**** BEGIN LOGGING AT Mon Mar 16 02:59:58 2015 Mar 16 08:56:30 good morning Mar 16 08:58:03 g'morning Mar 16 09:02:01 pcmanfm is incompatible with my host. Is that some kind of radio GUI or what. hmz Mar 16 09:02:22 I was wondering if someone could solve the mistery of the kernel location for me? How can I quickly compile a kernel change and test it by manually copying the zImage file? Mar 16 09:02:23 graphical filemanager. oh dear Mar 16 09:02:31 why do you need one? Mar 16 09:03:29 if I replace /boot/zImage and what it points at it simply has no effect. Even after deleting the files from boot, u-boot still manages to boot. Mar 16 09:03:44 pompomJuice: depends on your uboot setup Mar 16 09:04:02 It's so confusing Mar 16 09:05:15 When I try to manually build and do something like: -c compile followed by : -c package_write_ipk followed by : bitbake package-index followed by opkg update/upgrade... stil no effect Mar 16 09:06:04 But there is no way of doing it faster than just building a new image every time? It's so slow... and I am struggling with the kernel stuff which makes it even worse... Mar 16 09:06:59 Back in the day I remember that when I do a tftp to zImage it worked. Mar 16 09:07:12 but setting that up is a mission Mar 16 09:07:36 Michael Halstead, ping? you here? Mar 16 09:18:20 morning all Mar 16 09:18:54 hi bl Mar 16 09:19:15 hi woglinde Mar 16 09:20:17 pompomJuice: i sometimes use tftp for kernel dev. I just start a tftp server (i use the tftp_server.py that ships with tftpy python-module) and let it share ${WORKDIR}/linux-${MACHINE}../arch/arm/boot (where zImage is located). And then I just "bitbake -c compile -f kernel" when I make changes Mar 16 09:20:25 hey bluelightning & woglinde Mar 16 09:20:55 pompomJuice: this way i don't have to create packages or anything, i just reboot and let uboot tftp the zimage Mar 16 09:21:43 Ok mag0, looks like il do the tftp thing again.... But why is there a fat partion mounted at /boot that contains kernels? I don't understand. The kernel must be somewherE? Mar 16 09:21:55 Where is it located? Mar 16 09:21:58 pompomJuice: and i think with regards to your attempt in using opkg to update the kernel, you will probably need to increase PR every time you rebuild otherwise opkg will not consider it a new version even if you've rebuilt index Mar 16 09:22:20 I do a install --force-reinstall Mar 16 09:22:22 pompomJuice: i dunno where you keep your kernel :) kind of hardware specific Mar 16 09:23:08 but everything points to /boot/zImage... for example.... my u-boot says: loadimage=fatload mmc ${mmcdev}:${mmcpart} ${loadaddr} ${image} Mar 16 09:23:12 thats a lie! Mar 16 09:23:27 Is there a standard way to fix "contains bad RPATH" errors from a recipe? Mar 16 09:23:32 if I resolve all those partitions and variables Mar 16 09:23:35 it's simply a lie Mar 16 09:23:47 I can delete all files from /boot.... and the thing still boots. Mar 16 09:24:11 Therefore the /boot partition only serves as a decoy do confuse noobs Mar 16 09:24:35 So frustrating Mar 16 09:24:40 Logic breaks down Mar 16 09:25:08 but im gonna try that python-tftp... hopefully that works easily. Mar 16 09:26:57 woglinde: do you know if meta-java has any external host dependencies (other than libstdc++ that the README mentions); my java build doesn't build clean on our buildservers, it seems to fail when trying to link the hotspot launcher for icedtea6-native since it is unable to -ljvm. The only library searchpath is -L `pwd`, shouldn't it be looking at the JVM sysroot to find libjvm.so? Mar 16 09:27:10 pompomJuice: where is your /boot mounted