**** BEGIN LOGGING AT Mon Oct 28 02:59:58 2013 Oct 28 08:04:38 good morning Oct 28 09:18:51 morning all Oct 28 09:28:20 morning Oct 28 09:30:02 hi JaMa Oct 28 09:47:17 Some recipes (e.g. the ti-dsp-link) have problems with "out-of-tree" build directories. I had chosen the tmpdir to be at ${TOPDIR}../../build/tmp that leads to errors during build. When adding a python function that calls os.path.abspath on the TMPDIR those problems seem to be fixed, so I wonder if it wouldn't be a good idea to always run abspath on the tmpdir before starting building Oct 28 11:21:17 morning Oct 28 11:27:09 hi otavio Oct 28 11:56:22 howdy! is there some documentation on the best practises concerning a project workflow and SCM? Oct 28 11:56:51 e.g., what do i best keep in my companys SCM, and what is considered throw-away on the developer machine? Oct 28 12:04:09 LetoThe2nd: how do you mean, the best way (tm) is to push anything usefull upstream, and keep everything company specific in your own custom layer Oct 28 12:05:30 jackmitchell: sure, but i'm not yet sure of the totally pragmatic way how to SCM a sppecific product Oct 28 12:05:30 LetoThe2nd: if you mean how do you keep a specific yocto configuration, then the way I do it is to host a local.conf somewhere which everyone can download, then point them to the right versions of upstream layers, plus the company specific layer Oct 28 12:06:41 thats what i meant - just keep the local.conf, and have each developer again set up the layers for his own machine? Oct 28 12:06:52 sounds like pretty much could go wrong there. Oct 28 12:07:11 or rather use something ripped out of angstroms setupscripts? Oct 28 12:07:16 LetoThe2nd: indeed, that's why you usually have someone who controls the build, and dishes out SDK's to the developers Oct 28 12:07:38 you could use an automatic script like the Angstrom one sure Oct 28 12:08:27 hmhm Oct 28 12:09:08 unless your devs are doing build system changes, they don't need the buildsystem Oct 28 12:09:33 and if they are doing build system changes, they should really be able to manage downloading and setting up a few layers Oct 28 12:09:45 of course, but reproductibility is always important Oct 28 12:10:21 if you get the right commits, for all of the layers then it should be reproducable Oct 28 12:10:28 LetoThe2nd, as I understand it (I'm in the same position as you, but haven't yet actually started), the "company layer" also goes in scm, separate from upstream Oct 28 12:10:46 kmacleod: thats obvious, yes. Oct 28 12:10:54 but you're never going to get it complete deterministic unless you use a single build machine Oct 28 12:11:57 ^^ that's probably a bit exaggerated, but on the whole every machine is slightly different Oct 28 12:13:31 yes of course, but basically anybody who needs to be a holiday repleacement or successor of the build system dev should be able to reproduce the functionality without hunting down layers, configs, and all that. Oct 28 12:13:43 IMHO that should be like a checkout - go. Oct 28 12:13:57 LetoThe2nd: in that case, write a small shell script, or use a custom git repo with submodules Oct 28 12:14:19 there is also a combo-layer script in oe-core which does something like that I believe Oct 28 12:14:29 hmhm Oct 28 12:14:59 combo-layer is used to build the yocto git repo with meta-yocto + oe-core etc... Oct 28 12:15:19 will look it up, thanks Oct 28 12:15:27 hi, does one need to explicitly specify IMAGE_BOOTLOADER for a distro? It looks a bit strange at first, but I see something like that in distro config files. Oct 28 13:19:42 Hey, does anyone here know the perl recipe intimately? Oct 28 13:20:43 It seems to me that many core modules, f.x. Time::HiRes, even when RDEPEND'ed