**** BEGIN LOGGING AT Wed Jun 30 02:59:57 2010 Jun 30 06:05:59 03sledz  07org.openembedded.dev * rc45b19daad 10openembedded.git/recipes/linux/linux-2.6.24/hipox/defconfig: Jun 30 06:05:59 linux-2.6.24: enable DEBUG_FS and USB_MON for hipox machine Jun 30 06:05:59 Signed-off-by: Steffen Sledz Jun 30 06:05:59 git-svn-id: https://svn.dresearch.de/repos/openembedded/branches/2010-02-24_initial@92 ba11ecae-741b-462f-8724-1218f99f5906 Jun 30 06:51:34 hm, would have expected that the unpack stage would already copy the patches to the workdir (but not apply them), is it a bug or a feature that this does not happen? Jun 30 07:11:15 eFfeM_work: it works like this for me (I have them in ${WORKDIR} after -c unpack) Jun 30 07:11:45 hm, just tried it for gcc-cross-intermediate and they were not there Jun 30 07:12:25 wanted to have them copied then quilt push them until I hit the one that needed a patch, now did -c patch and popped back Jun 30 07:12:32 but was kinda surprised by this Jun 30 07:12:36 will retry Jun 30 07:25:04 btw is there a way to force printing a message when a certain machine is used when baking things? for nios2 I would like to print a message where to find the actual cpu file Jun 30 07:25:27 alternative is probably to have a readme or to add comment in the machine conf file Jun 30 07:49:50 morning Jun 30 08:22:51 good morning Jun 30 08:27:35 hi florian Jun 30 08:50:09 03Robert Schuster  07org.openembedded.dev * r748d27503b 10openembedded.git/recipes/llvm/llvm2.7_2.7.bb: llvm2.7 2.7: Split packages for each shared library. Jun 30 08:50:19 jo raster Jun 30 08:50:24 03Robert Schuster  07org.openembedded.dev * r3a0f681008 10openembedded.git/recipes/llvm/ (9 files in 3 dirs): llvm: Removed 2.5 and 2.6 and both native variants. Jun 30 08:50:25 03Robert Schuster  07org.openembedded.dev * r06be7fc026 10openembedded.git/recipes/llvm/ (llvm-native.inc llvm.inc llvm2.7_2.7.bb): llvm.inc: Moved packages dynamic stuff into the general llvm.inc file. Jun 30 08:59:46 woglinde: poop! Jun 30 09:15:48 hi mrmoku Jun 30 09:17:33 hi woglinde Jun 30 09:23:35 03Martin Jansa  07org.openembedded.dev * r723a7c8b21 10openembedded.git/recipes/xorg-driver/ (2 files in 2 dirs): Jun 30 09:23:36 xf86-video-glamo: better patch for nonDRM kernels Jun 30 09:23:36 Signed-off-by: Martin Jansa Jun 30 09:23:37 03Martin Jansa  07org.openembedded.dev * rc2c6f15e59 10openembedded.git/recipes/xorg-xserver/ (2 files in 2 dirs): Jun 30 09:23:37 xserver-xorg-1.8.99.903: update patch for freedesktop #28824 Jun 30 09:23:37 * https://bugs.freedesktop.org/show_bug.cgi?id=28824 Jun 30 09:23:37 Signed-off-by: Martin Jansa Jun 30 09:36:51 hi gnutoo Jun 30 09:39:52 woglinde, hi Jun 30 10:23:56 jo kergoth Jun 30 10:52:16 khem, ping? Jun 30 11:03:05 03Martin Jansa  07org.openembedded.dev * r58cdf0ae06 10openembedded.git/recipes/openmoko-3rdparty/mcnavi_0.2.10.bb: Jun 30 11:03:06 mcnavi: add 0.2.10 Jun 30 11:03:06 Signed-off-by: Martin Jansa Jun 30 11:17:18 03Robert Schuster  07org.openembedded.dev * rd5be18abe1 10openembedded.git/recipes/llvm/llvm2.7_2.7.bb: llvm2.7 2.7: Changed patch URL to new format. Jun 30 11:18:14 Hi All, one doubt pls.. bitbake only uses Internet for downloading sources tars and zips. ?? and no other purpose!!? Jun 30 11:19:17 Nikunj: and tinderbox if you have activated that Jun 30 11:20:09 and its not bitbake which accesses the net to retrieve sources but the programs that are needed to fetch a particular URI (eg. a SVN, GIT or HG repository) Jun 30 11:21:22 thebohemian: Thanks for the info, but if I want to restrict bitbake to use only whatever is present on the system ( the sources/ folder )and dnt look on the net..what can be done here then.? Jun 30 11:21:45 thebohemian: and how bad is goin in this way.? Jun 30 11:22:16 Nikunj: one way is to only run the tasks that fetch the sources once Jun 30 11:22:37 when that finishes the next run will not fetch any sources from the net Jun 30 11:22:58 (provided you do not change the metadata in between, e.g. git pull) Jun 30 11:23:17 but will ask VCS servers for latest revs if ie AUTOREV is used in some recipe Jun 30 11:23:27 (yuck) Jun 30 11:24:42 thebohemian: hmm..that means, If I unplug network cable and start bitbaking.. ( Assuming my sources/ folder has all needed tars and zips ) the chances that my