sep 02 15:11:51 micro.conf was overwriting the variable, and then Sep 02 15:11:51 micro was missing from OVERRIDES Sep 02 15:13:14 florian: ok Sep 02 15:18:18 srv: try micro-base-image with DISTRO=micro-uclibc Sep 02 15:21:49 srv: above combination is 800K for cpio.gz :-) Sep 02 15:22:53 is that with mtd-utils in? Sep 02 15:23:52 no, with mtd-utils, 1177027 bytes Sep 02 15:24:10 hum. that's still rather bigger than I am getting. Sep 02 15:24:22 oh, maybe you are getting arm code rather than thumb. Sep 02 15:24:57 pb__: yes, that is likely Sep 02 15:25:20 pb__: did thumb get turned off again recently in OE? Sep 02 15:25:23 looks like micro still defaults to arm, which is probably a bit silly. Sep 02 15:25:30 try adding: Sep 02 15:25:31 THUMB_INTERWORK = "yes" Sep 02 15:25:32 PREFERRED_ARM_INSTRUCTION_SET = "thumb" Sep 02 15:25:35 pb__: seems I recall some toolchain issues Sep 02 15:25:36 to your local.conf and see if the situation improves Sep 02 15:25:45 ok, will do Sep 02 15:25:59 fwiw, I am using: Sep 02 15:26:07 PREFERRED_VERSION_gcc-cross = "4.4.1" Sep 02 15:26:07 PREFERRED_VERSION_gcc-cross-intermediate = "4.4.1" Sep 02 15:26:07 PREFERRED_VERSION_gcc-cross-initial = "4.4.1" Sep 02 15:26:07 PREFERRED_VERSION_binutils-cross = "2.19.51" Sep 02 15:26:28 which seems to work. I think some other toolchain versions were a bit flaky with thumb. Sep 02 15:27:09 if it works for you then I will see about making thumb be the default for micro. in my experience it gives significantly smaller code and the resulting binaries run faster, so it seems like a win. Sep 02 15:28:09 pb__: 4.2.4 here Sep 02 15:28:33 pb__: do you expect compile or run-time problems with non 4.4.1 compilers? Sep 02 15:28:41 compile time, I think Sep 02 15:28:51 iirc, 4.2.x was a bit prone to ice when trying to compile uclibc Sep 02 15:29:19 by all means give it a go and see what happens. if you start to see problems along the lines of "couldn't find a register to spill in class LO_REGS" then you probably need a better compiler. Sep 02 15:30:20 * cbrake starts a fresh thumb build ... Sep 02 15:32:11 it mightn't be a bad idea to make 4.4.1 the default toolchain for micro on arm, at least when thumb is in use. I think it does still produce slightly worse code for ARM-mode than 4.2.x in some cases, but the advantage of being able to use thumb probably outweighs that. Sep 02 15:34:02 plus 4.4.x does have other advantages even in arm mode, I think you generally get better dcache usage with 4.4.x compiled code. Sep 02 15:35:53 ah, except, I remember now that c++ doesn't seem to work with uclibc using gcc 4.4.x. personally I don't care about that but I guess some other folks might :-} Sep 02 15:36:05 heh Sep 02 15:36:34 I don't think it should be that hard to fix for someone who understands the c++ runtime, it's just a case of forward porting the uclibc patches for libstdc++ Sep 02 15:37:16 iirc, if you try to build c++ against uclibc at the moment then you end up with some kind of ctype-related disaster. Sep 02 15:44:44 pb__: http://pastebin.ca/1551601 Sep 02 15:44:52 pb__: that is with 4.2 Sep 02 15:46:33 linuxtogo.org is down? :( Sep 02 15:46:44 no, is resting Sep 02 15:47:22 heh, bus error? that is a new one. Sep 02 15:47:39 Changing hosting or so? Sep 02 15:47:46 cbrake: oh, er, that was compiling a -native package. Sep 02 15:47:52 I am guessing it is the same problem as mailservers mentioned in topic Sep 02 15:48:41 pb__: hmm, restarting build ... Sep 