**** BEGIN LOGGING AT Wed Sep 02 02:59:58 2015 Sep 02 07:16:35 good morning Sep 02 08:10:29 How can I make sure only gstreamer1.0 packages are build and no gstreamer0.10? Sep 02 09:34:59 morning all Sep 02 10:15:15 bluelightning: ping Sep 02 10:17:32 bluelightning: I have a query regarding do_package_qa failure in the build. I am getting an error message saying: http://pastebin.ubuntu.com/12252189/ Sep 02 10:19:04 bluelightning: any clue what could be the reason? The user sujith using which I login to the build machine and trigger the build has uid: 1000. Not able to understand what is triggering this error. A helping hand would be nice :) Sep 02 10:47:12 hey.. one of my recipes splits usr/bin into PN and usr/lib into PN-dev, but PN-dev does not come into the image.. why? how to bring into the image? Sep 02 10:51:46 ericbutt1rs: if it's actual binaries going into the dev package you should probably just fix the splitting Sep 02 10:51:50 ericbutt1rs: depends if the pn-dev packaging is right or not - if its packaging plugin modules or non-versioned libraries into dev when it should be in pn then you need to edit your files Sep 02 10:52:15 speaking of which we need to add a sanity test for that dont' we Sep 02 10:52:23 ".so in dev package that isn't a symlink" Sep 02 10:59:53 rburton: I thought there already was Sep 02 11:00:16 to be honest I'd much rather prefer something that just fixed it for you (which you can opt out of if you want) Sep 02 11:01:12 it is libxyz.so, no version, no symlink Sep 02 11:01:22 how to change my recipe to get it in pn? Sep 02 11:02:04 adjust FILES_${PN}-dev Sep 02 11:02:29 for each file, the first package listed in order in PACKAGES whose FILES value matches the file, gets the file Sep 02 11:03:15 i added: FILES{$PN}-dev += "/usr/lib/*.so" -- but same result Sep 02 11:03:28 FILES_{$PN}-dev Sep 02 11:03:51 pn-dev already has the .so, that's the problem isn't it? Sep 02 11:03:58 right Sep 02 11:04:03 should this recipe have a dev package at all? Sep 02 11:04:08 no Sep 02 11:04:17 so set FILES_${PN}-dev = "" Sep 02 11:04:26 and add ${libdir} to FILES_${PN} Sep 02 11:10:03 thanks! now everything is in pn.. Sep 02 12:05:04 Humm the only solution i found is to execute chown root:root ${D} :P Sep 02 12:16:12 sujith_h: you might want to ask kergoth about that one Sep 02 12:40:56 bluelightning: sure Sep 02 15:10:02 kergoth: are you around? Sep 02 15:11:00 kergoth: I'm seeing some odd behaviour when using MIRRORNAME and BASENAME in PREMIRRORs when fetching hg or svn sources. It appears these sources append some path elements to the PREMIRROR url. Is this expected? Sep 02 15:11:47 hey, can I build valgrind for ARM with the recipe in meta-core? Sep 02 15:11:51 is there a bitbake var that allows one to do this? do these two fetchers do this automatically? Sep 02 15:11:58 meta/recipes-devtools* Sep 02 15:15:49 ERROR: Nothing PROVIDES 'valgrind' Sep 02 15:16:41 COMPATIBLE_HOST = '(i.86|x86_64|mips|powerpc|powerpc64).*-linux' Sep 02 15:16:41 COMPATIBLE_HOST_armv7a = 'arm.*-linux' Sep 02 15:17:00 if it doesn't fall into those compatible buckets, valgrind isn't availabel for your platform Sep 02 15:19:45 ERROR: valgrind was skipped: incompatible with host arm-foo-linux-gnueabi (not in COMPATIBLE_HOST) Sep 02 15:20:05 then your system is not compatible.. IA, MIPS(32), Power, or ARMv7a based Sep 02 15:21:09 lpapp: what's your machine? Sep 02 15:22:13 jedix: possibly, can you elaborate on that? Sep 02 15:22:56 for background, what we're doing is trying to add a mirror to the system that points to our customer support site.. Sep 02 15:23:10 so the site has to be accessed via https and have credientials added to it for each customer.. Sep 02 15:23:26 armv5tejl Sep 02 15:23:32 jedix should be able to show you what the format of mirror line(s) are and what it's being transitioned to.. Sep 02 15:23:45 lpapp, according to the recipe ONLY armv7a is supported by valgrind.. Sep 02 15:23:48 is this a Yocto and/or valgrind limitation? Sep 02 15:23:50 For hg, when the url scheme is changed, the mirror tarball basename replaces the path of the url, to ensure it fetches the tarball rather than leaving the rest of the url intact