run will be successful is as good as with network cable.? Jun 30 11:25:28 Nikunj: I think so, yes Jun 30 11:25:57 thebohemian : Thts great..!! :-) Jun 30 11:56:47 hi hrw Jun 30 11:57:08 re Jun 30 11:57:35 hrw could you please ack the patches from stefan? Jun 30 11:57:39 for stable Jun 30 11:59:10 will look Jun 30 12:00:03 ecj acked - something more? Jun 30 12:01:06 hm Jun 30 12:01:15 damn Jun 30 12:01:20 gcc isnt complete yet Jun 30 12:01:23 will talk with stefan Jun 30 12:01:25 khem, uclibc.org/~aldot/uClibc/nptl/ Jun 30 12:03:46 03Koen Kooi  07org.openembedded.dev * r52ebbfaf6c 10openembedded.git/recipes/jamvm/jamvm-initial_1.4.5.bb: jamvm-initial 1.4.5: fix dependencies, see http://tinderbox.openembedded.org/public/logs/task/6388490.txt Jun 30 12:05:40 03Koen Kooi  07org.openembedded.dev * r81adc0e1ce 10openembedded.git/recipes/xbmc/xbmc_svn.bb: xbmc: fix version and fixup python lib Jun 30 13:09:08 03Dmitry Eremin-Solenikov  07org.openembedded.dev * r40171e8aa6 10openembedded.git/recipes/libxsettings-client/ (4 files): Jun 30 13:09:08 libxsettings-client: switch to using INC_PR Jun 30 13:09:08 Signed-off-by: Dmitry Eremin-Solenikov Jun 30 13:09:09 03Dmitry Eremin-Solenikov  07org.openembedded.dev * r23d51f646b 10openembedded.git/recipes/libxsettings-client/libxsettings-client.inc: Jun 30 13:09:09 libxsettings-client: use xsettings-common.h from libxsettings-dev Jun 30 13:09:09 Signed-off-by: Dmitry Eremin-Solenikov Jun 30 13:09:10 03Dmitry Eremin-Solenikov  07org.openembedded.dev * re90cf71fe6 10openembedded.git/recipes/xcb/ (xcb-proto.inc xcb-proto_1.6.bb): Jun 30 13:16:20 can I extend FILESPATHPKG based upon architecuture? e.g. FILESPATHPKG_nios2 += "glibc-2.5:files:" Jun 30 13:16:24 ? Jun 30 13:20:03 eFfeM_work: yes, but why? Jun 30 13:20:54 zecke: just tested it, apparently it does not work, if I remove the _nios2 part it does work Jun 30 13:21:35 zecke: nios2 needs some additional glibc patches for 2.5 but apparently all patches from 2.5 are taken from the glibc-2.4 dir Jun 30 13:22:04 effem use add Jun 30 13:22:41 woglinde add? or append? FILESPATHPKG_add_nios2 ? Jun 30 13:23:15 actually glib2.5 defines FILESDIR, doesn't seem to sue FILESPATHPKG Jun 30 13:23:31 glibc Jun 30 13:23:41 this is somewhat of a mess Jun 30 13:28:32 anyway, got it going: FILESPATHPKG_nios2 = "glibc-2.5:glibc-2.4:files:" Jun 30 13:29:04 not the nicest solution but as people feel 2.5 should get its files from the 2.4 dir this seems the best I can do Jun 30 13:29:16 hm Jun 30 13:29:35 I think nobody really cared about this all Jun 30 13:29:41 and nobody cleaned it up Jun 30 13:30:14 i coined it on the ML, did get some negative feedback :-( will paste url Jun 30 13:32:47 effem best would be nios could take a actual libc Jun 30 13:32:50 actually it was not mailed as such, my original msg was in http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-June/020963.html Jun 30 13:32:53 an Jun 30 13:32:56 or uclibc Jun 30 13:33:20 might be someone here mentioned that 2.5 referred to 2.4 to save some space Jun 30 13:33:32 woglinde: I understand and share your feelings. This is on the todo list Jun 30 13:33:59 basically there is a nios2 git and that one has 2.5; will try a later version at a later tiem Jun 30 13:39:38 (btw tried uclibc for nios2+mmu but that needed some work,might also peek at eglibc); want to upgrade binutils (and maybe gcc) first Jun 30 13:44:13 effem a lot of work to do Jun 30 13:46:55 woglinde: yes; I'm already happy that I got the stuff building based upon standard releases i.s.o. windriver/csl stuff Jun 30 13:48:03 latest binutils is probably mostly done Jun 30 13:48:31 hehe Jun 30 13:49:01 gcc will be a pain, might have chagnes to go to 4.2, but 4.4 still gave a lot of issues Jun 30 13:50:14 I hope you get paid well Jun 30 13:51:46 hmm. that is not too bad but my boss is already happy if we have it runnign and is not too interested in moving to the latest gcc etc Jun 30 13:52:16 * eFfeM_work learned a lot about gcc machine descriptions recently, trying to find out why things work on 32bit but not on 64bit Jun 30 13:53:17 effem sure Jun 30 13:56:09 part of the fun ... Jun 30 14:12:39 hah, i forgot that originally referencing a var that didn't exist was an error. zecke added it, "VarExpandError" was raised in that case back