**** BEGIN LOGGING AT Mon Aug 25 02:59:58 2014 Aug 25 03:15:16 Ok I was able to patch the Makefile.am and stop it from trying to run the program during install. The -c was trying to create a score file during install Aug 25 10:16:29 hi guys Aug 25 10:17:05 I am planning to build uboot.srec and UImage for one R-car board Aug 25 10:17:54 but after loading uboot into the board, it hangs and cannot load uImage Aug 25 10:19:35 what could be the size of uboot.srec ? Aug 25 10:20:16 Now I have around 690k..is that correct? Aug 25 10:20:39 But the device memory map tells around 240KB Aug 25 11:22:14 that is about the right size for u-boot Aug 25 11:22:27 do you get any log output from the board? Aug 25 11:28:32 similar board has uboot.bin of 390k and uboot.map of 140k Aug 25 11:51:58 <_rink> anyone have any clues when my SDK suddenly will nonger install? :-/ Aug 25 11:52:10 <_rink> i'm getting errors relating to python2.7real: Aug 25 11:52:15 <_rink> Setting it up...Could not open /opt/yelem/1.6/sysroots/x86_64-pokysdk-linux/usr/bin/python2.7.real. File used by another process. Aug 25 11:52:18 <_rink> Please make sure you exit all processes that might use any SDK binaries. Aug 25 11:52:20 <_rink> SDK could not be set up. Relocate script failed. Abort! Aug 25 11:52:24 <_rink> but I rm -rf'ed the entire directory before installing Aug 25 12:39:43 hello Aug 25 12:39:52 just repeating the question Aug 25 12:40:17 what about the size of uboot.srec built using yocto and ponky Aug 25 12:40:54 I got as 690KB Aug 25 12:41:07 but my Secure RAM size is only 240KB Aug 25 12:41:23 how can I reduce? Aug 25 13:21:12 JaMa: to answer the question from friday, I built perf for qemu86-64 and linux-yocto-dev on Thursday, had a strange blow up on Friday (different than yours) and now it is building again. Another guy here in the office, also built it fine with no issues and linux-yocto-dev .. so I'm trying to hunt down whatever is racing in the build. Aug 25 13:27:02 zeddii: thanks I'll try to build it again today Aug 25 13:41:38 zeddii: hi there. I'm under the impression there's something wrong in the initramfs images since some time. Aug 25 13:42:11 I already found out armv4 bins are broken but armv5te cpio's init is also failing Aug 25 13:42:26 add to this I have to explicit rdinit= for th emips target... Aug 25 13:43:25 interesting. and by failing, is init just panicing ? Aug 25 13:43:30 I'll try this evening with a fairly simple cpio Aug 25 13:44:19 I have to add more debug but all I get is the kernel logo... Aug 25 13:45:50 this with 3.10, 3.14 and my locval 3.16-vanilla Aug 25 13:46:48 my bad I spent most time with the armv4...finally I got a clue thanks to the "Illegal instruction" yesterday Aug 25 13:52:33 sounds like a tuning gone insane. Aug 25 13:56:03 * LetoThe2nd just misread that as "sounds like a tuna gone insane" Aug 25 13:57:08 sounds like a good movie Aug 25 13:57:26 would compete with "sharknado" Aug 25 14:07:50 zeddii: fails the same here with today's master-next Aug 25 15:33:25 what does ^ mean in layer.conf like BBFILE_PATTERN_layername = "^${LAYERDIR}/" Aug 25 15:33:41 it's a regular expression. Aug 25 15:33:44 google :) Aug 25 15:33:52 got it Aug 25 15:33:53 ^ anchors the expression so it only matches at the beginning of the path Aug 25 15:34:24 Ok was looking at basic-syntax page for bitbake but I was looking at wrong spot. Thanks! Aug 25 15:34:33 np Aug 25 15:35:02 So I should look at python regex instead :) Aug 25 16:40:05 Would there be any objection to changing the oe-init-build-env scripts to add to BB_ENV_EXTRAWHITE rather than setting it? Aug 25 16:40:40 kergoth: what would that be for Aug 25 16:40:57 i'm sick of telling poeople they have to add to it *after* sourcing the setup scripts but not before Aug 25 16:41:02 since it'll blow it away Aug 25 16:45:48 IMO the setup script should inherit less and less from env Aug 25 16:46:13 otherwise its a slippery slope Aug 25 16:49:10 it's a fair point, but in this particular case it's a list of words specifically intended for the user to be able to alter, and it having duplicates would be harmless Aug 25 16:49:48 may be change it to pass it as an argument ? Aug 25 16:50:42 I'd like to see the setup scripts become more capable, the mel ones have a number of useful arguments, including args to add extra layers