**** BEGIN LOGGING AT Fri Jul 18 03:00:00 2014 Jul 18 08:23:49 what happends to binaries that are not installed after a build? Jul 18 08:25:41 Layder: they (silently) stay in WORKDIR Jul 18 08:26:17 and if you use "rm_work" class, they get deleted Jul 18 08:31:42 ndec: right, seems rm_work is used in my case Jul 18 08:35:31 when the project is autotooled and i dont write the fonction by hand, what happends when i add a do_install() , does it overwrite the existing do_install or just add to the existing, Jul 18 08:36:05 morning all Jul 18 08:36:33 Layder: overwrites - if you want to append / prepend to the default then you need to do do_install_append or do_install_prepend Jul 18 08:49:35 hi bl Jul 18 08:53:06 hi woglinde Jul 18 09:07:50 RP_: for the autotools changes in master-next am I expected to clean out TMPDIR? Jul 18 09:08:27 I get a .la base failure for shadow-native when reusing TMPDIR Jul 18 09:12:00 | x86_64-linux-libtool: link: cannot find the library `/build/linaro/build/build/tmp-eglibc/sysroots/x86_64-linux/usr/lib/libattr.la' or unhandled argument `/build/linaro/build/build/tmp-eglibc/sysroots/x86_64-linux/usr/lib/libattr.la' Jul 18 09:25:22 is anyone working (or planning to work) on mesa 10.2 upgrade? what's the plan for this release cycle about mesa? Jul 18 09:25:42 I have upgrade only for mesa-demos Jul 18 09:27:10 well, i was more concerned about the rest of it ;-) but that's good to know. Jul 18 09:31:12 ndec: you'd probably want to ask rburton, he's not in here atm (but is in #yocto) Jul 18 09:33:27 ndec: I'm more excited about 10.3 Jul 18 09:33:51 10.2 is a step toward 10.3.. Jul 18 09:34:02 :) Jul 18 09:35:11 10.3's not actually out yet though right? Jul 18 09:35:17 no, it's not. Jul 18 09:43:35 ndec: I suspect mesa 10.2 and 10.3 will need an llvm update as well Jul 18 09:43:47 hm.. if ohloh/black duck is right we have had in the last 12months Jul 18 09:44:15 281 contributors for the full openembedded and 225 for oe-core only Jul 18 09:44:35 which would indicate there are only 56 contributors for the non-core stuff Jul 18 09:45:35 take out the casual commits and you are much lower... Jul 18 10:13:55 in do_install_prepend, how do you specify in which package a file has to be installed? like -dev or others? Jul 18 10:18:58 anyone knows why llvm started to fail in QA check now after insane_qa changes? Jul 18 10:19:56 Layder: you don't do that in do_install, in fact. Jul 18 10:20:00 I wonder why this error wasn't shown before Jul 18 10:20:02 ERROR: QA Issue: non -dev/-dbg/-nativesdk package contains symlink .so: libllvm3.3-llvm-3.3 path '/work/armv5te-oe-linux-gnueabi/llvm3.3/3.3-r0/packages-split/libllvm3.3-llvm-3.3/usr/lib/libLLVM-3.3.so' [dev-so] Jul 18 10:20:16 in do_install you copy all the files you want to deploy in ${D}, for all packages Jul 18 10:20:49 then you control how files are distributed in each packages with the variables FILES_ Jul 18 10:55:49 \o/ Jul 18 10:55:59 less breakage than expected in master-next Jul 18 11:00:50 well, back to my reasoning, the supposed 56 contributors of meta-openembedded is a number almost matching the ol oe-members Jul 18 11:02:10 ant_work: yes, but look at number of patches of top 3 commiters :) Jul 18 11:02:32 being that some oe-core devs are also helping with meta-oe and counting the ex-developers (inactive now) we didn't gain much devs :/ Jul 18 11:06:39 btw hrw should update and add the missing layers Jul 18 11:08:49 JaMa: apart koen, you and me I only see oe-core/Yocto devs as top-committers Jul 18 11:09:28 mickeys has long been inactive Jul 18 11:10:31 JaMa, ant_work: I think we've succeeded in scaring away the hobbyists from OE Jul 18 11:10:59 the ones I see are using poky and can't grasp OE is upstream and won't send patches Jul 18 11:11:35 yes, sad but true Jul 18 11:11:57 frankly I'm not sure that's any more true now than it was in the OE-Classic days Jul 18 11:12:10 I'm luring some buildroot devs and they are already asking about that dicothomy Jul 18 11:12:21 people often used to just fork the entire OE tree and never sent anything back Jul 18 11:12:44 bluelightning: That's true for everyting, so long as some proportion send things back it's fine. Jul 18 11:12:56 layers are still a mixed blessing Jul 18 11:13:06 nothing's perfect... Jul 18 11:13:20 I can trust OE-core won't be broken and focus on my own layers Jul 18 11:13:31 and keep