**** BEGIN LOGGING AT Wed Dec 05 02:59:59 2012 Dec 05 03:20:40 Crofton: so it can work with ctrlport? I haven't had a chance to play with it yet, but it should prove to be interesting Dec 05 08:30:59 good morning Dec 05 09:25:04 morning all Dec 05 09:29:30 hi bluelightning, all Dec 05 09:29:36 hi mckoan Dec 05 09:36:48 morning all Dec 05 11:33:36 while updating a recipe, why do people remove the recipe and add new recipe name instead of just renaming it with "git mv" Dec 05 11:34:06 if "git mv" is used, we can see what was changed in recipe file, instead of seeing lots of "+ " lines in diff Dec 05 11:35:12 eren: you're right Dec 05 11:36:21 I guess it would be good to document it (if not), and encourage people Dec 05 11:40:23 mckoan: deleting the recipe and adding a new one isn't required, right? I'm asking whether the current recipe updating is intended Dec 05 11:40:38 s/intended/encouraged/ Dec 05 11:44:17 eren: I think you're right but if everybody do that way may be a reason Dec 05 11:51:14 eren: "git mv" isn't what determines whether something is a move/rename or a delete+add Dec 05 11:51:25 eren: that is figured out dynamically when printing the change Dec 05 11:52:19 git has a -M option for diff, format-patch etc. which will display moves as simple moves and not delete+add Dec 05 11:52:38 bluelightning: hm, that's not related with "git mv" then? Dec 05 11:52:55 nope, that just does a delete + add internally anyway Dec 05 11:53:15 even if we delete+add, we can still get a diff which displays moves as simple Dec 05 11:53:27 sure, just use -M Dec 05 11:53:44 ideally when people send patches they should use -M so it's clear what is actually being done Dec 05 11:53:53 yeah Dec 05 11:53:56 sometimes people forget though Dec 05 11:54:18 while reading commits via e-mail, I get lost frequently Dec 05 11:54:45 for multiple patches the easiest way to do the right thing is to push the branch somewhere and use the create-pull-request and send-pull-request scripts (in the scripts dir) Dec 05 11:55:15 that formats everything in the most acceptable way Dec 05 11:59:48 bluelightning: wouldn't it be good to encourage -M option by branch maintainers or {create,send}-pull-request script?. Of course, reading commits and understanding which is changed is more important for maintainers. If they are happy, it's OK but still, most of the commits cannot be understood via e-mail. Dec 05 12:00:59 eren: we do encourage it, usually if someone sends a patch without -M that involves moves/renames we ask for it to be resent Dec 05 12:01:11 I've just updated http://www.openembedded.org/wiki/How_to_submit_a_patch_to_OpenEmbedded Dec 05 12:01:18 somehow I missed having -M in the example command Dec 05 12:02:04 bluelightning: oh okkie Dec 05 12:02:51 bluelightning: subject prefix is used by patchwork, right? Dec 05 12:03:46 eren: only in that it puts everything in [] together into one (after removing PATCH) - otherwise that's just a convenient label Dec 05 12:04:18 the correct prefix is recommended if you want the appropriate maintainer to notice your patch though Dec 05 12:06:52 bluelightning: got it, thanks for the info Dec 05 12:07:18 * eren brb Dec 05 13:35:39 bluelightning: I've noticed your patches to tidyup qt example application, can you test qt4-4.8.4 with them? I'll have patch to upgrade qt4 soon Dec 05 13:36:28 JaMa: ah, sure Dec 05 14:21:53 hi Dec 05 14:22:07 friend asked me about docs about bootstrapping gcc Dec 05 14:22:14 any suggestions? Dec 05 14:30:04 presumably he already read the gcc manual, right? Dec 05 14:31:14 I think so Dec 05 14:39:18 hrw: http://crosstool-ng.org/hg/crosstool-ng/file/36aa2ae92c29/docs/9%20-%20How%20is%20a%20toolchain%20constructed.txt Dec 05 14:40:46 thx abelloni Dec 05 15:00:54 "ERROR: Function failed: libdwarf: LIC_FILES_CHKSUM points to an invalid file:" - why license check is part of configure instead of being between unpack/configure? Dec 05 15:02:12 agreed it would be clearer as separate task if it doesn't have big overhead Dec 05 15:02:58 configure step can take long time and then fail because of totally unrelated stuff Dec 05 15:06:26 yup and hard to see that, when there is whole log.do_configure printed around it Dec 05 15:11:14 hrw: I think there