to the bblayers configuration Aug 25 16:50:46 * kergoth shrugs Aug 25 16:51:31 thats valuable Aug 25 16:52:56 also automatically sorts the bblayers list by BBFILE_PRIORITY, to sync the two priorities up, with an option to override (defaults to lowering the meta-oe priority to 1) Aug 25 16:53:38 you mean in BBLAYERS ? Aug 25 16:53:48 order Aug 25 16:55:32 yeah Aug 25 16:56:01 yes you should submit such improvements Aug 25 16:56:12 k, will add to todo Aug 25 17:33:11 I wonder if MACHINEOVERRIDES should be emptied out for nativesdk recipes Aug 25 18:20:37 How to change fetch system so that it will search in local download folder first Aug 25 18:21:20 you can use the existing MIRRORS mechanism to fetch from a local path, use file:// uri Aug 25 18:21:26 If not found then only it should fetch Aug 25 18:21:38 but if the file is already in DL_DIR with a .done stamp then it'll already not be fetched Aug 25 18:22:03 that's how it already works, except that a .done stamp is written next to the file so we know it was a complete and not interrupted download Aug 25 18:24:21 If i will download file and put in download folder along with a dummyfile x.done then it will bot download from url Aug 25 18:24:44 Let me try this Aug 25 18:34:25 Hi, I am trying to make a recipe that only depends on a subset of the boost packages. Aug 25 18:36:10 Lukas_Bulwahn: not possible unless you divide boost recipe into multiple sub components with their own recipes Aug 25 18:36:34 if you mean the binary package, that's easy enough. but all of boost will be built Aug 25 18:36:36 I tried DEPENDS = "boost-serialization" instead of DEPENDS = "boost", but that did not work. I also tried with RDEPENDS_${PN} = "boost-serialization", but still all boost packages are in the image. Aug 25 18:37:07 DEPENDS are on recipe names and RDEPENDS are on output packages Aug 25 18:37:09 Hi khem and kergoth, the binary package is enough. Aug 25 18:37:49 So, I tried with RDEPENDS_${PN}, but still all boost packages were in the image. Aug 25 18:37:51 RDEPENDS should have worked but then boost-serialization might have deps too Aug 25 18:39:31 So, how are the runtime dependencies of the subpackages determined? Aug 25 18:41:53 RRECOMMENDS_${PN} += "${BOOST_PACKAGES}" Aug 25 18:42:20 Do you have boost somehow mentioned in IMAGE_INSTALL ? Aug 25 18:42:34 if you do then it will pull everything as you see above Aug 25 18:43:35 I do not have boost in IMAGE_INSTALL, only the package that RDEPENDS on 'boost-serialization'. Aug 25 18:45:26 Of course, the package DEPENDS on boost, but I believe that is needed. Aug 25 18:45:46 well then you get what you asked for Aug 25 18:46:41 boost seems to be a meta package so it must be empty just there to get dependencies set right Aug 25 18:47:14 so if you dont need all libraries from it. Remove it and instead mention particular packages from boost that you need Aug 25 18:48:32 khem: https://gist.github.com/kergoth/eb55e1527b6ada7430c0 - work-in-progress, currently a terrible hack shoving host binaries into a target package due to our sdk only having a target sysroot, but could create a nativesdk recipe based on it in theory to get out of tree kernel module support for cross dev with the sdk Aug 25 18:51:39 Erm, maybe itd be more cross-canadian, since its bound to both SDKMACHINE and MACHINE Aug 25 18:51:40 * kergoth shrugs Aug 25 18:52:17 Okay, so I must modify the boost recipe to make the boost package empty? Is that what you suggest, khem? Aug 25 18:52:43 Lukas_Bulwahn: it should be empty otherwise remove the RECOMMEND Aug 25 18:52:54 kergoth: step in right direction Aug 25 18:54:34 Okay, I have to check that. Aug 25 19:02:26 hi. what would be the typical size of systemd binaries? they should link against most libraries dynamically, right? Aug 25 19:04:06 sheesh, lmbench is getting a ton of love on the lists all of a sudden :) Aug 25 19:14:35 Okay, boost is not completely empty. I contains atomic and chronic and I guess through the rdepends the boost subpackages depends on that metapackage, and then all of the boost packages are pulled in. Could that be the reason? Aug 25 19:18:29 hmm... can i assume that something went wrong when none of the systemd binaries are dynamically linked against libdbus? Aug 25 21:08:15 khem: you mentioned that on-target development gets the kernel source