02 15:48:41 cbrake: so it seems unlikely that thumb can have caused this. Sep 02 15:48:46 pb__: nod Sep 02 15:49:05 looks like cosmic rays, maybe, or overheating cpu or something. Sep 02 15:50:52 pb__: :-) yeah, stranger things have happened. Still running. Sep 02 15:51:23 Skyld: yes - same machine Sep 02 15:51:37 Oh no. Is the old machine still up somewhere? Sep 02 15:55:04 fsck... my C3 beagleboard do not want to behave ;( Sep 02 15:57:49 gm Sep 02 15:57:52 pb__: now I'm to the LO_REGS error you mentioned earlier: http://pastebin.ca/1551614 Sep 02 15:58:00 khem: hello Sep 02 15:58:35 LOREGS error on uclibc ldso.c ? Sep 02 15:59:12 khem: ldso/ldso/dl-elf.c:913: error: unable to find a register to spill in class 'LO_REGS' Sep 02 15:59:24 khem: this is in thumb mode. ARM builds work fine Sep 02 15:59:33 cbrake: yeah I am aware Sep 02 15:59:45 Incidentally, does anyone know if, say, an angstrom base-image is meant to produce any kernel output or does that not really appear until you get to console-image stage? Sep 02 15:59:49 cbrake: you could lower the optimization for this one file Sep 02 16:00:50 * cbrake wonders what happened to devshell ... Sep 02 16:01:18 or does that not work with micro distros ... Sep 02 16:03:07 cbrake: uclibc_ldso_use_O0.patch is this applied ? Sep 02 16:03:33 khem: appears to be Sep 02 16:04:27 cbrake: this is 0.9.30.1 Sep 02 16:04:59 yes Sep 02 16:05:20 cbrake: you could build just that one package in arm mode (i.e. set PREFERRED_ARM_INSTRUCTION_SET_uclibc = "arm"), or use a newer gcc for that one package even if you don't want to use it globally. Sep 02 16:05:42 or lowering the optimisation level might possibly help although I am not too optimistic about that Sep 02 16:06:11 lower the optimisation level and raise the optimism level :) Sep 02 16:06:26 pb__: it surely will because uclibc ldso combines the files and the function grows very big Sep 02 16:06:42 and thumb has less registers to use Sep 02 16:07:13 we ask -O2 -fno-omit-frampointer Sep 02 16:07:47 I think even if we get rid of -fno-frame-pointer it will give some respite to gcc. Sep 02 16:07:56 not that its not a gcc bug Sep 02 16:08:19 I discussed it with pbrook sometime ago Sep 02 16:08:35 pb__: what is the CTYPE error you are getting on uclibc/c++ Sep 02 16:08:52 khem: I don't remember exactly, it's months since I tried it. Sep 02 16:09:04 I just set LANGUAGES = "c" Sep 02 16:09:15 pb__: I think if you enable XLOCALE it should fix the __ctype_loc undefined error Sep 02 16:09:55 cbrake: I will try to build uclibc in thumb mode with gcc 4.4.1 and see if I hit this error Sep 02 16:10:16 pb__: I think pb has already done this Sep 02 16:10:21 the ice? that doesn't happen with 4.4.1, only with 4.2.4. Sep 02 16:10:34 iirc in angstrom we dropped idea of using thumb for uclibc - it was not working with 4.2.4 Sep 02 16:11:07 hrw: its more of a binutils problems than gcc Sep 02 16:11:11 khem: if you can confirm that c++ works with 4.4.1 under uclibc nowadays then I think it should be safe to make 4.4.1 the default compiler for micro. Sep 02 16:11:28 pb__: ok Sep 02 16:11:39 I've been using 4.4.x for a while and it seems to be fine for me other than that. Sep 02 16:11:45 ok Sep 02 16:12:08 pb__: btw what do you think of the revised target OS patch Sep 02 16:12:24 it looks good to me although I haven't checked it in detail yet Sep 02 16:12:51 pb__: ok. I tried three builds of console-images so far. and all were successful Sep 02 