upon, does not provide all of the files it should. Oct 28 13:21:06 Such that I need a bbappend for perl that adds PM and .so files to FILES_${PN} for perl in order for core perl modules to work, even when RDEPEND'd upon. Oct 28 13:21:42 Stygia: there aren't automatic dependencies between the core perl modules, no Oct 28 13:22:09 bluelightning, I don't understand what you mean. Oct 28 13:22:30 If I, say, depend upon perl-module-time-hires, HiRes.pm and .so aren't included, unless I manually append to the perl recipe. Oct 28 13:22:55 That'd strike me as a bug, that the perl recipe needs a FILES_${PN}-time-hires += "..." somewhere? Oct 28 13:23:01 that sounds wrong to me yes Oct 28 13:23:55 Hmm, okay. Oct 28 13:24:13 I'll test it, but I'm trying to clean up the recipes I have, so I can commit them without my fudgly perl bbappend. Oct 28 13:24:20 Stygia: er, perl-module-time-hires contains HiRes.pm here Oct 28 13:24:52 Does it also contain auto/Time/HiRes/HiRes.so? Oct 28 13:25:40 yep Oct 28 13:25:42 http://pastebin.com/DFkuZSgY Oct 28 13:26:34 packages-split, of course, implying that it would be part of the image... Oct 28 13:26:39 Well, alright, weird. It must be something on our end, then. Oct 28 13:26:48 You have 5.14.3? Oct 28 13:26:58 Nevermind, I saw your pastebin, you do. Oct 28 13:27:37 PR=1 too. Oct 28 13:28:06 Well, okay, thanks bluelightning. I'll look into what's wrong on my end, but that at least implies I should be able to just sub in the core modules where necessary and it should work. Oct 28 13:28:26 bluelightning, Been fighting a rrdtool recipe for a week, so pardon me for yet again being delayed in pushing those recipes. Oct 28 13:29:26 what does it mean IMAGE_FSTYPES += "tar.bz2"? Oct 28 13:29:39 what is a tar.bz2 filesystem? Oct 28 13:30:20 is that for backup? Oct 28 13:32:54 bluelightning, Maybe you have some advice? The rrdtool recipe has --disable-perl as a EXTRA_OECONF flag, and I happen to need the perl binding. Oct 28 13:33:18 But the perl binding is in a separate Makefiel.PL, which gets called by make during do_compile (making tampering with the Makefile.PL and the makefile it generates difficult). Oct 28 13:34:14 After setting PERLCC and such, I finally got it to actually compile, but I am having trouble using it, at it fails in a way that - as far as I've found - implies improper versions of lbiraries, i.e. my perl binding from rrdtool and the CPAN module that calls it having different version of libraries. Oct 28 13:34:41 Now, I've copied the env variables from the cpan.bbclass and related recipes, to try and get a similar build environment, but I've been failing at this rather dramatically (And am now trying to build the perl debugger). Oct 28 13:35:30 In such a complex situation like this, any idea as how I fruitfully go about debugging it? It passes all the QA checks, so the correct arctitecture is used, and it's also using the same perl version. Further, I removed all HOST include paths and such, but still no luck. Oct 28 13:35:58 Although the module loads fine (i.e. can be 'use'd from perl), it fails once RRD::Simple (A CPAN module) tries to use RRDs.pm, which is provided by the rrdtool recipe once you remove --disable-perl Oct 28 13:39:23 The actual error I get is "Bizarre copy of UNKNOWN In scalar assignment", but from what I've found this implies incompatible lib versions.. Oct 28 13:43:24 Stygia: don't think I can help with that, sorry... Oct 28 13:43:54 lpapp: it's just a bzip2-compressed tar archive of the filesystem Oct 28 13:44:21 bluelightning, Quite fair. Oct 28 13:44:34 bluelightning, Thanks for reading it, though, and for writing me back. :) Oct 28 13:45:15 bluelightning, On another note. I'm still seeing the issue with