**** BEGIN LOGGING AT Thu Nov 20 02:59:58 2014 Nov 20 12:12:13 has anyone used cgexec (or some ther cgroups tool) to contain OE build so the desktop keeps responsive while compile is going on? Nov 20 12:13:23 basicly to be a bit more assertive at being nice than "nice" command Nov 20 12:48:50 Python.h: No such file or directory Nov 20 12:48:58 is there a proper way to fix this? Nov 20 13:34:33 yeah, use the proper include dir :) Nov 20 13:35:13 *g* Nov 20 13:40:16 hmm, pyqt has made some build system changes Nov 20 13:40:21 * Crofton|work crosses fingers Nov 20 13:41:22 suihkulokki, I carefully tune parallel make and bbthreads so there is a core for web browsing:) Nov 20 13:42:31 ... with lynx Nov 20 13:42:39 :) Nov 20 13:47:28 lynx? who would use such a thing? Nov 20 13:47:34 links is the best surely Nov 20 13:47:37 ;) Nov 20 13:48:20 suihkulokki: the default fedora setup seems to work 'OK' for that Nov 20 13:48:39 suihkulokki: the only issue is that the google hangout plugin won't get enough CPU time to do VP8 Nov 20 13:48:54 (which seems to nee 60% on 4 cores) Nov 20 13:49:09 : error: 'class QWidget' has no member named 'setAccessibleDescription' Nov 20 13:49:09 | sipCpp->setAccessibleDescription(*a0); Nov 20 13:49:23 any Qt gurus handy? Nov 20 13:57:48 why wouldn't qwidget have this method Nov 20 13:59:30 koen: guess why I was asking this right after the meeting :) Nov 20 14:06:10 Something seem to be breaking in Dizzy on RPM backend Nov 20 14:06:11 | DEBUG: Executing shell function BUILDSPEC Nov 20 14:06:11 | error: line 61: Illegal char '-' in EVR: Requires: libegl-mx6 >= 3.10.17-1.0.1-hfp Nov 20 14:06:11 | error: Package has no %description: piglit-1.0+gitr0+bbeff5d21b-r0.cortexa9hf_vfp_neon_mx6sl Nov 20 14:06:14 | Building target platforms: cortexa9hf_vfp_neon_mx6sl-poky-linux-gnueabi Nov 20 14:06:22 but it used to work just fine Nov 20 14:06:25 clues? Nov 20 14:22:43 Hi quys, I have a simple question. (Angstrom 2013.12) I have mesa-gl_9.1.6.bb. I can run "bitbake mesa-gl" and it's working. But i cannot add mesa-gl to my image.bb file. When i add to IMAGE_INSTALL += mesa-gl and the bitbake my image, i have an installation error. How to add mesa to my image? Nov 20 14:23:58 spaszkoPL: check what packages that recipe actually produces, you can look under packages-split under the work directory for the recipe to find out Nov 20 14:24:37 otavio: haven't seen that before... Nov 20 14:24:43 otavio: can you please file a bug? Nov 20 14:26:20 bluelightning: work directory for the recipe? You mean directory where is .bb file or deploy directory? Nov 20 14:27:08 spaszkoPL: I mean tmp/work/.../mesa-gl/9.1.6-rX/ Nov 20 14:27:29 spaszkoPL: you can get the full path by running bitbake -e mesa-gl | grep ^WORKDIR= Nov 20 14:28:33 thx a lot, i will investigate it :) Nov 20 14:30:20 bluelightning: I am trying to understand the issue so we can file it Nov 20 14:30:37 bluelightning: do you have any hint how I can gather more info? Nov 20 14:31:13 otavio: maybe the spec file that has been written out to the workdir for rpmbuild to use? Nov 20 14:31:32 bluelightning: where I can find it? Nov 20 14:31:50 otavio: libegl's workdir... Nov 20 14:32:10 Ok. Soo it produces a lot of packages, for example : libgl-mesa libegl-mesa libglapi-dev or even mesa-gl, but mesa-gl is empty inside. Nov 20 14:33:34 spaszkoPL: I guess it depends on which one you actually need for whatever you're doing Nov 20 14:34:10 bluelightning: http://privatepaste.com/633b43cee9 Nov 20 14:34:37 So instead of adding mesa-gl to IMAGE_INSTALL you suggest adding one of names of this packages under packages-split directory? Nov 20 14:35:36 spaszkoPL: correct yes Nov 20 14:36:00 ok, i will try it. Nov 20 14:37:04 In fact I'm trying to cross compile QT with QPA=eglfs for AM3352 which don't have a SGX graphics, soo i need mesa. Nov 20 14:39:15 spaszkoPL: I think you might want libegl-mesa in that case Nov 20 14:43:40 bluelightning: any hint where the culprit might be? Nov 20 14:46:01 otavio: no, in fact I can't even see a reference to that version string in the file Nov 20 14:46:26 otavio: oh wait, this would be in the piglit specfile, sorry I led you astray Nov 20 14:57:47 interesting, the pyqt configure-ng.py script may be a big help for us Nov 20 14:57:52 testing underway Nov 20 15:20:29 configure-ng.py ? Nov 20 15:33:09 bluelightning: oh ok; let me get it Nov 20 17:13:32 Quick question: We have OE build and dev machines that run different Linux distros (Mint 17 and RHEL 7), which currently makes -native packages