16:12:56 arm ones I could boot too Sep 02 16:13:05 for ppc I do not have hardware Sep 02 16:13:11 ok, great. Sep 02 16:13:29 you should get mickeyl to bless the patch, since it touches minimal, but apart from that I think I am happy with it Sep 02 16:14:46 pb__: ok I will bug mickeyl Sep 02 16:16:00 khem: go ahead Sep 02 16:16:51 mickeyl: danke schoen Sep 02 16:16:54 :) Sep 02 16:17:15 you see I am getting humbler in my deutsch Sep 02 16:18:34 Hmm :| I am awfully confused Sep 02 16:18:44 Haret gets as far as "Jumping to kernel" and then nothing happens Sep 02 16:19:06 indeed, good progress Sep 02 16:19:11 As far as I know, haret is configured properly since it's log ends with "go go go", and the images seem to have built successfully Sep 02 16:20:44 How do you know if the kernel is booting if it does not produce output :( Sep 02 16:30:52 umm... erk.. Sep 02 16:31:14 kergoth: ?! Sep 02 16:31:48 data_smart's expand() doesn't pass along 'os' in locals, so it's relying on 'os' being around as a global, for some of the path bits in bitbake.conf... this doesn't strike me as very safe Sep 02 16:33:14 i just, somehow, made a change that resulted in it not being able to find os from those python snippets.. not entirely sure how i did it, but it was just moving some vars around Sep 02 16:33:17 yeah, that is a bit of a generic problem with calls to python functions from .conf/recipe files Sep 02 16:33:19 * kergoth scratches head Sep 02 16:33:34 there isn't any way to declare the imports that you need, you basically just have to hope for the best. Sep 02 16:33:36 yep, right now we only provide 'd', and there's no mechanism to inject more bits into that environment Sep 02 16:33:37 right Sep 02 16:33:42 not good Sep 02 16:34:07 could 'def' wrappers for everything in base.bbclass, i suppose :) Sep 02 16:34:38 (aside: def'd functions don't go into globals, or into a special area for tasks/python snippets, they go into __builtins__!.. guh..) Sep 02 16:34:56 yeah, something like that would probably be the right way to proceed. in the longer term you probably want to get away from allowing bitbake content to poke at arbitrary bits of python runtime and instead confine it to a sort of walled-garden world. Sep 02 16:35:02 ah yeah, I think that was my fault. :-} Sep 02 16:35:06 hehe Sep 02 16:35:27 I can't quite remember why __builtins__ seemed like the right place for them at the time. Sep 02 16:35:30 heh Sep 02 16:35:32 yeah, I'd love to see the ${@} blocks be much more limited, more like a little DSL Sep 02 16:35:42 ideally they'd be side effect free Sep 02 16:36:08 right. Sep 02 16:36:11 * pb__ go home now Sep 02 16:36:13 later all Sep 02 16:36:18 pb__: cu Sep 02 16:36:20 later pb Sep 02 16:36:29 anyone seen zecke yet today? Sep 02 16:39:04 Can anyone help me? @ "jumping to kernel" and then stop Sep 02 16:39:53 never worked on kernels on such level Sep 02 16:39:58 ~seen zecke Sep 02 16:40:39 * kergoth thinks the next big step for parser-rework is to change the include/inherit ast nodes to parse their files at object construction time rather than eval, so those nodes would actually be StatementGroups consisting of the nodes for their files Sep 02 16:40:59 kergoth: congrats on mvl6 release Sep 02 16:41:04 :) thanks Sep 02 16:41:27 kergoth: sales@ are the only source of price? Sep 02 16:41:29 i think it fits a real need, those who want commercial support for an OE based setup Sep 02 16:41:36 right now, yeah, i haven't heard anything