**** BEGIN LOGGING AT Wed Jun 25 02:59:58 2014 Jun 25 05:18:25 wb Jefro Jun 25 06:49:55 Hi, I have a qestion about “native” packages. What are these type of packages? And what is their difference? For example “readline-native & readline” Or “binutils-natve & binutils-cross“ Jun 25 07:23:04 morning all Jun 25 07:23:21 sm_: native recipes are built for the build host (the machine the build is running on itself) Jun 25 07:47:05 bluelightning: thanks. so U mean that it's (native recipe) depends on host configuration and machine specification? Jun 25 07:48:27 sm_: to some degree yes, but we often build other native dependencies to avoid having to install / set up too much on the host (as well as in some cases to configure or patch them appropriately for the build system) Jun 25 07:53:41 bluelightning: so what will happened to cross compile when you install them on the host? Jun 25 08:18:05 good morning Jun 25 08:31:51 sm_: sorry, I'm not sure I understand the question Jun 25 08:32:11 sm_: native recipes are built with the build host's compiler Jun 25 09:32:20 bluelightning: wrty [yocto] File system for alternate partition msg Jun 25 09:32:38 " there’s cruft that I don’t really need in the FS" Jun 25 09:33:01 same happens with i.e. kexecboot initramfs Jun 25 09:33:28 ant_work: yes coincidentally I was just trying that out here after reading that Jun 25 09:33:33 ;) Jun 25 09:33:55 not much space tbh Jun 25 09:34:10 I can't actually seem to create an empty image as Seth has, not without any errors anyway Jun 25 09:34:47 worth opening a bug? Jun 25 09:35:10 could be yes Jun 25 09:36:23 we could empty the IMAGE_INSTALL of the initramfs recipe Jun 25 09:36:28 to test Jun 25 09:38:55 that's basically what I'm doing Jun 25 09:39:01 well, I did PACKAGE_INSTALL Jun 25 09:39:35 initramfs-kexecboot-image.bb has already some anti-cruft countermeasures Jun 25 09:39:57 not enough apparently... Jun 25 09:40:31 though, the modules stuff could come from some machine EXTRA Jun 25 09:41:41 so i was wondering yesterday if we need another layer of image class which just has the "make an image from this directory" logic in Jun 25 09:42:13 so if you really just want a few packages, or a few files, you can use that directly instead of battling against image.bbclass Jun 25 09:43:27 rburton: might be useful, but there is one benefit to constructing your image from packages - it encourages practices that enable license tracking Jun 25 09:43:31 rburton, honestly with little adjustments we are not far, only opkg, initscripts and k. modules files Jun 25 09:45:28 actually setting IMAGE_INSTALL = "" rather than PACKAGE_INSTALL = "" worked Jun 25 09:45:35 here are the results: http://pastebin.com/QWJ2BeW2 Jun 25 09:46:24 yes, 3 sources Jun 25 09:47:18 some code is sneaking in unconditionally Jun 25 09:54:14 well, I'm just testing it now but I think setting LDCONFIGDEPEND = "" should kill off the /lib/modules/ stuff Jun 25 09:54:29 I can't help feeling that ought to be automatic though Jun 25 09:55:00 hmm, that didn't do it... strange Jun 25 10:00:09 afair it did the trick long ago when it was added Jun 25 11:01:52 bluelightning: thank you for responding Jun 25 11:36:27 sm_: no problem Jun 25 12:49:22 denix: After configuring arago using arago-daisy-config.txt, I attempted a build of arago-amsdk-image. I got the following error: build/arago-tmp-external-linaro-toolchain/work/am335x_evm-oe-linux-gnueabi/omap3-sgx-modules/5.01.01.01-r22a/Graphics_SDK_5_01_01_01/GFX_Linux_KM/services4/3rdparty/dc_ti335x_linux/omaplfb_linux.c:134:27: fatal error: omapfb/omapfb.h: No such file or directory Jun 25 12:49:27 denix: Have you seen this before? Jun 25 13:57:39 has anyone seen these PAM sshd errors before with daisy: http://pastebin.com/ysjQ1SBz ? Jun 25 13:58:32 part of it seems that yocto doesn't install /etc/pam.d/system-auth at all Jun 25 14:55:55 Anyone have any suggestions about what I can do about connmand[468]: eth0 {add} route 0.0.0.0 gw 10.2.7.1 scope 0 followed by a {del} of the same thing then followed by an add again? Thousands a second. Jun 25 19:55:42 trying to build for allwinner a10 and have set PREFERRED_VERSION_gcc-cross ?