then? Mar 16 09:28:53 wait a minute mag0 Mar 16 09:29:32 seems that I have a "dud" boot folder that exists on my root partition... but I know I have a fat partition also with the same files... Mar 16 09:29:47 mag0 ljvm is inside the sourcecode Mar 16 09:29:54 libjvm I meant Mar 16 09:29:54 -/boot is not mounted at all Mar 16 09:29:57 pompomJuice: yeah, so that's probably where your logic breaks down Mar 16 09:30:25 Let mee see if I replace the kernel file on the fat partition Mar 16 09:30:32 mag0 so there is maybe a problem compiling libjvm Mar 16 09:30:56 woglinde: and that would be jamvm, right? (if my java-native is jamvm)? Mar 16 09:31:03 no Mar 16 09:31:10 icedtead/openjdk Mar 16 09:31:26 why should openjdk use the libjvm from jamvm? Mar 16 09:31:35 okay, weird. I don't see any errors in the log before hitting this linking issue with the hotspot launcher Mar 16 10:04:43 woglinde: to be honest i don't fully understand the whole bootstrapping process with java. I'm trying to build jamvm for target, which depends on classpath and a native javac. If I understand things correctly, OE begins by building jikes-native, which then compiles java-initial (cacao-initial-native in my case). It then builds a Java compiler that is implemented in Java (ecj-initial-native), so OE can compile 1.5 Mar 16 10:04:45 code (although the VM is 1.4). It can then finally build jamvm. So where does openjdk come into this? And icedtea? Mar 16 10:21:54 mago we compile from scratch and we need to start with java 1.4 and build us up to java 1.6 and hight Mar 16 10:22:15 because jamvm and cacao needs a java 1.6 compiler Mar 16 10:22:58 jikes is the only javac compiler which does not need another javac Mar 16 10:31:16 woglinde: okay, and openjdk is the most complete javac that meta-java provides (java version compatibility-wise)? but what does then virtual/java-native represent? is it the JVM that OE use to build OpenJDK? Mar 16 10:34:08 virtual/java-native presents java executible in our case jamvm or cacao Mar 16 10:34:52 mag0 in our case it provides the vm which can build openjdk/icedtea-native Mar 16 10:35:37 we played around with using jamvm and cacao directly but that leaded to errors when using heavy parallel threads Mar 16 10:36:01 so we decided to minimize the errors build icedtea-native first Mar 16 10:36:12 and than use it to build the target one Mar 16 10:36:40 sorry we have no site for design decisions Mar 16 11:16:50 this diff makes sure some flags (such as -mfloat-abi) are not stripped in u-boot build: http://pastie.org/10029961 - any comment? Mar 16 11:17:34 the error that lead me to this is 'foo.o uses VFP register arguments, output does not' Mar 16 11:43:11 woglinde: so the bringup chain is jikes-native+classpath-initial-native (1.4 JVM) -> jamvm-native||cacao-native -> icedtea-native -> openjdk-jre && openjdk-javac ? Mar 16 12:56:43 is TEST_IMAGE="1" with core-image-sato supposed to work when building on a remote box? Mar 16 13:18:24 blindvt`: you do have to have some means of displaying graphics, either VNC or an X server Mar 16 13:19:27 bluelightning: yes, X11 forwarding works fine for my user but i guess sudo is in the way Mar 16 13:19:39 oh, right Mar 16 13:20:01 you should create the tap devices with scripts/runqemu-gen-tapdevs Mar 16 13:20:38 that's not related to being on a remote machine though Mar 16 13:21:57 pristine scripts/runqemu-gen-tapdevs will not run since some smart guy uses "[[" in there Mar 16 13:22:06 * blindvt` sighs Mar 16 13:22:48 blindvt`: in its defence it does #!