on prices yet personally Sep 02 16:41:42 i'll see if i can find out Sep 02 16:43:32 thx Sep 02 16:44:16 ok, time to visit castorama and buy some mounting parts for shelf in daughter's room Sep 02 16:44:23 bye Sep 02 17:16:04 wow, if you def a function in a .inc loaded by bitbake.conf, early on, the parse explodes Sep 02 17:16:10 * kergoth grumbles Sep 02 17:47:49 hello! i tried to build angstrom, but for beagleboard-demo-image it breaks with libwnck: "configure: error: gdk-pixbuf-csource executable not found in your path - should be installed with GTK" Sep 02 17:55:59 i'm rather new to openembedded/angstrom, but have been using poky for some time now Sep 02 17:56:45 i can't get bitbake to list the tasks available for a certain package. in fact, if i do `bitbake -c stage gdk-pixbuf-csource-native -f` it replies that it attempted 0 tasks Sep 02 18:03:07 * cbrake wonders why mdev does not create /dev/mtdX entries Sep 02 18:05:36 mrc3_: bitbake -c listtasks gdk-pixbuf-csource-native Sep 02 18:05:37 ta da Sep 02 18:05:38 :P Sep 02 18:06:49 hey, kergoth! i did try that, but nothing comes out of that (0 tasks of which 0 didn't need to be rerun and 0 failed) Sep 02 18:07:22 listtasks is a nostamp task Sep 02 18:07:28 its always run when you ask for it to be run. Sep 02 18:08:58 i know. something seems broken with my bitbake Sep 02 18:09:19 oddly enough, i get the tasks for gtk+ (for instance) Sep 02 18:10:02 broken BBPATH could easily be the culprit Sep 02 18:10:18 if its preferring the bitbake classes/base.bbclass, you woudln't get the stock tasks provided by the OE one Sep 02 18:16:58 re Sep 02 18:33:39 Skyld: what is Haret? Sep 02 18:49:41 likewise: http://handhelds.org/moin/moin.cgi/HaRET Sep 02 18:55:24 santamarta: tnx Sep 02 19:10:12 hmmm changing linux-uclibcgnueabi to linux-uclibceabi means I need to to bump up OELAYOUT_ABI too in conf/abi_version.conf Sep 02 19:26:17 ping mickeyl Sep 02 19:26:23 hi all ! Sep 02 19:26:34 hi gremlin[it] Sep 02 19:26:45 hi khem Sep 02 19:27:03 hi gremlin[it] Sep 02 19:27:31 hi pb___ Sep 02 19:37:28 pb___, you are involved with the General Assembly ? Sep 02 19:37:55 pb__: I was wondering if people could attend remotely Sep 02 19:38:12 dial in or something ? Sep 02 19:39:13 can anyone recall the motivation behind allowing empty -dbg & -dev packages, by cahnce? Sep 02 19:43:13 Hum Sep 02 19:43:24 hi Sep 02 19:43:44 I think at one point, there were some dumb deps around that try and always pull those packages in, even if they didn't exist Sep 02 19:44:00 thats what i figured, but i dont really see the point now Sep 02 19:44:13 task-foo-dev for example uses rrecommends to pull them in, not deps Sep 02 19:44:21 Or maybe it was the other way around and that's what got fixed? ie -dev/-dbg always depended on a main, and main was empty since everything was in some other -sub package Sep 02 19:44:40 I know I got some problem like that fixed, a while back Sep 02 19:44:42 hmm Sep 02 19:44:50 well, either way, i think empty -dev/-dbg packages are pointless Sep 02 19:44:53 just clutters things up Sep 02 19:45:09 easier to spot msising debug info / stripping without them, too, can just look in the ipk dir and see which dont' exist Sep 02 19:45:26 * kergoth 'll test a build without that allow_empty Sep 02 19:46:19 3b701993d63e2c2f697bbf546db72d359f3bae93 is what I was thinking of Sep 02 19:46:53 But it's likely we allowed for always-empty dev/dbg for similar reasons as to why Koen objected to