Sep 02 15:23:54 same with git Sep 02 15:24:02 beyond that, i'd need more etail, the exact lines, as you suggest Sep 02 15:24:05 detail Sep 02 15:24:22 you can adjust there recipe and try it elsewhere.. but someone added the limitationf or a reason.. testing or upstream support Sep 02 15:25:11 yeah, I am wondering... Sep 02 15:25:21 because of upstream says that it should just work... Sep 02 15:25:27 and if not... Sep 02 15:26:13 I'd like to see the mirror tarball support moved out of the core, or at least that url mangling portion, and use a replacement like MIRRORNAME to make the change to the tarball in the mirror line itself rather than in the fetch core,b ut there are complications that need sorting out to move in that direction, since we'd still need a url replacement specified for the tarball filename, so the individual fetch method would need to be able to inject Sep 02 15:26:13 replacements. then we'd need to pass along the mirror urldata to the fetch method download(), so it knows what was actually fetched Sep 02 15:26:21 but that's longer term.. Sep 02 15:26:52 ok. (I think jedix stepped away from his desk for a sec.. he should be back in a minute) I don't have an example of what it's doing unfortunately Sep 02 15:27:01 k, np Sep 02 15:30:34 lpapp the issue may be related to thumb support BTW. thumb may not be supported by valgrind Sep 02 15:30:51 sorry, back Sep 02 15:31:47 let me pastebin Sep 02 15:32:27 fray: hmm. Sep 02 15:33:15 a lot of tools don't support the older thumb1/thumb2 instructions for non armv7 platforms. Sep 02 15:33:33 (thumb was rarely used in the workstation and consumer devices that people built the tools on) Sep 02 15:34:42 kergoth: http://pastebin.com/L7GgmaGw Sep 02 15:35:49 fray / lpapp: maybe it's worth building -marm and see if it works? Sep 02 15:36:19 lpapp: assuming upstream says that for arm-no-fpu? Sep 02 15:36:21 kergoth: in this example, the svn source for unfs3 is adding a /code/{pkg} for BASENAME and a svn.code.sf.net.p.unfs3.code/{pkg} for MIRRORNAME Sep 02 15:37:14 what's the original url, and what are you wanting it to fetch vs what it's actually fetching? also, what does the do_fetch log show regarding the mirror processing? Sep 02 15:37:34 * nerdboy on his first armv5 machine, so... Sep 02 15:40:39 fray: nerdboy hate to say that, but I cannot comment on any of this. It is far beyond my league ^_^ Sep 02 15:40:49 I'll add that Sep 02 15:40:59 if you have any ideas, I am happy to give it a go and see if it works. Sep 02 15:41:25 it is just that my colleague has the valid request of getting valgrind on the board for benchmarking purposes if possible. Sep 02 15:45:09 lpapp: you'd need to muck with the TUNE parameters and set it to allow your machine Sep 02 15:46:24 basically rebuild -marm -mno-thumb i think Sep 02 15:47:52 So I found the bug that was making pseudo act up, turns out that realpath(path, NULL) was broken. Also no one should ever use realpath() because it's horrible. But it's fixed. Sep 02 15:47:53 not sure if it's worth it for a valgrind "benchmark" unless you have no other way to meet the req Sep 02 15:48:04 Will send out 1.7.1 patch today I think. Sep 02 15:49:19 lpapp: if you have time for an experimental build give it a try... Sep 02 15:50:39 kergoth: http://pastebin.com/6qPPT585 Sep 02 15:51:19 kergoth: is that the info you wanted? Sep 02 15:51:48 kergoth: I think I should remove BASENAME/MIRRORNAME and just have the URL for what I want and handle a leading slash on the server side Sep 02 15:52:00 but it seems odd that this only happens on certain packages Sep 02 15:52:08 almost all svn and hg Sep 02 15:53:47 lpapp: https://github.com/VCTLabs/meta-raspberrypi/blob/master/conf/machine/raspberrypi.conf <= kind like that? Sep 02 15:58:24 *kinda even? Sep 02 15:58:28 jedix: the thing is, svn doesnt' have mirror tarball handling. it doesn't know how to fetch a tar.gz and use it instead of contacting an svn uri, but your changing the url scheme in your mirror line activates the basename mangling in uri_replace() which is done for mirror tarball handling. Sep 02 16:00:47 kergoth: what about BASENAME? Sep 02 16:01:06 it's tacking "code/package", and I would expect "package" Sep 02 16:02:24 yeah, looks like BASENAME Is currently the entire path, not the basename. really unintuitive Sep 02 16:02:59 but only for SVN and hg? Sep 02 16:03:47 no, that's always the case, unless mirror tarball handling is active, in which case the path is overridden anyway Sep 02 16:03:51 DEFAULTTUNE ?