= "linaro-4.8.1" in the conf/local.conf, but, bitbake keeps complaining that "preferred version linaro-4.8.1 of gcc-cross-initial not available (for item virtual/arm-poky-linux-gnueabi-gcc-initial)" and then tells me "versions of gcc-cross-initial available: 4.7.2 4.8.1" Jun 25 19:56:10 the name you use must match the name the system is expecting.. Jun 25 19:56:16 gcc-cross becoems gcc-cross- Jun 25 19:56:28 oh, thanks. doh Jun 25 19:56:35 so you need to do PREFERRED_VERSION_gcc-cross-x86_64 = ... (or similar) Jun 25 19:56:43 too many hours staring at screen. Jun 25 19:56:53 * T0mW goes for coffee Jun 25 19:57:03 easiest way to know what that is.... run bitbake gcc-cross and look at the sugegstions.. ;) Jun 25 19:57:24 (I ran into this issue a few hours ago for a different reason..) Jun 25 20:36:49 hello everybody! I got a question on the usage of the SDK's sysroot. Is it normally acceptable to chroot into the sysroot and install packages through a pkg manager (smart, for example)? Jun 25 20:37:21 no that isn't really supported Jun 25 20:37:39 i see Jun 25 20:38:00 the SDK is a snapshot of the base system and compiler configuration... if you want to provide different libraries or combinations of components, you should rebuild teh SDK.. Jun 25 20:38:10 what's the recommended way to install packages (such as opencv) in the sysroot? Jun 25 20:38:27 we've had requests in the past for package level upgrade of the sysroot, it might be possible.. but there are issues with that approach. Jun 25 20:38:45 (I don't think we can do native sdk components at all due to the relocation code) Jun 25 20:38:58 depends on how you are generating the SDK.. Jun 25 20:38:58 i'd like to my users to be able to simply d/l the baseline sdk/sysroot and pull necessary packages and libraries from my repository Jun 25 20:39:12 if you are using the image recipe, just add it to your image.. since you'll need the libraries to run code that are linked against.. Jun 25 20:39:39 if you are using a dedicated toolchain recipe, i.e. meta-toolchain then you can add them directly via your configuration Jun 25 20:39:52 (image recipes use the bitbake -c populate_sdk Jun 25 20:40:17 I prefer the populate_sdk / image approach since I know if it's available as a runtime library in my image, it's available on the SDK's sysroot Jun 25 20:40:18 yes -c popoulate_sdk is how i got the sdk - Jun 25 20:41:02 so say i generate sdk, and realized i didn't have opencv library included in my recipe, so it's not in the sdk sysroot, i need to rerun the populate_sdk command? Jun 25 20:41:04 so assuming it's a runtime/dev time component pair.. then add it to your image.. IMAGE_INSTALL_append = " (note the leading space) in your local.conf is one way Jun 25 20:41:17 hat is currently the best approach, yes Jun 25 20:41:45 i see - and setting up a pcakge manager may be possible - but not proven? Jun 25 20:41:49 the first build of the SDK is fairly slow, subsequent builds (in the same build directory) should be signigicantly faster.. Jun 25 20:42:05 yes, it may be possible (only for the target components), but as far as I know, nobody has gotten it to work.. Jun 25 20:42:11 i see - i was hoping to let my users avoid building an entire image Jun 25 20:42:37 hmm Jun 25 20:42:42 to make it work yourself, you would need to follow the steps in the image creation for using the package manager you selected.. that means environment configuration and related -- plus include the package manager binaries into the SDK Jun 25 20:42:51 can you take a look at section 4.2 in http://www.yoctoproject.org/docs/1.6/adt-manual/adt-manual.html Jun 25 20:43:06 i think the section 4.2 is exactly what i need - except it's opkg Jun 25 20:43:51 I certainly put that in the category of should work, but I've not seen people use it regularly.. Jun 25 20:43:59 (I also use RPM packages primarily and not IPK) Jun 25 20:44:26 it's possible some of those instructions are older, from before we had the image recipe SDKs.. I'm not sure Jun 25 20:44:45 n either case though (opkg or rpm) the package manager components need to be in the SDK for any of that stuff to have a chnace of working Jun 25 20:45:07 (sorry my network is saturated currently, and I keep dropping characters) Jun 25 20:45:43 sd card creation is a mess Jun 25 20:46:48 hmm i guess i need to try it first Jun 25 20:47:05 adma_ -- if you are using IPK packages, and that section does not work.. please open a defect in the Yocto Project bugzilla so we can fix it.. Jun 25 20:47:25 i see the smart binary is included in the sysroot Jun 25 20:47:27 if it does work, let me know. :) and if you are using RPM, there are already enhancement requests to make it work.. but so far we've not implemented htem Jun 25 20:47:38 yes you are right, that's the firs thing i