in the original expansion code Jun 30 14:15:22 *g* Jun 30 14:16:45 * kergoth thinks itd be nice to do that again :) Jun 30 16:05:57 hi all Jun 30 16:06:39 hi Jun 30 16:07:32 i had built a custom kernel with external toolchain ... but when loading the image using uboot it give " BOOTING IMAGE AT xxxx ....BAD MAGIc number " s Jun 30 16:07:52 can some one let me know what could be th erason Jun 30 16:09:56 wrong kernel config? Jun 30 16:10:10 ill check that Jun 30 16:10:11 you didnt create an uImage Jun 30 16:10:21 i created an uImage Jun 30 16:10:24 okay Jun 30 16:10:27 from ZImage Jun 30 16:17:40 newbie_sreddy: and how did you build uImage ? Jun 30 16:17:57 did you use kernel make target Jun 30 16:19:17 i build using makeimage tool form my toolchain Jun 30 16:20:07 khem, hi. see above for the *patch we talked about yesterday Jun 30 16:20:40 i just compiled the kernel using OE to zImage .. but after that using the using tool chain I created the UImage Jun 30 16:21:21 so you dont needed oe at all Jun 30 16:22:22 finally i need .... it but as of now for tesing I am using OE upto do_compile task Jun 30 16:23:10 newbie_sreddy: and you are sure that you did the right thing Jun 30 16:23:37 blindvt: hmm those look invasive Jun 30 16:24:03 blindvt: btw noq I feel better after fixing errno on mips. Jun 30 16:24:22 khem hehe Jun 30 16:24:26 khem, they don't just look like ;P Jun 30 16:24:32 I will check once again ... is there anything which i should look into apart from kernel configuration and using of correct toolchains ? Jun 30 16:28:39 khem, yea, thanks alot for that. Jun 30 16:42:02 RP: ping Jun 30 16:54:57 hm sed guru here Jun 30 16:55:17 I want to place a newline containing stuff before the string I am searching Jun 30 16:56:52 "place"? as in, include a newline in the replacement text? Jun 30 16:57:19 hm now Jun 30 16:57:22 ups now Jun 30 16:57:25 args no Jun 30 16:57:39 I want find the string and place line above this string Jun 30 16:57:54 that doesn't make it any clearer :) Jun 30 16:57:58 elaborate or give an example please Jun 30 16:58:12 I want find foo somewhere in the file Jun 30 16:58:33 and one line before foo I want make a newline with some stuff Jun 30 16:59:00 so yes, you want to include a newline in the replacement text Jun 30 16:59:07 you want to find foo and replace it with newline, some stuff, foo Jun 30 16:59:15 no? Jun 30 16:59:21 ah good idea Jun 30 16:59:30 now its clear Jun 30 16:59:36 how it will work Jun 30 16:59:38 thanks Jun 30 16:59:58 not sure offhand if it'd be easy to do, but i'd think something like s/foo/\nbar\nfoo/ or perhaps \r instead.. Jun 30 17:00:10 if that doesn't work, i think theres other ways, but thats worth trying first Jun 30 17:00:56 yes worked Jun 30 17:01:00 with grouping Jun 30 17:01:20 sed -i -e "s|\(^sdf*\)|foo\n\1\n|" testfile.txt Jun 30 17:01:55 nice Jun 30 17:02:27 hm the seconde \n is unecessary Jun 30 17:04:16 hm cool Jun 30 17:42:10 I deleted a bunch of images for my machine and would like to rebuild the uImage again, what recipe would i need to clean to make that happen? Jun 30 17:42:37 ???? Jun 30 17:43:01 uImage is the wrapped z or bzImage Jun 30 17:43:09 a normal image contains filesystem Jun 30 17:43:24 and maybee the uImage Jun 30 17:43:36 yes, I'd like to get the packaged kernel rebuilt Jun 30 17:43:49 as uImage Jun 30 17:45:06 Do I make sence? Jun 30 17:45:15 yes Jun 30 17:45:43 hm uImage is only generated when makeing an normal image Jun 30 17:45:55 so I tried bitbake -c clean console-image, and as you know it doesnt really clean deep enough Jun 30 17:46:08 sure its doing Jun 30 17:46:17 delete stuff under delpoy Jun 30 17:46:21 images Jun 30 17:46:46 run bitbake console-image again Jun 30 17:47:54 after bitbake -c clean console-image Jun 30 17:50:33 thats how it started, I deleted a bunch of old images. Then I tried to clean console-image and rebuilt it Jun 30 17:50:53 I guess I'll try to clean the kernel somehow Jun 30 17:50:55 kergoth: I want to setup a clone of a git repo but so that I could share it internally do have some advise for me Jun 30 17:51:43 kergoth: I still want the repo to be pulled from community repo but publish the internal repo so people can clone it. it wil also have few internal patches Jun 30 17:51:47 on top Jun 30 17:58:02 damn Jun 30 