herding cats for other layers :/ Jul 18 11:13:50 but glass houses and all Jul 18 11:13:56 koen: you shouldn't need to clean TMPDIR, that looks like a dependency bug where the depends on libattr may be floating Jul 18 11:32:35 koen: bluelightning: I feel we are getting to the point we have too many layers. Even with the best intentions it is impossible to do heavy changes in oe-core and evaluate the possible fall-out Jul 18 11:32:51 maybe once cleaned we could refill all under meta-oe Jul 18 11:33:05 so it will be easier to grep... Jul 18 11:33:35 ndec: im asking this because i added install -d {bindir} and install file ${D}{bindir} and couldnt find it in the ipk Jul 18 11:33:55 ant_work: but that makes the opposite problem, one we already have, worse - pretty much just one person left to do the maintenance of a huge block of recipes Jul 18 11:34:33 ant_work: the hope with splitting out the recipes by theme is that you can attract new people to areas of responsibility they are interested in Jul 18 11:34:40 I think we have done that in some cases Jul 18 11:36:55 I think we could have 2 or max 3 levels: oe-core oe-extra oe-compatibility(obsolete stuff) Jul 18 11:37:35 ofc maintainerwship would be pro-sublayer Jul 18 11:38:20 it's really hard to explain to newcomers which layers they need and why it is so split Jul 18 11:38:38 (I know Poky does some merge-down but is specific) Jul 18 11:38:57 Layder: i guess you meant install -d ${D}${bindir}, right? Jul 18 11:39:16 exacty Jul 18 11:39:51 anyway, the black hole seems now meta-openembedded/meta-oe Jul 18 11:40:04 the others are mostly maintained Jul 18 11:40:15 Layder: well, there is clearly a problem somewhere ;-) Jul 18 11:40:36 if the file was installed in ${D} but not packaged in any .ipk, bitbake would complain about it. Jul 18 11:40:47 something like 'files installed but not shipped' Jul 18 11:41:13 how many packages does you recipe generate? Jul 18 11:41:23 3 Jul 18 11:41:37 including dev und dbg Jul 18 11:43:05 are you using any of the standard class to build like autotools, cmake? Jul 18 11:43:17 RP: do you think it would be possible for the Yocto Project to find some sponsoring for the meta-openembedded layers? Couple of dedicated devs? Jul 18 11:43:56 autotools Jul 18 11:45:24 autotools_do_install will call make install. maybe the makefiles cleans bindir first... i think you mentioned you use do_install_prepend. can you try do_install_append? Jul 18 11:45:38 in fact it's probably better to use _append Jul 18 11:45:38 tried both Jul 18 11:45:59 is your file installed properly? e.g. can you check ${D} Jul 18 11:49:42 how can i do thath? Jul 18 11:50:46 ${D} is set to "${WORKDIR}/image" Jul 18 11:51:18 so you can do bitbake -c install , then look at tmp/work/...//-/image Jul 18 11:51:52 do_install actually copies files from the 'build' directory aka ${B} into the 'image' folder Jul 18 11:52:11 then the packaging task creates the packages by pulling files from ${D} into ipk Jul 18 12:01:49 ant_work: I'd love to say yes but realistically, it needs a business case and those are hard to put together Jul 18 12:26:21 ndec: in image i have usr Jul 18 12:26:32 usr/lib usr/include Jul 18 12:26:53 do they contain the files you want to put in th ipk? Jul 18 12:28:20 doesnt seem to Jul 18 12:29:28 any chance you can show your do_install_append? Jul 18 12:29:54 you mean how it is in tehrecipe? Jul 18 12:30:55 yes Jul 18 12:32:51 like i said : do_install_append () { install -d ${bindir} install -m 0755 ${WORKDIR}/test ${D}{bindir} Jul 18 12:33:17 it's ${bindir} Jul 18 12:33:23 a '$' is missing Jul 18 12:33:41 ${D}${bindir} to be exact Jul 18 12:33:43 i think i tried both but lets try it this way Jul 18 12:33:45 ${D}${bindir} Jul 18 12:33:52 D and bindir are 2 variables Jul 18 12:34:19 also, it's install -d ${D}${bindir} Jul 18 12:34:35 otherwise it will do 'install -d /usr/bin' on your PC Jul 18 12:38:41 ow right in this case it couldnt install it in the packag Jul 18 12:39:34 some 'bytes' are more important than others ;-) Jul 18 12:39:38 QA issue Jul 18 12:39:49 file were installed but not shipped Jul 18 12:40:44 /usr/bin Jul 18 12:41:32 hmm. i bet this is a left over from the previous trials. can you clean the workdir completely, and retry? Jul 18 12:41:45 yep Jul 18 12:42:00 bitbake -c clean Jul 18 12:42:09 and make sure that 'image' is gone Jul 18 12:42:44 koen: related with that do_install, there are many slashes in the paths and at the end in the metadata around. Time for a janitoring task? Jul 18 12:44:39 I think we just need to grep for '//' Jul 18 12:45:11 iirc opkg added some of its own Jul 18 12:45:15 well, excluding the SRC_URI... Jul 18 12:45:17 I think Paul fixed that Jul 18 12:45:51 I've seen some. If not too many I'll send a cumulative patch Jul 18 12:48:54 ahem.. install -m 0755 kexecboot ${D}${bindir}/ Jul 18 12:49:09 ovO Jul 18 12:50:39 ndec: thx it is worlking Jul 18 12:50:43 finally Jul 18 12:50:45 ok, good! Jul 18 12:51:30 Now im having trouble lauching my program on the target (lol) Jul 18 13:09:25 RP, looks like subsurface uses cmake now Jul 18 13:09:31 I should update recipe Jul 18 13:21:13 Crofton: I just got rid of the annoying build failure Jul 18 13:32:35 I know Jul 18 13:32:57 I've been poking at subsruface again on pc Jul 18 13:33:09 basically, no need to upstream patch :) Jul 18 13:33:20 I need subsurface for holiday Jul 18 13:33:37 diving? Jul 18 13:34:58 yeah Jul 18 14:26:39 oh my god is rbash useless Jul 18 14:29:11 typing bash Jul 18 14:29:20 gives you a real shell again Jul 18 14:30:35 such a standard security questions and no good simple solution Jul 18 14:30:50 restrict a user via ssh to only a few commands Jul 18 14:35:43 hm looks like it can be done with appamor Jul 18 14:47:49 hm or selinux Jul 18 14:47:53 but both are not simple Jul 18 14:54:00 woglinde: to do it properly you need to chroot the user like ftp does Jul 18 14:56:19 xora nah, you can easily chroot out of that Jul 18 14:56:40 selinux looks good but is not easy to configure Jul 18 14:56:51 thats what the selinux is supposed to prevent ;-) Jul 18 14:57:15 but that can not be, it is a really simple task Jul 18 14:57:42 users with group xyz should not cd and only execute program a,b,c Jul 18 14:58:06 and we have no easy tools under linux to achieve that Jul 18 15:01:01 is there a bitbake option/command to remove old builds and only keep the latest? Want to clean up my disk a bit Jul 18 15:01:36 (I have disks arriving tomorrow to expand, but until then...) Jul 18 15:03:49 there are some scripts to cleanup sstate Jul 18 15:04:09 and you could delete your tmp folder and recreate all from sstate Jul 18 15:04:31 thanks Jul 18 15:04:36 yeah, I usually do the later Jul 18 15:04:48 just wondered if there was a quick(er) way Jul 18 15:11:35 try the sstate cleanup script first Jul 18 17:21:19 JaMa: three more fixes on http://git.yoctoproject.org/cgit.cgi/poky-contrib/log/?h=rpurdie/metaoe, will post later Jul 18 17:56:35 RP: thanks, I already have exo and was planing to push back syslog-ng and wireshart on submitters, because they aren't in master yet, but will take your fix instead Jul 18 19:57:46 JaMa: if when we have a gdal patch for the B thing, i'll push it back to Frank Jul 18 19:58:00 not that he's been on IRC lately... Jul 18 19:59:42 lulz wireshart Jul 18 19:59:58 nerdboy: ok Jul 18 20:01:20 * nerdboy hadn't really thought about GIS usage in yocto either Jul 18 20:01:45 do you know if there's a user base for that? and how big? Jul 18 20:02:36 i took care of a lot of that related/orphaned stuff in gentoo Jul 18 20:04:21 gdal/ogr/tiff/gpsd/gpsdrive/mapnik/blah/blah/blah Jul 18 20:05:21 not that i actually get to do much gis stuff anymore, but apparently i still care... Jul 18 20:49:44 I guess .so files should have 755 permissions ? Seems to be no problem with 644 though.. Jul 18 20:50:40 kroon: iirc ld-*.so has to have executable bit Jul 18 20:52:44 JaMa, yeah, those I can even run from the commandline, it takes arguments and everything I discovered. But "normal" .so files, like libbz2.so Jul 18 20:53:29 On my fedora system libbz2.so.1.0.6 is 755 Jul 18 20:53:48 actually running it, I get a core-dump though Jul 18 20:55:09 meta-freescale's gpu-package sets all its .so-files to 644 Jul 18 20:57:31 gentoo also sets 755 for all libs Jul 18 20:58:47 seems to be a linux-specific thing, that .so doesn't need to be executable. Some unixes require it though Jul 18 20:59:14 Apparently ubuntu has 644 for most of them Jul 18 22:26:45 JaMa: ah, that is why those issues suddenly appeared! :) Jul 18 22:26:58 JaMa: hadn't realised they we're in yet :/ Jul 19 00:14:16 no pseudo 1.6.1 up on yoctoproject.org yet, or maybe there just is a delay ? **** ENDING LOGGING AT Sat Jul 19 03:00:00 2014