/bin/bash ... ;) Mar 16 13:23:13 though I agree it should be very easy to de-bashify it Mar 16 13:23:24 you were running it with sh I assume? Mar 16 13:23:33 yea, done. Mar 16 13:24:04 i'm using busybox's ash, or msh, depending on if i have an MMU on my terminal Mar 16 13:25:34 for ((index=0; index < $COUNT; index++)); *wince* seq, anyone? Anyway, let's see Mar 16 13:33:22 for index in `seq 0 $COUNT`;do ... done? Mar 16 13:34:50 yep Mar 16 13:35:03 hrw $() though Mar 16 13:35:28 as? Mar 16 13:35:44 cause I usually use that form so would love to improve ;d Mar 16 13:35:53 instead of the ancient, non-nestable, ugly `` Mar 16 13:36:13 did i say discouraged? :) Mar 16 13:36:24 ;) Mar 16 13:52:18 hrw, http://git.openembedded.org/openembedded-core-contrib/log/?h=aldot/fixes fwiw Mar 16 13:53:54 ok. so are the pre-populated tap picked up automagically or do i need to somehow hit runqemu with a clue-bait.. Mar 16 13:55:39 works automagically. Nice :) Mar 16 13:59:58 blindvt`: thanks Mar 16 14:00:05 each day something new Mar 16 15:38:47 meta/recipes-core/initscripts/initscripts-1.0/bootmisc.sh: SYSTEMDATE=`date -u +%4Y%2m%2d%2H%2M%2S` Mar 16 15:38:56 wth? Mar 16 15:39:07 heh Mar 16 15:39:16 what's it do with that? Mar 16 15:39:47 length qualifiers for strftime? who invented that? Mar 16 15:40:51 that is a little odd, don't the format specifiers already sort of include an implicit length? Mar 16 15:41:11 save-rtc.sh, bootmisc.sh and image.bbclass do that and of course fail miserably for me Mar 16 15:41:27 kergoth: yea, Y is 4 by definition, same for the other fields Mar 16 15:42:01 kergoth: glibc handles these as extension, it seems :-/ Mar 16 15:42:56 i'm more inclined to fix these 3 spots in OE than to support that non-standard format though. Mar 16 15:43:29 I doubt anyone would argue about fixing that, as long as the results are what they should be Mar 16 15:43:42 heh, i doubt those scripts have gotten much love recently Mar 16 15:43:43 of course Mar 16 15:43:57 * kergoth remembers swiping the initial startup scripts from the debian packages back when we created the first oe recipes :) Mar 16 15:44:52 2078af333d704fd894a2dedbc19cef5775cdadbb did that. Gary Thomas and Sau! Mar 16 15:44:56 oh well, fixing.. Mar 16 15:46:16 huh Mar 16 15:46:39 on the plus side everything but qemuarm boots just fine and runs flawlessly, on armv5whatever the Xserver somehow segfaults on me which is not something i enjoy to debug nor fix Mar 16 15:47:04 what are you working on, out of curiosity? Mar 16 15:48:11 quemu68k core-image-minimal-dev still fails to link busybox, ld from binutils master is somehow confused about GOT layout or something like that Mar 16 15:48:28 kergoth: i'm updating my uClibc testing stuff Mar 16 15:50:00 kergoth: just preparing long overdue 0.9.34 release Mar 16 15:50:19 ahh, nice Mar 16 15:51:16 POSIX.1-2008 now also allows the format specifications %+xC, %+xF, %+xG, and %+xY (where 'x' is a string of decimal digits used to specify printing and scanning of a string of 'x' decimal digits) with leading zero fill characters and a leading '+' sign character if the year being converted is more than four digits or a minimum field width is specified that allows room for more than four digits for the year. This allows date providers and consume Mar 16 15:51:37 huh Mar 16 15:52:04 that's a new thing but the %4Y is definitely not portable. I'll keep that new format thing in mind for later Mar 16 15:53:18 kergoth: see http://pubs.opengroup.org/onlinepubs/9699919799/functions/strftime.html RATIONALE Mar 16 15:54:37 well, if people fancy such odd stuff but i agree it might make sense in some cases. Mar 16 15:56:46 blindvt`, You