17:58:05 damn Jun 30 17:58:12 depenendcy stuff Jun 30 17:58:19 why I am not allowed Jun 30 17:58:47 to use packages without ${PN} Jun 30 17:58:51 package Jun 30 18:06:04 he crofton Jun 30 18:06:15 you are in germany? Jun 30 18:06:20 woglinde, what do you mean use packages without ${PN}? Jun 30 18:06:33 yeah Jun 30 18:06:38 Friedrichstafen ... Jun 30 18:07:02 kergoth I want PACKAGES = "${PN}-dev ${PN}-dbg ${PN}-doc " Jun 30 18:07:06 a nd Jun 30 18:07:15 PACKAGES_DYNAMIC = "llvm-*" Jun 30 18:07:27 Friedrichstafen is a cool place Jun 30 18:07:40 Crofton: I use to go there quite a lot Jun 30 18:07:42 beacause ${PN} make not really sin for llvm Jun 30 18:07:57 woglinde, would probably have to tweak RDEPENDS_${PN}-{dev,doc,dbg}, make sure they don't depend on ${PN} Jun 30 18:07:58 cool Jun 30 18:08:07 but the ACKAGES_DYNAMIC = "llvm-*" depneds on ${PN} Jun 30 18:08:08 I am going to the flying museums in the morning Jun 30 18:08:15 even it isnt there Jun 30 18:08:15 hmm Jun 30 18:08:18 Crofton: you should also go to bodensee Jun 30 18:08:23 Fly out of Frankfurt friday mornuing Jun 30 18:08:29 that is the water thing Jun 30 18:08:38 need to go outside and look around Jun 30 18:08:43 I am now try to work around it with empty ${PN} Jun 30 18:08:47 and get of interweb until is it dark :) Jun 30 18:08:50 bye guys Jun 30 18:08:53 thanks for the tips Jun 30 18:09:00 Crofton: enjoy Jun 30 18:09:07 bye crofton drink some beer Jun 30 18:09:23 woglinde, of course :) Jun 30 18:09:52 Crofton: ulmer muenster :) Jun 30 18:10:19 sigh Jun 30 18:10:55 and the ${PN} depends on all PACKAGES_DYNAMIC = "llvm-*" even its empty Jun 30 18:11:36 or the index generator suckz Jun 30 18:11:56 args Jun 30 18:15:48 kergoth any idea? Jun 30 18:20:29 hm okay Jun 30 18:20:37 fixed now Jun 30 18:25:06 re florian Jun 30 18:25:16 re Jun 30 18:45:44 Its slow the compilation? Jun 30 18:46:02 ????? Jun 30 18:46:04 slow is relative term Jun 30 18:46:07 dont chat as root Jun 30 18:46:11 you might get hacked Jun 30 18:46:22 i am using slackware.... Jun 30 18:46:36 dont depends on the host distro Jun 30 18:46:41 depends only on hw Jun 30 18:46:42 and i cannot use another user to chat.... Jun 30 18:46:49 sure you can Jun 30 18:46:56 su - newuser Jun 30 18:46:58 i get a error..... Jun 30 18:47:06 adduser newuser Jun 30 18:47:23 well you get hacked thats your headache isnt it :) Jun 30 18:47:33 whats your problem with OE Jun 30 18:47:56 none...but is slow to compile..... Jun 30 18:48:08 task 194 of 3274 Jun 30 18:48:11 whats your build box configs Jun 30 18:48:33 I using P4 1gb Memory Jun 30 18:48:50 thats a slow machine Jun 30 18:49:09 we are building libc twice and gcc third Jun 30 18:49:17 yes I have one of those too and OE has made me think its worthless so I gave it away to kids to play Jun 30 18:50:02 hum.....Whats machine is good for this job? Jun 30 18:50:16 4-qay Jun 30 18:50:21 ups 4-way Jun 30 18:50:26 bye for now Jun 30 18:51:08 Have any rootfs opeenbbeed of a1200 to download Jun 30 18:51:13 ????? Jun 30 18:51:14 Angelox_123: people have core i7 Jun 30 18:51:22 Wow!! Jun 30 18:51:41 Angelox_123: you mean prebuilt Jun 30 18:51:41 03Koen Kooi  07org.openembedded.dev * r1f1a416035 10openembedded.git/recipes/xbmc/xbmc_svn.bb: xbmc: add missing deps, bump SRCREV Jun 30 18:52:07 yes khem Jun 30 18:53:49 Angelox_123: I dont know off hand Jun 30 18:54:12 Angelox_123: OE is about building yourself :) Jun 30 18:54:20 Hum..... Jun 30 18:55:08 what a pity..... Jun 30 18:56:07 03Khem Raj  07org.openembedded.dev * r48dfe40e52 10openembedded.git/contrib/qemu/run-qemu.sh: Jun 30 18:56:07 contrib/qemu/run-qemu.sh: Dont search for hdimage when doing a nfsboot Jun 30 18:56:07 Signed-off-by: Khem Raj Jun 30 20:12:16 03Dmitry Eremin-Solenikov  07org.openembedded.dev * r866fc2f108 10openembedded.git/recipes/libxsettings/libxsettings_0.11.bb: Jun 30 20:12:17 libxsettings: bump PR to force rebuild to restore xsettings-common.h in sysroots Jun 30 20:12:17 Signed-off-by: Dmitry Eremin-Solenikov Jun 30 20:51:40 03Martin Jansa  07org.openembedded.dev * r48cbac51b7 10openembedded.git/conf/distro/include/sane-srcrevs.inc: Jun 30 20:51:40 EFL: bump SRCREV a bit more for tasn's fix for Arabic shaping Jun 30 20:51:40 Signed-off-by: Martin Jansa Jun 30 20:51:41 03Martin Jansa  07org.openembedded.dev * r6e70d7daf1 10openembedded.git/recipes/xorg-driver/ (2 files in 2 dirs): xf86-video-glamo: bump SRCREV, remove applied patch Jun 30 21:40:52 03Klaus Kurzmann  07org.openembedded.dev * r6918af9576 10openembedded.git/recipes/linux/ (2 files in 2 dirs): Jun 30 21:40:52 linux-openmoko-2.6.32_git.bb: add a patch to make jack input events work Jun 30 21:40:52 Signed-off-by: Klaus Kurzmann Jun 30 21:41:58 Heya guys! My ship is sinking! So, im building openembedded for the gumstix overo, and my hard drive space is quickly dwindling! I added rm_work or whatever to my local.conf, which is supposed to cut down on space, and ran bitbake to build my distribution again, but it seems that it didn't actually go about cleaning up whatever files are taking up all of the space before it went to compile all over again... Help me #oe, your my only hope! Jun 30 21:42:49 you could try rebuilding from packaged staging Jun 30 21:43:04 remove all in your tmpdir BUT /pstage Jun 30 21:43:12 thanks! Jun 30 21:43:25 keep /pstage ! Jun 30 21:44:42 What does rm_work do anyways? Jun 30 21:45:23 it does remove the files in workdir but being the packages are staged (Already compiled once) these are left untouched... Jun 30 21:45:48 thus, rebuilding is needed Jun 30 21:46:46 wow, so if i rm_work, that means that every time i build an image, i will have to wait 12 hours? Jun 30 21:47:12 well, only on first build Jun 30 21:47:22 from pstage should be much more rapid Jun 30 21:47:41 i.e. 15mins vs. 2 hours Jun 30 21:48:01 in my case on my host Jun 30 21:48:22 s/should be/is/ Jun 30 21:49:25 gotcha Jun 30 22:09:06 bkinman: another thinh to change is the debug option Jun 30 22:10:05 re ant Jun 30 22:10:21 bkinman: I think its using -g3 at the moment and it will eat up your harddisk Jun 30 22:10:29 ant__: hello Jun 30 22:10:34 hwy woglinde ...bah.. netsplit Jun 30 22:10:44 hello khem Jun 30 22:10:51 oh, awesome. Jun 30 22:10:55 Ill look into that. Jun 30 22:12:09 bkinman: it will also speed up your build if you used =g instead of g3 Jun 30 22:12:57 Where do i change that at? Jun 30 22:13:35 bkinman: whats your DISTRO Jun 30 22:14:03 I'm using angstrom. Jun 30 22:14:24 ok open conf/distro/include/angstrom-glibc.inc Jun 30 22:14:39 and you should see something like FULL_OPTIMIZATION = "-fexpensive-optimizations -frename-registers -fomit-frame-pointer -O2 -ggdb3" Jun 30 22:15:00 replace -ggdb3 with -g Jun 30 22:15:08 -O2 -ggdb3? Jun 30 22:15:18 thats crazy but thats how it is Jun 30 22:15:29 hm.. looking strange :) Jun 30 22:15:37 uclibc was 232M when built with -ggdb3 :) Jun 30 22:15:49 angstrom maintainers like it Jun 30 22:15:58 omg Jun 30 22:15:59 as they are blessed with large harddisks Jun 30 22:16:12 well.. sleep anyway :) Jun 30 22:16:20 I stopped using angstrom for very same reason Jun 30 22:16:24 I use minimal Jun 30 22:16:27 its sleek Jun 30 22:16:37 bkinman: did you get it ? Jun 30 22:16:40 * Jay7 -> sleep() Jun 30 22:16:52 i'm resizing my virtual disk partition, hehe. Jun 30 22:17:45 Well, -ggdb3 makes debugging deployed stuff easier i think Jun 30 22:17:52 * Tartarus needs to find the time to confirm this, one of these days Jun 30 22:18:13 Tartarus: may be but at what cost Jun 30 22:18:15 * Tartarus is busy kicking bitbake world -k -c fetchall actually finish, with strict checksums disables Jun 30 22:18:32 khem: I'm with the Angstrom folks, sorry Jun 30 22:18:48 Not that I advocate minimal/micro going -ggdb3 too Jun 30 22:19:00 disk is cheap, even laptop disk Jun 30 22:19:01 Tartarus: I have found that -g is good for debugging Jun 30 22:19:07 -g3 is overkill Jun 30 22:19:20 khem: I think the point of -ggdb3 is that you don't need sources around then Jun 30 22:19:23 THat's what I want to confirm anyhow Jun 30 22:19:26 unless you compress debug info Jun 30 22:19:29 which gcc does not Jun 30 22:19:52 Tartarus: why you dont need sources ? Jun 30 22:20:04 how would you do source level debugging without them Jun 30 22:21:03 khem: What I suspect and would like to confirm is that -ggdb3 gives you what you need when you don't want to debug libpcap (for example), but your libpcap using app and you're stepping in to see what's going on Jun 30 22:21:19 That's when you don't care about having libpcap sources around Jun 30 22:21:24 but do want source level