= "arm1176jzfshf" Sep 02 16:04:05 hmm, that is an interesting set of characters :) Thanks! Sep 02 16:04:05 see lines 854 to 859 in bitbake/lib/bb/fetch2/__init__.py for the replacements like BASENAME Sep 02 16:04:28 that said, you don't need to use BASENAME to get the BASENAME Sep 02 16:04:35 you can use regex replacements and group references Sep 02 16:05:17 foo://.*/.*/ bar://oururl/oursubdir/ or foo://.*/.*/(.*) bar://outurl/oursubdir/\1 Sep 02 16:05:21 for example (untested) Sep 02 16:06:09 oh, okay Sep 02 16:06:26 that might work better since I may ditch the leading slash on my side instead of the server Sep 02 16:08:10 just remember that the url is split, *then* the regexes apply. e.g that gets broken up into multiple. foo is one regex pattern match, .* is the one for the host, and .*/ is the one for the path (or .*/(.*) for the latter example) Sep 02 16:08:28 so e.g. each component can use ^ and $ and that anchors just for that component (though ^ is implicit) Sep 02 16:12:30 thanks kergoth Sep 02 16:12:39 no problem Sep 02 16:13:09 the svn fetcher seems odd to me. it generates a tarball with a specific filename, but afaict nothing ever uses its tarball anywhere. doesn't handle it like a mirror tarball, doesn't unpack it ever Sep 02 16:13:17 * kergoth scratches head Sep 02 16:18:00 it's time for fetcher3? It looks like many features which were working in fetcher1 didn't ever make it into fetcher2 Sep 02 16:19:11 * kergoth wants to prototype a standalone fetch tool we could use, not part of bitbake at all Sep 02 17:07:13 so I did a ton more testing, and so far as I can tell, pseudo 1.7.0 really did have exactly one missing allocation causing those weird random failures. Sep 02 17:07:21 whee. Sep 02 17:07:34 Also found an XFS problem that could affect 64-bit inodes and probably fixed it. Sep 02 18:08:56 So, yocto builds subversion-native right...to fetch source from a svn:// URL? How does it know it needs to do that - what's the magic behind? Sep 02 18:09:58 line 528 of meta/classes/base.bbclass. Sep 02 18:10:11 528 to 530, that is Sep 02 18:10:20 * gunnarx checking Sep 02 18:10:48 also grep for subversion-native.. Sep 02 18:11:22 aha... Sep 02 18:12:42 pregnant-kitty-with-the-bottomless-pit just caught a finch and brought it in the house to play with Sep 02 18:13:19 fray: did i mention she was pretty spry for a giant lump of kittens? Sep 02 18:13:39 nerdboy :) Sep 02 18:13:56 apparently the finches are suckers for the sunflowers Sep 02 18:14:20 cats around here like to 'farm' bunnies.. :P Sep 02 18:14:50 they've figured out how to keep the bunnies in certain areas of the property.. and then they selectively pick them off over the period of a month or so in the spring.. :P Sep 02 18:15:04 now i know why she spends so much time lurking in the garden... Sep 02 18:15:45 in the morning and evening.. if I forget to feed the (inside) cats.. they start looking at me like I'm food.. :P Sep 02 18:16:17 that's the thing, if you drop dead now, your "inside" cats will eat you... Sep 02 18:16:36 well, I think the dogs would protect me for a while.. THEN the cats would eat me.. ;) Sep 02 18:16:50 kergoth: that's a whole lot of python code doing the magic. OK thanks Sep 02 18:17:01 it's 3 lines for what you're talking about. Sep 02 18:17:08 lots of python code doing lots of other things too, yes Sep 02 18:17:26 and they'd just look at your wife like "what? i was hungry..." Sep 02 18:19:14 likely yes Sep 02 18:19:30 but they are really good at keepign the mice out of the house.. ;) (well 2 of them are anyway) Sep 02 18:20:10 ours always like to bring stuff in from outside.. Sep 02 18:20:58 we only live about 100 yards from the road.. I got done playing undertaker.. they're not allowed out any longer.. Sep 02 18:21:04 the outside cats aren't allowed in... Sep 02 18:21:06 problem solved Sep 02 18:35:25 sounds like letting