that change initially on the ML Sep 02 19:47:08 s/allowed for/have/ Sep 02 19:51:02 * kergoth 'll have to read that thread Sep 02 19:51:04 thanks Sep 02 19:58:29 gremlin[it]: I'm not really involved with the general assembly itself, I think that is a matter for the oe eV board of directors. Sep 02 19:58:39 but I am involved with the wider logistics of the oedem, yes Sep 02 19:58:40 * chouimat wanders in Sep 02 19:58:43 hi * Sep 02 19:58:44 hi pb Sep 02 19:58:58 hi chouimat Sep 02 19:58:59 hi woglinde Sep 02 19:59:30 khem: that's a possibility but I can't promise anything at the moment. depends what venue we end up with. Sep 02 19:59:51 pb___: it would be great for people like me Sep 02 19:59:54 kergoth: yah, agreed, we should probably ditch the empty -dev packages. Sep 02 20:00:03 pb___, I'm not able to contact Mickey, i cannot attend at the GA and i don't know if my vote is necessary or not ... so to provide a proxy document for that ... Sep 02 20:00:40 kergoth: -dbg is maybe more questionable. depends if there's a way of predicting whether a given package will have -dbg or not. Sep 02 20:01:42 pb___, about OEdem ... have you a barelly idea about cost for a place to stay in Cambridge ? ... money at now is one of my first issues about partecipate ad OEdem or not Sep 02 20:05:19 gremlin[it]: you'd need to talk to mickeyl, florian or zecke about the GA proxy thing, I don't know anything about that. Sep 02 20:06:03 gremlin[it]: I think the cheapest accomodation in Cambridge is on the order of £20-£30 per night. Sep 02 20:06:32 might be able to get a youth hostel room for £20 (try yha.org.uk), otherwise there are various bed and breakfast places around the £30 mark. Sep 02 20:06:39 if you want a full service hotel, probably double that Sep 02 20:07:33 pb___: http://kergoth.pastey.net/124559 Sep 02 20:07:35 pb___, thanks for info ! Sep 02 20:11:06 kergoth: I'm not quite sure I understand what that's doing. Sep 02 20:11:27 no ${PN}-dbg, there's a -dbg for each package with binaries Sep 02 20:11:39 and no empty dbg packages Sep 02 20:12:25 hm youth hostel has free beds Sep 02 20:13:23 kergoth: ah right, got it Sep 02 20:14:32 kergoth: so, my reservation about not having empty -dbg packages is that, if you've got an existing filesystem and you just want to install all the debug info, you can't just iterate over the installed packages and do "ipkg install %s-dbg" Sep 02 20:14:46 ahh, right Sep 02 20:14:55 well, you can, but it will error Sep 02 20:15:09 but arguably that error is a better indicator of your inability to debug that package than installing nothing would be Sep 02 20:15:19 well, yeah, and you then can't tell the difference between a broken feed and a genuinely nonexistent debug package Sep 02 20:15:19 03Koen Kooi  07org.openembedded.dev * r5777be6816 10openembedded.git/ (37 files in 6 dirs): Sep 02 20:15:19 linux-omap-* 2.6.29: add l1 cache performance tweak from l-a-k ml Sep 02 20:15:19 * also sync patch between pm and regular kernels Sep 02 20:16:01 I guess if there was some kind of marker in the corresponding binary package, say "Debug-Info-In: libfoo-dbg", that'd help Sep 02 20:18:11 hi ant Sep 02 20:18:30 hello woglinde Sep 02 20:21:18 hi gnutoo Sep 02 20:21:33 hi Sep 02 21:17:11 03Koen Kooi  07org.openembedded.dev * rc73cd5fdeb 10openembedded.git/classes/gnome.bbclass: gnome bbclass: remove to more auto generated cruft that breaks package management Sep 02 22:41:10 damnit Sep 02 22:41:29 forgot that