debug Jun 30 22:22:01 Tartarus: I dont get you. Do you just want frame info and symbols ? Jun 30 22:22:06 for those Jun 30 22:22:17 they will be there with -g as well Jun 30 22:22:38 g3 adds macro debugging capabilities the biggest source of bloat Jun 30 22:22:52 and it will generate it for every occurance and use of the macro Jun 30 22:22:54 lame Jun 30 22:23:24 there is a fine compromise you have to have with opt level and debug info Jun 30 22:23:45 if you are asking for O2 lot of that bloated crap is wrong anyway Jun 30 22:24:04 gcc emitted that just because you asked for it. Jun 30 22:24:22 If you used something like -O0 -g3 then ok Jun 30 22:24:30 you will get super duper debug info Jun 30 22:24:41 and world class debugging experience Jun 30 22:24:52 khem: Like I said, it's something I want to confirm before I speculate more Jun 30 22:25:02 and I'm also assuming that there's a good and tested reason for it in Angstrom Jun 30 22:25:14 Tartarus: I have used both so I am speaking from experience Jun 30 22:25:35 there could be a good reason but I dont know that Jun 30 22:25:45 Yes, that's what I'm speculating about Jun 30 22:25:49 generally we should approach it differently Jun 30 22:26:16 we should have options sets which dont harm debugging that much Jun 30 22:26:37 and then another option set which doesnt care about debugging at all that would be need for speed kind Jun 30 22:27:01 imho, anything that says "we build differently for debug vs not" is a deal breaker Jun 30 22:27:16 You will then get the bug that shows up in the 'not' case Jun 30 22:27:38 I agree. However this is an option for daring ones Jun 30 22:27:55 we might go conservatively with the option set thats good for debugging Jun 30 22:28:19 but we should also have an option to deteriorated debugging experience but faster speed Jun 30 22:28:59 the current option set we use is crazy Jun 30 22:29:29 we tend to optimize for size but then we ask for frename-registers and expensive-optimizations Jun 30 22:29:35 thats counter intuitive Jun 30 22:35:56 er, wait, how so? Jun 30 22:36:26 In that both of them are perf and necessarily size related? Jun 30 22:36:29 Tartarus: let me try out something like -g3 -feliminate-dwarf2-dups -femit-struct-debug-baseonly -femit-struct-debug-reduced Jun 30 22:36:55 Tartarus: no expensive-optimizations break debug info Jun 30 22:37:33 khem: Other than the usual debugging optmized code is hard? Jun 30 22:37:36 and same holds for rename-register Jun 30 22:41:45 Tartarus: with gcc 4.5 debugging optimized code is much improved Jun 30 22:41:54 I just debugged uclibc last night :) Jun 30 22:42:01 it could tell me all about locals Jun 30 22:42:16 at anypoint in execution in a function Jun 30 22:42:35 I am all for ggdb3 but it should be managable Jun 30 22:43:19 I think you should make a case for Angstrom folks to change the flag once they're ready to switch to 4.5, and what to switch it to :) Jun 30 22:43:27 and make it to them Jun 30 22:43:40 my playground is minimal Jun 30 22:43:41 atm Jun 30 22:44:02 but most of OE users use angstrom so yes you are right Jun 30 22:44:24 let me experiment with minimal and come up with a good opt combo for debugging Jun 30 22:44:44 * khem is sleep deprived Jun 30 22:45:27 * ant__ is following the debug saga Jun 30 22:45:42 Tartarus: but 500K of uclibc code and 202M of debugging size for it is something to worry dont you agree Jun 30 22:46:10 * ant__ could once build in tmpfs with only 8GB ram Jun 30 22:46:28 ant__: gone are days when people use gcc 2.95 :) Jun 30 22:46:43 he..it was not so long ago Jun 30 22:46:45 now gcc can shower so many bits on you if you ask for it Jun 30 22:46:57 before git native expansion Jun 30 22:47:30 problem with lot of debug info is that static linker (ld) takes a lot longer to link apps Jun 30 22:47:31 and before console-image pollution Jun 30 22:47:42 with extra X packages Jun 30 22:47:45 which will prolong the build time even more Jun 30 22:47:49 yes Jun 30 22:47:51 khem: I do not care about size, seriously Jun 30 22:48:02 Tartarus: how about build time ? Jun 30 22:48:20 For OE? hahah Jun 30 22:48:25 Seriously tho Jun 30 22:48:28 Tartarus: try to debug such a uclibc system on say qemu or some other embedded board Jun 30 22:48:32 khem: it is the same moving from core2 to