perl core modules not having their files follow along, as I added perl-module-term-readline to my -dev image, and don't see it's PM or so files. Oct 28 13:45:42 Stygia: if you look in packages-split for perl are the files in there? Oct 28 13:45:46 bluelightning, So something is up, but it could be our end. Bit fishy, though, as this has been like this since pretty much day one of our perl + yocto workings. Oct 28 13:46:50 bluelightning, Yes, they're in there for perl-5.14.3-r1, but not in the image. Oct 28 13:46:54 bluelightning: that is not a filesystem Oct 28 13:47:19 They're currently found in IMAGE_INSTALL_append, but a similar situation I've seen when adding them in RDEPENDS for packages. Oct 28 13:47:40 lpapp: the files that were to make it up if it were to be an actual one, then... Oct 28 13:48:25 Stygia: it sounds like those packages aren't really being installed then Oct 28 13:48:53 Stygia: are they mentioned as being installed in the log.do_rootfs for the image? Oct 28 13:50:00 bluelightning, Yes, they are mentioned as "to be installed: ..." Oct 28 13:50:47 bluelightning, I grep'd for "term-readline", and the only mentions of it seems to be in perl/perl-rdepends..., where it mentions its dependencies. Oct 28 13:51:17 No mention of which files should be included, which does seem to be my issue - the PM and .so files don't get included (Although Term::ReadLine doesn't seem to have an so file, but still). Oct 28 13:51:37 We have been "fixing" (or working around, at least) this by adding the files to FILES_{PN} For a perl bbappend Oct 28 13:52:03 Stygia: you mean pulling the files into the main package? Oct 28 13:52:05 bluelightning, Do the files wind up in your actual image, and not just perl's own packages-split? Oct 28 13:52:16 bluelightning, What? I'm not sure I understand. Oct 28 13:52:27 I have a FILES_${PN} += ... where I add the files I need for perl, yes. Oct 28 13:52:32 As one big blob, this is awful, I know. Oct 28 13:53:19 bluelightning: not sure what you mean. Oct 28 13:53:36 lpapp: I'm not sure what your question is either... Oct 28 13:53:39 It should be FILES_${PN}-term-readline AFAIK, but still, the point is I don't see the PM files and such getting into the actual image when depended upon, only in packages-split for perl. Oct 28 13:53:48 Stygia: if the files end up in packages-split that means they will be in the package in whose directory they appear Oct 28 13:53:59 Stygia: if they aren't in your image it means that package is not being installed Oct 28 13:54:12 bluelightning, Hmm. They're shown in packages-split for perl. Oct 28 13:54:37 Stygia: you mean packages-split/perl ? Oct 28 13:54:59 bluelightning, tmp/work/armv7a-vfp-neon-poky-linux-gnueabi/perl/5.14.3-r5/packages-split/perl-module-term-readline/ Oct 28 13:59:47 the 'r5' thing is my fdault, so never mind that. Oct 28 14:00:39 Hi all, Any idea how to get v4l2src working. Previously I embed v4lc(http://layers.openembedded.org/layerindex/recipe/1157/), but when I do gst-launch it says no element "v4l2src" Oct 28 14:02:26 typo v4l2* Oct 28 14:03:16 bluelightning, Another thing I noted, actually, is that it doesn't seem like perl here RPROVIDES all of the core modules. For example, it only mentions the RDEPENDS of ${PN}-module-term-readline, but trying to using in another packages RDEPENDs gets a complaint that it's not found in the base feeds. Oct 28 14:03:38 zaif: I would suggest confirming that v4l2 support is enabled in gstreamer (or whichever gst-plugins-* is meant to provide it) and that the appropriate plugin package is installed in your image Oct 28 14:05:47 Stygia: the perl module packages are dynamically generated, and the perl recipe does PACKAGES_DYNAMIC += "^perl-module-.