pinged me earlier. Mar 16 15:56:57 I'm available now. Mar 16 15:59:14 halstead: Already done, thanks. I managed to try to push with the wrong user instead of git, sorry for the noise Mar 16 15:59:59 blindvt`, No problem. Sorry I didn't take care of it over the weekend. (I didn't know you were the same person.) Mar 16 16:54:30 anyone have a suggestion for a vncserver buildable under oe? Mar 16 17:16:38 isn't xvnc in there somewhere? Mar 16 17:18:18 we have the server part, but somewhat of a dearth of clients it would seem Mar 16 17:18:34 (based on a quick search of the layer index) Mar 16 17:19:09 yup meta/recipes-graphics/x11vnc/x11vnc_0.9.13.bb Mar 16 17:19:58 bluelightning: you're he wants a client... Mar 16 17:20:35 hmm, maybe I do need new glasses Mar 16 17:20:42 somehow I read client... Mar 16 17:21:01 * nerdboy can't even type assuming... Mar 16 17:39:43 looking for server - thanks i'll try that out Mar 16 17:42:16 not sure how I missed that :-P (maybe it's me that needs new glasses!) Mar 16 17:44:03 hmm. it builds. that's a good sign ;) Mar 16 17:45:08 challinan: of course, you expected otherwise? ;) Mar 16 17:49:28 * challinan has learned to expect challenges along the way - its great when it Just Works(TM) Mar 16 17:58:21 indeed Mar 16 18:04:28 * nerdboy harassing beagleboard gsoc students this year Mar 16 18:38:00 abelloni: it seems it is not easy to hook wic on a task Mar 16 18:38:17 abelloni: as wic uses bitbake -e and it fails due an existing bitbake instance Mar 16 18:39:16 yeah Mar 16 18:39:55 fact is taht if you are unning from a task, you already have access to the environment Mar 16 18:40:34 I would personnally change that behaviour if we are to make it compatible with buildroot anyway Mar 16 19:29:04 Hi all -- quick question about the package names for Boost libraries. Mar 16 19:29:20 I'm on OE Dizzy, and when I build boost, I see the following: Mar 16 19:29:51 'boost' has the package name as expected Mar 16 19:30:17 as do 'boost-serialization' and 'boost-test', which are defined separately in boost.inc. Mar 16 19:30:27 However, the other libraries are named differently. Mar 16 19:30:45 e.g. libboost-date-time1.56.0 Mar 16 19:30:56 I'm trying to figure out where that name gets generated, and I haven't had any luck. Mar 16 19:31:05 I see this in log.do_package_write_ipk: Mar 16 19:31:18 Packaged contents of boost-date-time into /ssd/perforce/ThirdPartyExports/NIOpenEmbedded/trunk/3.0/objects/zynq/linuxU/armv7-a/gcc-4.7-oe/release/build/tmp-glibc/work/cortexa9-vfpv3-nilrt-linux-gnueabi/boost/1.56.0-r0/deploy-ipks/cortexa9-vfpv3/libboost-date-time1.56.0_1.56.0-r0.3_cortexa9-vfpv3.ipk Mar 16 19:31:49 just one more github repo couldn't hurt... Mar 16 19:31:52 they're named based on the SONAME of the shared library, per debian policy, see debian.bbclass Mar 16 19:31:53 Any ideas where that might be coming from? Mar 16 19:32:34 Cool, thanks -- I'll have a look. Mar 16 19:32:35 * nerdboy just denying the pending apocalypse out loud... Mar 16 19:34:54 generally INHERIT_DISTRO pulls in debian, see meta/conf/distro/defaultsetup.conf Mar 16 19:35:28 there are definite advantages to it. you can install in parallel multiple versions of a library and thereby still be able to run old binaries that link against the old soname Mar 16 19:36:20 OK, that makes sense. Thanks! Mar 16 19:42:10 Is the package name still missing a '-' when the SONAME is not just a single number? Mar 16 19:42:28 As in Table 5.1 at https://www.netfort.gr.jp/~dancer/column/libpkg-guide/libpkg-guide.html#naminglibpkg ? **** ENDING LOGGING AT Tue Mar 17 02:59:58 2015