core4 one year after :/ Jun 30 22:48:38 gdb takes ages to load that uclibc debug info Jun 30 22:49:12 khem: On what vintage of a machine are you working tho? Jun 30 22:49:13 Tartarus: everyone is not blessed with faster machines and lot of hd we should be considerate to them :) Jun 30 22:49:33 * ant__ already bricked 2 HDD building furiously Jun 30 22:49:33 Tartarus: heh I have fairly good machine its a T61 Jun 30 22:49:38 but my hd it 100G Jun 30 22:49:45 'fairly good' ? Jun 30 22:49:46 and I can only have one angstrom image Jun 30 22:50:15 Tartarus: well its a core2duo 2GHz Jun 30 22:50:21 with 2G ram Jun 30 22:50:30 and 100G hd Jun 30 22:50:38 should be enough in modern times Jun 30 22:50:53 we are not paid by intel's and amd's to create killer apps :) Jun 30 22:51:00 heh Jun 30 22:51:28 Well, that's why I care about the case of people debugging stuff where it wasn't built Jun 30 22:51:39 And what's really needed to make that happen, without OE around Jun 30 22:52:04 fair enough Jun 30 22:52:16 Since OE is a full system of the world, rather than an app development environment Jun 30 22:52:24 yes Jun 30 22:52:42 Your laptop should def. be enough to write and debug apps, but no, building a full system on there is a stretch Jun 30 22:52:59 Tartarus: I dont agree :) Jun 30 22:53:00 Heck, my IT-approved developer laptop is a bit of a stretch at times Jun 30 22:53:17 because then I dont have other machine to work on Jun 30 22:53:20 hehe Jun 30 22:53:50 khem: It's the difference between what resources you need to build a single big thing (even gcc/kernel/Eclipse/whatever) vs a full system of software Jun 30 22:53:54 imho Jun 30 22:54:00 I am saying we should not generate stuff blindly just because we can. Jun 30 22:54:05 and that's why I don't care so much about disk space or build time Jun 30 22:54:12 khem: I'll agree there too Jun 30 22:54:17 we should weigh if it really makes sense to have that much of debug info in practice Jun 30 22:54:31 If it adds value, it should be done, if not, not Jun 30 22:54:36 We just disagree about what value is :) Jun 30 22:54:48 And really, you should argue with the angstrom folks to drop down from -ggdb3 Jun 30 22:54:56 Or switch to minimal/micro :) Jun 30 22:55:33 yes for me if I can have same level of debugging with -g and -g3 may be -g3 is a bit more but then it needs triple the size and increased build time and load time in gdb Jun 30 22:55:48 I would think -g is more valuable as practical solution Jun 30 22:56:23 I use to fight star craft with unlimited reasources :) but thats not reality Jun 30 22:57:34 more than one OE user have complained about the problem bkinman ran into with angstrom Jun 30 23:00:01 wrt this channel, that's an argument to make minimal or micro 'better' somehow Jun 30 23:00:29 yes Jun 30 23:00:29 or slugos or whatever. And 'better' might have been better said as more popular or visible or whatever Jun 30 23:00:43 I am going to experiment with improvised debug opts Jun 30 23:01:10 angstrom like OE is a brandname where as minimal/micro are not Jun 30 23:01:28 I am happy that mwester chose eglibc for slugos now Jun 30 23:02:23 minimal is much more appealing than before now Jun 30 23:03:01 I'd say is bleeding-edge wrt toolchain (thx khem :) Jun 30 23:03:19 shr too Jun 30 23:06:30 brb Jun 30 23:21:37 ant__: thx Jun 30 23:21:50 ant__: I am sad that italy exited the WC so early Jun 30 23:22:00 I was rooting for them Jun 30 23:22:01 oh, *that* matter... Jun 30 23:22:22 both winner and runner up of last wc out in first round Jun 30 23:22:35 indeed Jun 30 23:23:08 my bicycle has italian made components from bianci and scattante Jun 30 23:23:22 big brands;) Jun 30 23:23:25 although made in taiwan :) Jun 30 23:23:47 ant__: I like the parts though Jun 30 23:23:53 tdf starts this weekend. woot! Jun 30 23:23:56 my bike runs like a billy goat Jun 30 23:24:09 yeah grg Jun 30 23:24:11 hello Jun 30 23:24:16 morning Jun 30 23:24:37 grg: opkg needs to do better :) Jun 30 23:24:48 khem, of course :) Jun 30 23:25:55 grg: which city are you in Jun 30 23:26:04 khem, adelaide Jun 30 23:26:41 grg: I was once authorized to visit linux.conf.au Jun 30 23:26:52 but could not make it as my daughter was to be born Jun 30 23:27:10 that was my only chance to visit australia thus far :) Jun 30 23:27:16 