*" Oct 28 14:06:05 Stygia: which means that perl-module- should be provided Oct 28 14:06:48 bluelightning, Ah, yea okay. So why, when I add perl-module-term-readline to IMAGE_INSTALL_append do I get told it's not in the base feeds? Because it really isn't, since it's dynamic? Oct 28 14:07:19 if you get that error it means that the package manager can't find it Oct 28 14:07:32 Presumably because it's dynamic? Oct 28 14:07:35 no Oct 28 14:07:50 it just means it doesn't exist as far as the package manager knows Oct 28 14:07:59 that usually means it ended up empty Oct 28 14:08:29 your FILES_${PN} += hacks would actually cause that situation I think Oct 28 14:10:28 bluelightning, Oh? How so? Oct 28 14:10:46 well if the files were forced into the main package they couldn't be picked up by the dynamic packaging Oct 28 14:10:48 bluelightning, It's supposed to add the files dynamically, and never will since I'm touching FILES_${PN} at a higher PR? Oct 28 14:10:51 -> no package produced Oct 28 14:11:16 bluelightning, Huh, alright. Oct 28 14:11:27 but the failure doesn't occur until do_rootfs because the perl recipe has promised, through the PACKAGES_DYNAMIC statement, that that package will exist Oct 28 14:11:31 I'll try and look into removing this bbappend completely, then, and see if maybe we won't have to maintain this anywmore. Oct 28 14:12:26 bluelightning, And right. Thanks! Hopefully we'll get out of this mess, so I can fix my depends to perl-module-X without problems. Oct 28 14:12:42 I wonder how it dynamically knows which files to include, then, but I could read the bbclass, no worries. Oct 28 14:12:52 The perl recipe I know. Oct 28 14:14:55 Stygia: there's a section in the manual on how it works here: http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#handling-optional-module-packaging Oct 28 14:16:48 bluelightning, Sorta weird to see a 'python' function delimited with braces. But, thanks, I"ll read up on that and get to cleaning my env perl-wise. Oct 28 14:21:45 bluelightning, Hmm, right... yea, okay, I definitively see how this works now. Thanks for that link. Oct 28 14:21:51 np Oct 28 14:22:46 bluelightning, I don'tget 100% why my own FILES_${PN} would interfere with this, but I'll cut it out either way. Oct 28 14:24:35 bluelightning: tar.bz2 is a backup image. Oct 28 14:24:48 it is not a filesystem, like jffs, ext3, reiserfs, ubifs, btrfs, etc. Oct 28 14:33:35 lpapp: it's not a backup at all Oct 28 14:34:44 bluelightning: what is it then? Oct 28 14:35:31 it is a bzip2-compressed tar archive of the files that are in the image Oct 28 14:36:11 it is quite frequently used when you wish to install the image onto a device that already contains a filesystem, e.g. a pre-formatted SD card Oct 28 14:36:30 bluelightning: got some hints from this(http://gumstix.8.x6.nabble.com/Re-e-con-camera-modules-td2521910.html) regarding v4l2src. Will try these steps and get back for any help. Oct 28 14:40:24 bluelightning: but why is it exposed as IMAGE FILESYSTEM?\ Oct 28 14:40:29 it is not a filesystem as the rest. Oct 28 14:40:36 that's not important Oct 28 14:47:38 well, for you. Oct 28 14:47:46 it caused a few hours thinking, and it just seems to be a bad API. Oct 28 14:48:00 bad as in confusing. Oct 28 14:48:09 thanks. Oct 28 14:50:53 hi all Oct 28 14:51:05 if I define my own var e.g. USE_TEST = 1 Oct 28 14:51:41 can I use for "${@base_contains("USE_TEST", "yes", "xx", "yy", d)} ? Oct 28 14:52:01 to add list of files to FILES_${PN} ? Oct 28 14:53:20 mbelisko: certainly Oct 28 14:54:42 base_conditional maybe? Oct 28 14:57:26 ant_work: thanks Oct 28 14:58:51 though you probably want to create a separate package instead Oct 28 15:19:53 halstead: did you get a meeting