should really try Jun 25 20:47:45 i see Jun 25 20:47:57 alright, let me try the sdk doc method, and will let you know ;) Jun 25 20:48:02 thanks for your insight! Jun 25 20:48:23 no problem.. (I'm theoretically responsible for fixing this if it is broken BTW..) Jun 25 20:48:44 hahahah then thanks for keeping things in order Jun 25 20:53:25 is there a way to query bitbake for which package is pulling in a specific dependency? Jun 25 20:54:29 staylor: you can look at the dependency graph - bitbake -g -u depexp Jun 25 20:55:07 staylor if the graph doesn't show you (or is too complex for an item) you can blacklist the item and you'll get an error that shows you why it was pulled in.. Jun 25 20:55:36 I've blacklisted features, what would the variable name be for blacklisting a specific package? Jun 25 20:55:49 add 'blacklist' to your USER_CLASSES.. then add an entry: PNBLACKLIST[recipe] = "Reason" Jun 25 20:56:13 when you try to build 'recipe' it will of course tell you the 'Reason', but an error will occur showing you what can't build due to the blacklist.. :) Jun 25 20:56:24 'er.. sorry when you -build- (not build the recipe)... Jun 25 20:56:50 i.e. bitbake core-image-lsb -- if you blacklist 'bash' you should see everything that requires bash in that image Jun 25 20:58:40 I see, perfect thank you Jun 25 20:59:06 Hmm, that doesn't show the dependency that pulled it in (and now is failing) Jun 25 20:59:30 scratch that last comment my mistake Jun 25 20:59:34 is the dependency in your image, but not in the graph, then it's a recommended package Jun 25 21:04:11 Yeah that worked great, so packagegroup-qt5-toolchain-target requires qtwebkit *rolls-eyes* Jun 25 21:05:16 staylor: bitbake -g Jun 25 21:11:35 it appears that oe_runmake has its' own ideas of what LDFLAGS should be? I tried export LDFLAGS="blah blah" in the do_compile of the recipe, but nothing changes in the run.do_compile. Suggestions? Documents? Jun 25 21:12:22 usually you pass options using the EXTRA_OEMAKE ... one 'option' could be 'LDFLAGS="..."' Jun 25 21:12:41 note you really don't want to deviate from the system ldflags, or you are likely to break something.. (hashes, abi, etc..) Jun 25 21:44:55 T0mW: oe_runmake just runs make with ${EXTRA_OEMAKE}, it has no internal knowledge of LDFLAGS. that said, by default in non-autotools recipes extra_oeconf includes -e, which makes the makefile prefer our exported variables rather than what's defined in the makefile. as fray says, it's better to explicitly pass in things than rely on the -e behavior, but even then, the makefile has to be constructed so as to know how to handle such vars being Jun 25 21:44:56 passed in. either split what ld flags are required to build from things like -L and optimization, or use 'override LDFLAGS +=' to append to ldflags the required bits even when ldflags is passed in. cflags is the same way, separate things like -I. and -I $(topdir)/src from things like -O2 -Wall Jun 25 21:45:30 every buildsystem for a piece of software should consider how to obey the things the distro wants it to, if they want integration with that buildsystem, rpm for example is no different :) Jun 25 22:48:47 zecke, you awake? Jun 25 23:29:23 kergoth: thanks Jun 25 23:47:14 yocti, Jun 25 23:47:24 yocti, 6481 Jun 25 23:47:25 Crofton: Error: "6481" is not a valid command. Jun 25 23:47:29 yocti, help Jun 25 23:47:29 Crofton: (help [] []) -- This command gives a useful description of what does. is only necessary if the command is in more than one plugin. Jun 25 23:47:37 yocti, help bugzilla Jun 25 23:47:37 Crofton: Error: There is no command "bugzilla". Jun 25 23:47:51 yocti, list Jun 25 23:47:51 Crofton: Admin, Bugzilla, Channel, ChannelLogger, Config, Karma, Misc, Owner, Praise, Seen, Services, and User Jun 25 23:48:00 yocti, bugzilla Jun 25 23:48:01 Crofton: Error: "bugzilla" is not a valid command. Jun 25 23:48:05 yocti, bugzilla help Jun 25 23:48:06 Crofton: Error: The "Bugzilla" plugin is loaded, but there is no command named "help" in it. Try "list Bugzilla" to see the commands in the "Bugzilla" plugin. Jun 25 23:48:17 yocti, list bugzilla Jun 25 23:48:17 Crofton: add, attachment, bug, and query Jun 25 23:48:26 yocti, bug 6481 Jun 25 23:48:27 Crofton: Bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=6481 major, Undecided, ---, richard.purdie, NEW , wic produces traceback Jun 25 23:48:59 summary, Monkey types wic. Error message not helpful :) **** ENDING LOGGING AT Thu Jun 26 02:59:58 2014