i've not been to a linux.conf.au Jun 30 23:28:33 always seem to be doing something else at that time of the year... Jun 30 23:28:56 Has any thought been given to randomising the bitbake queue in order to pick up missing build dependencies? Jun 30 23:29:35 * mwester wonders how that would help Jun 30 23:30:30 well, lots of build deps are explicit in the recipe, they just happen to be built because of a rather obscure dependency chain Jun 30 23:30:38 s/are/aren't/ Jun 30 23:31:33 e.g. xserver-kdrive_1.5.3.bb appears to depend on openssl, which is not in its dependency list and it only fails now because i removed openssl as a dependency from my opkg recipe Jun 30 23:32:12 i don't know if xserver-kdrive depends on opkg, but it probably doesnt Jun 30 23:33:01 Ah, ok, so randomizing would have the effect of uncovering these missing dependencies. Jun 30 23:33:08 grg: hmm yeah there are several case where apps build because they find a dependency because someone else pulled it in for them Jun 30 23:33:17 yeah Jun 30 23:33:30 I would think if we had enough BB threads running, that would do the same, no? Jun 30 23:33:48 mwester, probably Jun 30 23:33:58 i've just seen bitbake die from multiple BB threads Jun 30 23:34:07 * mwester goes to search for a 1000-core processor. Jun 30 23:34:09 it was just an idle thought.... Jun 30 23:34:14 I guess we could have a mode in bb where it traps the open syscall and notes what file is being opened Jun 30 23:34:30 that will be best way to track dependencies Jun 30 23:34:46 may be a kernel module or something that talks to bb Jun 30 23:34:57 Custom filesystem Jun 30 23:35:05 could even be a user-space fs Jun 30 23:35:09 this way we will find complete depchain Jun 30 23:35:14 Just has to report file operations Jun 30 23:35:22 and pass them to teh underlying fs Jun 30 23:35:43 commercial products exist that do that already. Jun 30 23:35:45 yeah something like that Jun 30 23:35:49 hmm cool Jun 30 23:36:19 I did not know that Jun 30 23:36:26 I am working on something similar Jun 30 23:36:32 for my paid work :) Jun 30 23:36:45 IBM's ClearCase, and Electric Cloud's emake and electrify products Jun 30 23:37:11 ah cool Jun 30 23:37:16 emake sounds nice Jun 30 23:37:39 (Disclaimer: I worked for the company that originally wrote ClearCase, and my current employer does emake/electrify) Jun 30 23:37:52 grg mweser rational? Jun 30 23:37:56 *g* Jun 30 23:37:57 Atria Jun 30 23:38:06 hm Jun 30 23:38:15 I worked for Rational for a year after they aquired Atria Jun 30 23:38:22 hm hehe Jun 30 23:38:31 They were too big a company, no fun at all. Jun 30 23:38:59 I once had to work around clearcase not avaliable encrypted transport Jun 30 23:39:23 I used the email patch support and gpg Jun 30 23:39:24 mwester: Put OE into emake before? Jun 30 23:39:32 or just usesyou mean? Jun 30 23:40:05 Tartarus: I tried, failed -- electrify is a better fit, but I'm on another project and have had absolutely no time to work on it. Jun 30 23:40:26 k, thanks Jun 30 23:40:35 woglinde: Clearcase was slow without encryption, i can just imagiine how painful that must have been! Jun 30 23:41:16 mwester hm ah I remeber email was the only way we got internetconnection to the other company Jun 30 23:41:36 was around 2001 Jun 30 23:42:32 Tartarus: the issue is that emake would only distribute threads at the makefile level, i.e. do_configure, do_compile tasks. I wanted to be able to distrbiute bb tasks out. I think it can work, but it will be tough. I need to convince my employer that it would be a great learning project and let me try it during working hours. :) Jun 30 23:46:14 heh Jul 01 00:00:18 fun..broken gpe screen calibration...Touch the crosshairs...no crosshairs! Jul 01 00:02:36 :) Jul 01 00:02:37 ant__, yeah, i got that too Jul 01 00:02:46 * buZz tries Jul 01 00:03:08 run xtscal from the command line and you get a libxcb assertion failure Jul 01 00:04:51 libxcb was touched 2 days ago Jul 01 00:06:28 xtscal: xcb_io.c:452: _XReply: Assertion `!dpy->xcb->reply_data' failed Jul 01 00:13:28 same for opie...only you see the crosshairs but you are in a loop :p Jul 01 00:14:39 ok, good night Jul 01 00:16:24 and the XCALIBRATE extension is missing with the 1.5.3 xserver-kdrive recipe... *sigh* **** ENDING LOGGING AT Thu Jul 01 02:59:57 2010