tree prepared to do kernel module builds, but that cross didn't. where does that happen, or did i misunderstand? I'm interested in how to cross-compile the scripts directory host stuff Aug 25 21:48:57 * kergoth sighs Aug 25 21:49:09 the mirror protocol checks completely choke on certain completely valid mirror patterns Aug 25 21:49:20 e.g. https?://, (ftp|http)://, etc Aug 25 21:51:33 i can't say i like storing the valid protocols in a list in the python code rather than a variable, either Aug 25 21:51:43 particularly since it's technically possible to add a new bitbake fetcher from the metadata Aug 25 21:51:48 (i've do so before) Aug 25 21:53:57 s/do/done/ Aug 25 21:54:24 I really hate sanity.bbclass. it needs revamping to work similar to insane, wrt WARN_QA/ERROR_QA to control which sanity checks are in use Aug 25 21:57:19 kergoth: yes Aug 25 21:58:00 when you add linux-dev to images then you can effectively develop external modules on target and we bundle a prepared kernel src tree Aug 25 21:58:03 on targett Aug 25 21:58:16 but for cross SDK there is a gap **** BEGIN LOGGING AT Mon Aug 25 22:00:32 2014 Aug 25 22:03:09 khem: thats what confuses me, don't you still need to run 'make scripts' in /usr/src/kernel first, on target? Aug 25 22:06:57 * kergoth hacks on the mirrors protocol sanity test Aug 25 22:07:20 kergoth: the one from master-next? Aug 25 22:09:09 it's on master now and biting me Aug 25 22:09:24 ah, I had some comments on ML too Aug 25 22:09:24 biting me as well **** BEGIN LOGGING AT Mon Aug 25 22:13:09 2014 Aug 25 22:13:10 also switched to using better variable names, and fixed the \n vs \\n thing that was mentiond on the list Aug 25 22:13:21 thanks Aug 25 22:13:55 * kergoth will submit patches shortly Aug 25 22:49:02 JaMa: sent to oe-core, let me know if it works for you, if you would Aug 25 23:07:33 er, somehow i broke my builds, they're hanging at runqueue preparation Aug 25 23:07:35 whatd i do? Aug 25 23:08:18 * kergoth bisects Aug 25 23:14:18 did my sanity changes break sanity? that'd be amusing Aug 25 23:14:27 * kergoth rolls eyes, definitely Monday Aug 25 23:14:35 * kergoth digs Aug 25 23:14:41 * mranostay hands kergoth a beer Aug 25 23:14:54 that's exactly what's needed, good call :) Aug 25 23:28:37 mranostay: a chilled one plz Aug 25 23:30:37 khem: a PBR okay? :P Aug 25 23:32:52 will do man Aug 26 01:27:39 hi all Aug 26 01:27:46 excuse me if this is a really dumb question Aug 26 01:27:57 but I've been googling the heck out of this and can't seem to find anything Aug 26 01:28:11 is it possible to download precompiled binaries? Aug 26 01:28:37 I've tried adding php to my recipe and it keeps failing and all I really need is a .deb file to put on an ARM device that has been built with Yocto Aug 26 01:29:02 its really not my wheelhouse and I just need to get php up and running so I can do some web-development for the device Aug 26 01:32:46 what do you mean by 'adding php to my recipe'? Aug 26 01:56:51 kergoth sorry I don't fully know what i'm talking about Aug 26 01:57:00 maybe I mean my bit-bake file? Aug 26 01:57:12 I'm not psychic. How am I supposed to know what *you* are talking about? Aug 26 01:57:33 no you're right, I wasn't expecting you too Aug 26 01:57:38 there's a php recipe in meta-oe already, are you not using it/ Aug 26 01:58:07 I initially thought what I was saying might have made sense, and then after gauging your response I realized how little I actually know Aug 26 01:59:16 I think what I'm actually doing is creating an image Aug 26 01:59:34 and I tried adding 'php' to the IMAGE_INSTALL variable Aug 26 01:59:41 but it failed due to some issues with mariadb Aug 26 02:00:06 and rather than fighting with it, I was hoping maybe there was such a thing as an existing binary I could just drop onto the device so I could continue with my development Aug 26 02:00:34 ah. no, that's unlikely, unless your'e using angstrom. angstrom heavily leverages binary package feeds, and provides a web ui to generate an image from the feed Aug 26 02:00:41 not sure if they build php for it, though Aug 26 02:02:40 gotcha Aug 26 02:03:01 I'm thinking my best bet is to just manually install any packages I need from the device itself Aug 26 02:03:07 knowing this completely breaks the entire point of yocto **** ENDING LOGGING AT Tue Aug 26 02:59:58 2014