the cats play in traffic was also "problem solved" ;) Sep 02 18:41:12 my would like to keep stray kitty in the house but i don't think that's recommended Sep 02 18:41:56 cat is to used to being outside, really only likes to come in when hungry or whatever Sep 02 18:42:03 *too even Sep 02 18:42:38 most of her favorite nap/sleep/lurk spots are outside Sep 02 18:57:36 hi guys Sep 02 18:58:03 may i know, if we have yocto support for nVidia jetson k1 board? Sep 02 18:59:23 this is #cat-cuddlers, no one here to answer your yocto questions... Sep 02 18:59:37 miaou Sep 02 18:59:47 gunnarx: means? Sep 02 18:59:49 sorry, weston, just kidding :) Sep 02 18:59:59 kergoth: I still had to strip the leading slash off on the server side, but I have my issue resolved. Thanks Sep 02 19:00:12 my bad english Sep 02 19:00:14 weston: before you asked, everyone was talking about cats... Sep 02 19:00:27 instead of yocto. I'm glad you are here to ask :) Sep 02 19:00:44 gunnarx: hehe.. i dont like cats, i like dogs..they are sincere Sep 02 19:00:53 agree Sep 02 19:01:09 gunnarx: I have a dog, broonie Sep 02 19:01:38 nice. anyhow, you are asking for a BSP for nVidia I think? I don't know, but hopefully someone does Sep 02 19:02:14 gunnarx: any body here aware of it? Sep 02 19:02:27 hopefully... Sep 02 19:03:22 Hi guys, sell me the answer! Sep 02 19:04:14 Googled. Found this: https://github.com/cubicool/meta-jetson-tk1 Sep 02 19:04:27 does not look very active but maybe it's a start Sep 02 19:09:17 gunnarx: how to build this using yocto? Sep 02 19:09:53 add to bblayers, set machine in local.conf Sep 02 19:10:14 judging by the number of times people come here to ask about the tk1, nvidia should make an official bsp :) Sep 02 19:10:21 I'm not expert. I suppose you include the layer like any other and start working. Most of all this should provide you with a kernel and maybe most other things you can get from other layers. Sep 02 19:10:42 graphics is always tricky but you should be getting some of the opengl stuff from this layer most likely Sep 02 19:10:57 gunnarx: sorry, I need to know all the steps for this Sep 02 19:11:02 yocto build steps Sep 02 19:12:00 someone else jump in with your standard how-to get started with yocto answer... Sep 02 19:12:10 weston: The yocto docs are really quite good Sep 02 19:12:29 gunnarx: okay.hope some one jump in to help me Sep 02 19:13:07 I'm sure they will, but normally you need to have some specific questions. Why not start by getting familiar with the GDP project on other platforms, for example :-) Sep 02 19:13:44 since I'm guessing you're doing automotive development? Sep 02 19:15:09 gunnarx: yes, you are right...I am from automotive Sep 02 19:15:17 gunnarx: this is my first board Sep 02 19:15:24 do you know GDP? Sep 02 19:15:32 and boss asked to me to learn yocto Sep 02 19:15:36 :-) Sep 02 19:15:36 gunnarx: no Sep 02 19:15:51 weston: if you need to learn yocto then we'll all recommend reading the documentation Sep 02 19:16:39 http://git.projects.genivi.org/genivi-demo-platform, it builds for other platforms so that's a start. Then you could try to exchange the BSP to build for something else. Sep 02 19:17:00 rburton: okay Sep 02 19:17:31 The requested URL /genivi-demo-platform was not found on this server. Sep 02 19:17:46 add .git to the end Sep 02 19:18:51 I bet there are others working on porting it to jetson also, try genivi-projects@lists.genivi.org for support when you have learned yocto basics, (or if you fail to build it for one of the already supported platforms) Sep 02 19:20:57 gunnarx: thanks Sep 02 19:21:48 you are welcome Sep 02 19:25:23 gunnarx: you from? Sep 02 19:25:31 weston: you need this also http://wiki.projects.genivi.org/index.php/GENIVI_Demo_Platform Sep 02 20:03:25 I'm trying to get a unit file in place at the same time i'm doing the instalation Sep 02 20:03:58 but, I'm getting the error ERROR: QA Issue: smack-userspace: Files/directories were installed but not shipped in any package: Sep 02 20:24:58 aj_c: you need to add those files to a package or delete them from the install root Sep 02 20:25:18 or not install them in the first place... **** ENDING LOGGING AT Thu Sep 03 02:59:59 2015