pkgdata doesn't include PROVIDES Info, so this shlibs based missing dependency checking isn't full proof Sep 02 22:41:43 says i need to depend on glibc, when I'm depending on virtual/libc :) Sep 02 22:43:03 bah, i wish the CACHE was a db like persist_data Sep 02 23:37:34 I was wondering what are native packages really we muck with that I need to override on my build system Sep 02 23:38:08 say I have something sane build machine like ubuntu/debian fedora suse etc. Sep 02 23:38:26 I think most of native packages can be used as provided by those Sep 02 23:49:41 khem, I tried to ASSUME_PROVIDED a whole lot of the basic native packages a few weeks ago. Things broke in weird ways. Sep 02 23:50:30 grg: well I will be conservative and still weed out some stuff like python and perl and gettext Sep 02 23:50:49 and keep reducing Sep 02 23:50:58 sure Sep 02 23:52:17 i suppose having all those native packages means that hitting an obscure build issue gets fixed for everyone and doesn't exist for only a few people running certain combinations of versions Sep 02 23:56:15 yes, cumbersome for me where I build from scratch everyday Sep 02 23:56:27 I know the advantages are there Sep 03 00:03:12 grg: this list here http://khem.pastey.net/124571 seems to be ok on karmic Sep 03 00:03:31 ofcourse you also need to make sure that build machine has the packages installed Sep 03 00:03:50 what is karmic? Sep 03 00:04:02 ugh ubuntu 9.10 (future) Sep 03 00:04:28 oh. ok Sep 03 00:05:10 my tasklist went down from 2000 to 1000 for a console-image Sep 03 00:05:17 approx Sep 03 00:05:24 excellent! Sep 03 00:05:28 I've just binished building the latest TI codec engine, dsplink, dmai etc. Sep 03 00:05:37 anyone interested in patches? Sep 03 00:06:05 I am sure there are people Sep 03 00:06:19 hyc, is that for omap35xx ? Sep 03 00:06:24 hyc: send the patches to the mailing list Sep 03 00:06:26 omap3530 yeah Sep 03 00:06:32 ok Sep 03 00:06:56 grg: How is gcc 4.4.1 going on mips Sep 03 00:07:05 its very nice Sep 03 00:07:09 (and the important point - it actually works.. ;) Sep 03 00:07:17 hyc: nice Sep 03 00:07:33 grg: What mips hardware do you have Sep 03 00:07:44 and what ABI do you use N32 O32 or N64 Sep 03 00:07:45 4.4.1 configured --with-mplt is faster than gcc 4.3.3 Sep 03 00:08:06 i have a jz4730 based board, o32+plt abi Sep 03 00:08:10 grg: Do you have perf numbers ? Sep 03 00:08:25 nope. Sep 03 00:08:30 it will be very interesting to me to justify new compiler Sep 03 00:08:33 subjective testing only Sep 03 00:08:54 ok. I think you could try some benchmarks Sep 03 00:09:28 How much perf gain do you see Sep 03 00:09:36 yeah, i know i should, but there's only me here working on this and i've got plenty else to be working on Sep 03 00:09:53 grg: ok where do u work :) Sep 03 00:10:04 small company, ubiq technologies Sep 03 00:10:14 in adelaide, australia Sep 03 00:10:33 so you must know cricket :) Sep 03 00:10:45 yep Sep 03 00:10:56 i know we suck Sep 03 00:11:00 (right now) Sep 03 00:11:04 heh Sep 03 01:42:16 ok, posted dsp patches to openembedded-devel, dunno if the list requires subscribe first. I didn't subscribe yet Sep 03 02:04:21 hyc, i don't see your mail on the list... Sep 03 02:04:26 Yeah, not here either Sep 03 02:13:42 I guess I should subscribe and repost Sep 03 02:19:06 ok, I see it in the archive now http://lists.linuxtogo.org/pipermail/openembedded-devel/2009-September/013553.html **** ENDING LOGGING AT Thu Sep 03 02:59:57 2009