are certain packages that output a license file based on configure selections, I think that's why it's been done that way Dec 05 15:11:32 granted that is unusual Dec 05 15:11:53 could easily be a separate task that happens after do_configure though Dec 05 15:12:00 then maybe separate task between configure and compile Dec 05 15:13:07 bluelightning: addtask check_license after do_configure? Dec 05 15:13:20 something like that yeah Dec 05 15:13:23 or even s/$/ before do_compile/ Dec 05 15:15:01 yes otherwise long do_compile can run in parallel again Dec 05 15:16:01 years ago it was simpler... Dec 05 15:24:33 hrw: years ago we didn't have squat for license checking at all :) Dec 05 15:25:24 kergoth: I know Dec 05 15:26:19 kergoth: years ago I was asking more stupid questions here as well Dec 05 15:26:29 hehe Dec 05 15:28:59 when there is a file in sysroot - how can I check where from it is? Dec 05 15:29:13 grep sstate-control directory Dec 05 15:29:57 thx Dec 05 15:31:48 ~curse authors assuming CC=BUILD_CC=HOST_CC Dec 05 15:31:49 May you be reincarnated as a Windows XP administrator, authors assuming CC=BUILD_CC=HOST_CC ! Dec 05 15:32:06 old curse got better after so many years Dec 05 15:33:40 Is there a way to clone BitBake somewhere other than inside of oe-core and still use "oe-init-build-env"? Dec 05 15:34:11 as long as bitbake/bin is in PATH it should work Dec 05 15:34:25 iirc there was issue with bitbake/bin in PATH twice leading to fork bomb Dec 05 15:35:24 I don't think that is an issue anymore at least if it is as described Dec 05 15:35:41 JaMa, ok, that's doable. It'd be nicer if I could maybe change something in local.conf. Can I modify PATH in there? Dec 05 15:35:57 if you run oe-init-build-env twice you will get the bitbake/bin directory in PATH twice, but it never caused me any issues Dec 05 15:40:13 bluelightning, JaMa, nope, even with bitbake/bin in my path I get "Error: The bitbake directory (/home/jawilson/src/picoflexor-oe/oe-core/bitbake) does not exist! Please ensure a copy of bitbake exists at this location" after sourcing "oe-init-build-env"... Dec 05 15:40:29 JaMa: having bitbake/bin in the path twice isn't an issue, but having the wrapper path twice can be (though unlikely), but as long as the paths are the same (that is, one isn't via symlink while the other is), the wrappers will succeed at removing their own paths, and it won't be an issue Dec 05 15:40:38 heh Dec 05 15:41:35 kergoth: it was at least year since I've seen this issue for last time, since then I was carefull :) Dec 05 15:41:52 and right, it was because of wrapper path, not bitbake itself Dec 05 15:42:36 at mentor we're using a wrapper around the wrapper, i got bit by that a couple times before i worked out the kinks :) Dec 05 15:43:17 * JaMa is using different init-build-env script Dec 05 15:43:47 * kergoth too Dec 05 15:46:28 aside: i ran across this online: PATH="$(printf "$PATH" | awk -v RS=: -v ORS=: '!arr[$0]++' | sed 's/::/:/g; s/:$//; s/^://')" Dec 05 15:46:32 removes duplicates Dec 05 15:46:37 not often needed, but occasionally useful Dec 05 15:47:45 are you using your own? I may do that, doesn't look terribly complicated Dec 05 15:47:58 I used to use own Dec 05 15:48:14 now I use OE one as I want to have same as on CI Dec 05 15:48:37 ~kill dwarf Dec 05 15:48:37 * ibot shoots a charged anti-proton gun at dwarf Dec 05 15:49:24 mentor uses our own based on the upstream ones due to slightly differing needs (e.g. only including the bsp layer you're targeting in bblayers.conf, avoiding the need to specify machine when operating against a released version of mel, and wanting a customized message shown when sourcing it) Dec 05 15:49:47 might be nice to make the upstream bits a bit more flexible/extensible eventually Dec 05 15:50:03 indeed, patches welcome ;) Dec 05 15:50:49 I'm trying to keep things as unmodified as possible, and I'm using a git repo with my build/conf in it and all the layers as git submodules, unfortunately, submodule in a submodule (oe-core/bitbake) doesn't work... Dec 05 15:53:24 jawilson: FYI you can set BITBAKEDIR to your bitbake location before running oe-init-build-env Dec 05 15:54:03 alternatively you can specify it as the second argument to oe-init-build-env (after the build directory name) Dec 05 15:55:23 bluelightning, huh, either of those don't seem to like it: "Error: The bitbake directory () does not exist! Please ensure a copy of bitbake exists at this location" Dec 05 15:55:40 sigh Dec 05 15:55:43 * bluelightning tests Dec 05 15:56:08 for reference oe-core and bitbake are clone into the same directory Dec 05 15:56:28 so I'm running ". oe-core/oe-init-build-env build/ ${PWD}/bitbake/" Dec 05 15:56:48 kergoth: but you know what? those nearly 9 years ago I would not though that I will use OE to build Facebook stuff Dec 05 15:58:02 ~curse dwarf author Dec 05 15:58:03 May you be reincarnated as a Windows XP administrator, dwarf author ! Dec 05 15:58:12 jawilson: worked here with ../bitbake in either case... Dec 05 15:58:28 hi Dec 05 15:59:52 jawilson: your example also worked Dec 05 16:00:18 hmmm, maybe the script has changed, I'm on denzil, a couple months back Dec 05 16:00:29 the distro I am building uses webkit-gtk_svn as the preferred version, but I have found that the compilation crashes and I would like to change to 1.2.3. I have added the PREFFERRED_VERSION_webkit-gtk ?= "1.2.3" in the my-image.bb but again after bitbake my-image it keeps compaling _svn version Dec 05 16:00:44 what should I change to make work as I want? Dec 05 16:01:24 jawilson: oh, ok... I think the second argument wasn't available in denzil but BITBAKEDIR seems to be Dec 05 16:02:15 well on my checkout even BITBAKEDIR isn't Dec 05 16:02:18 I'll rev-up Dec 05 16:02:27 jawilson: heh... actually there is a reference to it in denzil but it is overridden, so it can't be used with denzil either Dec 05 16:02:42 argh Dec 05 16:03:05 heh Dec 05 16:11:20 hmm proftpd from meta-networking fails do_install Dec 05 16:11:23 | rmdir: failed to remove '/OE/jansa-test/shr-core/tmp-eglibc/work/armv4t-oe-linux-gnueabi/proftpd/1.3.4b-r1/image/usr/lib/proftpd': No such file or directory Dec 05 16:13:00 bluelightning, after the first run, does oe-init-build-env really do anything other than modify PATH with bitbake? Dec 05 16:13:25 quit Dec 05 16:13:43 ah, nm, I see it exporting some things Dec 05 16:15:13 jawilson: well, there are two pieces. set of a couple vars (BB_ENV_EXTRAWHITE, PATH, etc), and then the second piece, creation of the build directory Dec 05 16:15:20 bluelightning: meta/recipes-qt/qt-apps/qmmp_0.6.1.bb seems to be using hosts paths in some cases, will try without your cleanup patch if it's fixed now Dec 05 16:19:00 this is WEIRD: Dec 05 16:19:01 NOTE: recipe udev-182-r2: task do_package_write_ipk: Started Dec 05 16:19:01 ERROR: Error executing a python function in /OE/jansa-test/shr-core/openembedded-core/meta/recipes-core/udev/udev_182.bb: Dec 05 16:19:04 OSError: [Errno 2] No such file or directory: '/OE/jansa-test/shr-core/tmp-eglibc/work/armv4t-oe-linux-gnueabi/systemd/git-r19/packages-split/libgudev' Dec 05 16:19:15 why udev recipe is touching systemd workdir? Dec 05 16:20:42 mm, that's interesting. I've seen a few instances of that sort of thing in the past but never been able to isolate exactly why it happens. Dec 05 16:21:32 jawilson: it sets BUILDDIR I think Dec 05 16:22:09 I think I'm just going to use a symlink... Dec 05 16:22:19 JaMa: I wouldn't imagine I've somehow fixed that, I wasn't aware of the issue... Dec 05 16:39:44 bluelightning: that fix for proftpd isn't enough imho.. ${libdir}/proftpd does not exist Dec 05 16:39:51 so it should be in "if" too Dec 05 16:41:16 JaMa: it fixes the failure here... Dec 05 16:43:00 can you add both in if? Dec 05 16:43:04 why? Dec 05 16:43:08 17:11:27 < JaMa> | rmdir: failed to remove '/OE/jansa-test/shr-core/tmp-eglibc/work/armv4t-oe-linux-gnueabi/proftpd/1.3.4b-r1/image/usr/lib/proftpd': No such file or directory Dec 05 16:44:47 why am I not seeing that failure here? Dec 05 16:45:08 no idea Dec 05 16:45:23 so is that the error you get before the patch I just sent or after? Dec 05 16:45:25 can you grep /lib/proftpd in temp.log? Dec 05 16:45:38 before, but your patch does not seem to fix this Dec 05 16:45:42 no, it does Dec 05 16:45:53 we are no longer trying to delete that directory twice Dec 05 16:46:11 since ${libexecdir} is ${libdir}/proftpd as well Dec 05 16:46:38 ah I see Dec 05 16:46:47 