invite from me? Oct 28 15:19:57 for this week? Oct 28 15:28:56 does anyone know if there is a good reason for i586 to be the default 32bit target tune, rather than i686? Oct 28 15:29:18 or even, why i686 doesn't seem to exist as a tune at all? Oct 28 16:36:49 goddamnit, even with my latest sstate reuse fixes, i'm still seeing cases where a massive number of setscenes run, but it runs the real tasks anyway Oct 28 16:36:53 * kergoth grumbles and digs Oct 28 16:43:41 jackmitchell: IMO x32 should be the default 32 bit target, but we are far from that ideal world. Oct 28 16:46:17 lpapp: are there any architectures that can't build for x32, atoms for example? Oct 28 16:46:21 lpapp: that wouldn't work on 32-bit hardware Oct 28 16:47:28 hmm Oct 28 16:48:59 bluelightning: that is why I said ideal world.... I do not think there is much point in 32 bit processors. Oct 28 16:49:39 ok, I was referring specifically to the default 32bit tune Oct 28 19:11:56 I am using gumstix-console-image. After build is success, is there any way to run it with qemu? Oct 28 20:16:38 What's the logic behind incrementing the self-referencing PRINC variable by a value other than 1? I apologize if I missed it somewhere in the docs. Oct 28 20:26:25 crunchex: PR increment is just for keeping track of any updates/patches. Oct 28 20:28:11 crunchex: Explanation: http://www.multitech.net/developer/products/multiconnect-ocg/development/writing-bitbake-recipes/ Oct 28 20:39:04 zaif: I understand the purpose of it, but I see that in some recipes, it increments by 2, for example: PRINC := "${@int(PRINC) + 2}" Oct 28 20:39:25 Why would that number need to be anything but 1? Oct 28 21:09:42 crunchex: in the past, PR had to be bumped manually. when the metadat changed, they bumped it Oct 28 21:09:51 crunchex: it started as 1, then got incremented to 2 when the content was changed again Oct 28 21:09:56 pretty straightforward Oct 28 21:13:34 i'm sure i've had this question answered here before, but can't seem to find it in my logs: how can i find out which recipe created a given file? Oct 28 21:13:50 (in this particular case, it's /etc/init.d/networking) Oct 28 21:17:15 easiest to just use the apckage manager on the target Oct 28 21:17:18 so it depends on what the package manager is Oct 28 21:17:29 alternatively, on the build side you could look at buildhistory Oct 28 21:28:08 kergoth: like, the PM on the device? don't think i've got one (certainly haven't intentionally) Oct 28 21:29:15 kergoth: thanks. so it really doesn't matter, but that was where it was at when it became automatic. Oct 28 21:29:16 anyway, found my answer for this time by doing grep -r on the recipes tree; turns out the forked rpi BSP is the problem Oct 28 21:29:24 plenty of images include it by default :) Oct 28 22:33:33 when i connect a USB cd-rom drive, no /dev/sr0 device appears Oct 28 22:33:48 dmesg shows that it's been recognised, by usb-storage, but no device file appears Oct 28 22:34:04 how should i go about fixing this? something to do with udev's config? Oct 28 22:59:37 Hi Oct 28 23:00:32 How do I get a patch included in 1.5. Does stuff get backporte automatically or should I specify in that patch that the patch is suitable to be applies to dora ? Oct 28 23:06:02 SorenHolm, gett the patch in master, the ask the dora maintainer to backport Oct 28 23:08:58 Crofton|work: ok - where can I find the list of maintainers? Oct 28 23:09:34 I was afraid you would ask that :) Oct 28 23:09:54 Crofton|work: :D Oct 28 23:10:14 I'll ask as a reply to my patch. Thanks. Oct 28 23:10:19 yeah Oct 28 23:10:24 that would be best Oct 29 02:41:08 Is configuring u-boot for a new board beyond yocto's scope? **** ENDING LOGGING AT Tue Oct 29 02:59:58 2013