rebuild needlessly Nov 20 17:13:45 (since the sstate cache from one is not seen as usable on the other). Nov 20 17:14:05 We've been told that we can use "sstate cache mapping tables with mirror syntax" to make this work Nov 20 17:14:23 e.g. declaring that the two distros produce compatible binaries. Nov 20 17:14:31 What's the syntax for making this happen? Nov 20 17:15:01 you guys are masochists :) Nov 20 17:21:05 agreed, but we'd like to be masochists with faster builds! Nov 20 17:23:01 use small chroot identical on all machines just for builds and you will get faster and more reliable builds :) Nov 20 17:27:10 anybody need some dead? http://www.dead.net/30daysofdead Nov 20 17:28:13 * nerdboy preparing for actual bbb project Nov 20 17:29:13 denix: is bbb kernel stuff fully mainlined now? i can just use yocto-3.17 or whatever is latest? Nov 20 17:32:35 bluelightning: https://www.dropbox.com/sh/pgi8l77k6kzb2qn/AABKUv-zgkgU-opMRx5Lm6qEa?dl=0 Nov 20 17:57:21 bluelightning: any hint? Nov 20 19:29:42 In oe-classic, is there a way to enable/disable stripping of binaries? Nov 20 19:31:46 beshelto: SSTATE_MIRRORS uses the same syntax as PREMIRRORS. regex matches Nov 20 19:33:40 beshelto: https://github.com/MentorEmbedded/meta-mentor/blob/master/meta-mel/classes/sstate-mirror-sites.bbclass is what we use for this at mentor. we add distros we know are compatible with the current one to SSTATE_MIRROR_DISTROS and urls to mirrors to SSTATE_MIRROR_SITES, and it combines them to construct SSTATE_MIRRORS. see https://github.com/MentorEmbedded/meta-mentor/blob/master/meta-mel/conf/distro/include/sstate.inc#L16-L40 and https://github.com/Mento Nov 20 19:33:40 rEmbedded/meta-mentor/blob/master/conf/local.conf.sample#L116-L117 Nov 20 19:33:55 beshelto: obviously you could just construct your SSTATE_MIRRORS manually, this is just how we do it Nov 20 19:34:44 we do our automated builds on rhel6, and all the engineers run who knows what distros, so with that configuration the rhel6 native sstates are used everywhere Nov 20 19:36:34 kergoth: Thanks! Nov 20 19:38:57 beshelto: for a concrete example: SSTATE_MIRRORS = "file://RedHatEnterprise-6 http://easource.alm.mentorg.com/sstate/RedHatEnterprise-6;downloadfilename=PATH\n file://.* http://easource.alm.mentorg.com/sstate/PATH;downloadfilename=PATH\n " Nov 20 19:39:09 erm, ignore the url, should've clipped that Nov 20 19:39:28 but you can see the idea. the second entry is a general mirror download, the first changes the distro portion to a different distro Nov 20 19:39:40 erm, no Nov 20 19:39:53 you want file://${NATIVELSBSTRING} for the first part Nov 20 19:40:09 then /SomeAlternativeNATIVELSBSTRING for the second part Nov 20 19:40:33 so it replaces the distro part of the sstate path with the remote with a different distro Nov 20 19:42:01 accidentally pasted the expanded, not unexpanded value, the first time :) Nov 20 19:42:24 beshelto: of course, there are two other options: use a chroot, or use richard's uninative stuff (not 100% where it needs to be yet, but promising) Nov 20 19:42:40 i set up aliases with schroot, so whenever i run bitbake, it really runs it in a pristine minimal chroot Nov 20 19:42:47 kergoth: sure, and we plan to migrate to uninative as soon as it's available. Nov 20 19:42:50 In oe-classic, is there a way to enable/disable stripping of binaries? Nov 20 19:43:20 awozniak: pretty sure even in oe-classic we did splitting of -dbg packages, in which case there's no point to disabling it, just install the -dbg packages to get the debug symbols Nov 20 19:43:57 but yes, see INHIBIT_PACKAGE_STRIP if you really want to do that Nov 20 19:44:02 same var in oe-core as in oe-classic Nov 20 19:44:20 My sandbox builds are much bigger than the builds on my production machine. Looking into it, the difference is whether or not the build produces stripped binaries. Nov 20 19:44:38 I'm wondering If I've tweaked some local config that caused the change. Nov 20 19:47:53 thanks kergoth Nov 20 21:30:49 how do I amke a rebuoild remove a work dir? Nov 20 21:31:01 cleansstate seems to leave stuff behind Nov 20 21:40:04 Crofton|work: clean and cleansstate both clear out workdir. cleansstate == clean + remove sstate binaries Nov 20 21:40:16 i think they leave behind the temp dir with the logs and whatnot, but other than that.. Nov 20 21:40:19 * kergoth shrugs Nov 20 21:41:37 hmm Nov 20 21:42:11 at least, thats the intent, if not, i'd open a bug :) Nov 20 21:49:13 direcotry is there, but empty **** ENDING LOGGING AT Fri Nov 21 02:59:58 2014