sorry for noise then and thanks for fix Dec 05 16:46:51 np Dec 05 16:47:54 I wonder if this was actually the correct fix, maybe it ought to be -d ${libdir} Dec 05 16:48:02 er Dec 05 16:48:09 -d ${libexecdir} Dec 05 16:48:37 yup that would work even for custom layout Dec 05 16:58:50 zeddii, ant_work: thank you for your time, I have to go now, I'll be back later or tomorrow Dec 05 16:58:57 have a nice rest of the day Dec 05 17:04:03 JaMa: sent, thanks for prompting me to fix it properly :) Dec 05 17:11:09 stamps/foo.do_configure.sigdata is created usually at the same time asi stamps/foo.do_configure.hash? I'm trying to find corresponding stamps for times of segfaults, but with rm_work original stamps are removed and only sigdata kept :/ Dec 05 17:12:13 JaMa: it's not rm_work that deletes the files, it's the stamp removal code Dec 05 17:12:29 AIUI it was a bug you filed that got Richard to make it do that more reliably ... ;) Dec 05 17:13:18 too reliable now? :))) Dec 05 17:13:37 well, it was deleting the sigdata files until recently, that was a bit too much Dec 05 17:13:49 at least if you wanted bitbake-diffsigs -t to work Dec 05 17:14:18 pity that cooker log does not have timestamps in it :) Dec 05 17:15:45 I think it would be fairly easy to tweak the logger such that it does Dec 05 17:15:46 and rm_work does remove some stamps, doesn't it? and converts some others to _setscene Dec 05 17:16:01 I haven't looked at the code recently Dec 05 17:16:19 bluelightning: true, I just don't want to run bitbake world again after dumping sstate-cache and tmpdir :) Dec 05 17:16:49 NOTE: Running task 4360 of 22159 Dec 05 17:16:51 do you suspect some corresponding event to the crash? Dec 05 17:17:00 otherwise I'm not sure what the timestamp would help with Dec 05 17:17:17 it's for sgw for this bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=3232 Dec 05 17:17:25 to find some smaller reproducers then webkit-efl Dec 05 17:18:08 hmm Dec 05 17:18:13 could you use the pid perhaps? Dec 05 17:18:26 oh but you have enabled rm_work... Dec 05 17:18:40 hmm right, temp/log* are still there Dec 05 17:18:57 ah yes so that could be one way to link it Dec 05 17:19:14 and they have "right" creation time too Dec 05 17:19:16 thanks Dec 05 17:19:56 pid of autom4te probably won't be listed in log, but log.do_configure timestamps should help Dec 05 18:06:16 jawilson, ./oe-biuid-iit ../build ../bitbake is what I use Dec 05 18:06:55 Crofton, yeah, but it looks like that's not supported in the denzil branch Dec 05 18:07:26 denzil, is hat old? Dec 05 18:07:34 * Crofton gets all these names mixed up Dec 05 18:09:06 Crofton: 1.2 Dec 05 18:09:26 anyway, I know it works in modern bitbake :), I use it all the time Dec 05 18:16:42 it looks like danny is new enough, so I'll have to upgrade sooner or later Dec 05 18:18:02 yeah Dec 05 18:18:09 I would try and do that Dec 05 21:59:54 hi Dec 05 21:59:58 the distro I am building uses webkit-gtk_svn as the preferred version, but I have found that the compilation crashes and I would like to change to 1.2.3. I have added the PREFFERRED_VERSION_webkit-gtk ?= "1.2.3" in the my-image.bb but again after bitbake my-image it keeps compaling _svn version Dec 05 22:00:03 what should I change to make work as I want? Dec 05 22:00:50 add recipe for 1.2.3 version first Dec 05 22:02:50 it is currently there Dec 05 22:03:28 org.openembedded.dev/recipes/webkit/webkit-gtk_1.2.3.bb Dec 05 22:05:07 well my-image should not select versions of anything.. that's distro policy Dec 05 22:06:05 aham Dec 05 22:07:00 ok thx, going to add it conf/distro Dec 05 22:09:41 JaMa|Skiing, where are you skiing? Dec 05 22:13:46 uhmm I am modifying the omap3-desktop-image.bb to create a custom image, but I can't see where the DISTRO used is set... where is the place where the distro is configured? I can't see it in conf/machine/overo.conf either Dec 05 22:13:55 * mr_science forsees bitbake and a tree merging... Dec 05 22:19:54 Crofton: Austria Dec 05 22:20:00 cool Dec 05 22:20:14 Several years ago we went to St Anton for a few days Dec 06 02:43:45 hey. how can i stop bitbake/oe/whatever from converting lib-only packages to packages named "libqtmultimediakit1" ? **** ENDING LOGGING AT Thu Dec 06 02:59:58 2012