**** BEGIN LOGGING AT Fri Sep 30 02:59:58 2016 Sep 30 11:14:51 when a change in the sysroot causes that "bitbake -c unpack -f foo; bitbake foo" creates a different .ipk (as seen through its deps in deploy/ipk), what will cause the old foo.ipk (with the old deps) to get back into deploy/ipk ? Sep 30 11:15:16 (working around that by bumping PR, but I'd like to understand) Sep 30 11:18:25 sounds like foo has non-deterministic dependencies and is finding more pieces it can link against in the sysroot Sep 30 12:54:22 Hi All. I am running an qemux86-64 image . I want to enable it for connection to internet. How can I do that? Sep 30 12:56:05 newguy10: you might have to be a bit more specific than that? Are you asking about passing network connect to it from your host machine? Sep 30 13:03:18 CTpollard: I have installed docker in qemux86-64 using smart. So inorder to get docker images I want qemux86-64 , connect to the net. Can you tell me how it is possible for it(qemu) to connect to net through host machine? Sep 30 13:08:29 newguy10: http://wiki.qemu.org/Documentation/Networking Sep 30 13:08:43 runqemu uses tap by default Sep 30 13:08:59 so you likely just need to set a DNS in /etc/resolv.conf Sep 30 13:11:52 rburton:Thanks. I will look into that. Sep 30 13:40:15 Hello. It seems POCO library bitbake recipie was not present in FSL community BSP Fido release. I was wondering how can one conveniently find out when a certain recipie was added to Yocto? I can not see such info on the OE layers page. Sep 30 13:40:50 i.e. I want to find out into which yocto release the POCO libraries were included for the first time Sep 30 13:42:31 https://layers.openembedded.org/layerindex/recipe/45079/ Sep 30 13:42:36 for nativesdk packages, are they supposed to be able to depend on native (not sdk) packages? If so, what is the right way to get LDFLAGS set properly? Just append -L${STAGING_LIBDIR_NATIVE} ? Sep 30 13:43:31 the page I showed does not say: is POCO included since Jethro or Krogoth? how can I easily tell? Sep 30 15:14:20 It seems like just adding the `-L` to link in libs from a native package isn't quite right: bitbake complains about the generated binaries lacking providers for those libraries (file-rdeps) Sep 30 15:14:38 Seems like nativesdk packages should only depend on other nativesdk packages? Sep 30 15:17:30 If it helps, the background here is that I'm trying to get nativesdk-qt4-tools (from meta-qt4) to build Sep 30 15:24:05 Hmm... if the -native package is using STAGING_LIBDIR_NATIVE and STAGING_INCDIR_NATIVE (to search for libs & includes), what would be appropriate for a nativesdk package? Sep 30 15:25:48 yes, nativesdk can only depend on other nativesdk packages, and you don't need to add any -I/-L, we pass —sysroot= to the compiler and linker for you Sep 30 15:26:01 as long as you're obeying our CC/LD variables and whatnot, you're fine Sep 30 15:30:39 Good evening :) Sep 30 15:30:57 is it possible to inherit my new layer from existed one? Sep 30 15:31:10 that doesn't really make sense. Sep 30 15:31:21 you can create a new layer that depends on another layer. is that not good enough? Sep 30 15:32:57 yes, it is Sep 30 15:33:00 thank you Sep 30 15:33:19 where can I read about dependencies? Sep 30 15:33:33 i'm guessing you can search either the mega manual or bitbake manual for LAYERDEPENDS Sep 30 15:33:50 great, thank you kergoth Sep 30 15:33:59 but you just set LAYERDEPENDS_yourlayername = "anotherlayername asecondlayername", where the names are what goes into BBFILE_COLLECTIONS in the different layer.conf files Sep 30 15:34:04 np Sep 30 15:58:40 Someone identified a bug in pseudo and submitted a patch to poky@. The bug, and patch, are both correct and it's merged in master. Sep 30 15:59:02 So there was a hilariously stupid bug in rename(), and I never fixed it in renameat(), which no one noticed because who would use renameat()? Sep 30 15:59:15 ha Sep 30 16:00:40 For some reason, long ago, probably 2012-2013 or earlier, I concluded that tests for "existing file with the name rename wants to rename to" and "existing database entry for the file being renamed" were logically equivalent. Sep 30 16:00:46 I can't really explain how I could have reached such a conclusion. Sep 30 16:50:07 seebs: out of curiosity, which tool did actually use openat() and uncovered tis? Sep 30 16:50:15 Because I've never seen openat(2) in the wild Sep 30 17:35:31 not openat, renameat. And I don't know. Sep 30 17:35:42 I know I had renameat() dummied out to abort if called for a couple of years. Sep 30 17:36:13 But we did finally hit it sometime before 2013. Sep 30 17:39:28 Quick question: It trying to save disk space during compiles, are there flags or options to reduce build artifacts (*.o, *.so, unnecessary images)? All I need is the *.direct images. Ideas? Sep 30 17:39:35 I'm Sep 30 17:40:44 rm_work is the first step, otherwise the unpacked source trees, etc hang out during and after the build, which takes up way m ore space than a duplicated image Sep 30 17:42:18 RP: wondering if the task elapsed time should be suffixed with ' elapsed'. when you're seeing the first ton of tasks going 0s/1s, it's not entirely clear what it's talking about Sep 30 17:44:37 RP: uh, is it expected that there aren't task progress bars anymore? I'm no longer seeing a linux-yocto do_fetch progress bar, only an elapsed time. haven't seen a single task progress bar since is tarted this build.. Sep 30 17:44:50 hmm Sep 30 18:20:44 Hmm, best way to debug a taskhash mismatch after the fact? there's no sigdata for both of them, just one Sep 30 18:31:18 kergoth: We use a gitlab runner (*.yml file) that caches (stores) the following directories (build/cache, build/downloads, build/sstate-cache) to save time for each yocto build. Could these directories be pruned further to save space/time while recreating environment each time or would setting up a local repository prove faster? Just wondering what other project do. Thx Sep 30 18:32:54 mrk377: sstate-cache and downloads available via network and configure them as sstate mirror and premirror; bitbake will only fetch what it needs Sep 30 18:33:16 mrk377: add a "make " in front to make the sentence parse Sep 30 18:38:55 panic: Thx. looking at that now. Sep 30 18:52:07 how does the cache/ thingy work anyway? does it hash the text in the file, reparsing it if it changes? Sep 30 18:52:54 mrk377: don't forget to set BB_GENERATE_MIRROR_TARBALLS (useful for git repositories) Sep 30 18:54:02 looks like that might be it. adding a single character to a comment gives Sep 30 18:54:06 Parsing of 867 .bb files complete (866 cached, 1 parsed). 1318 targets, 50 skipped, 0 masked, 0 errors. Sep 30 18:56:19 mario: if you set BB_GENERATE_MIRROR_TARBALLS = "1", then will every downloaded source code package create a tar.gz that can be stored in a local repo? Just checkin' Sep 30 18:57:26 mrk377: looks like it's just for git repositories Sep 30 18:57:51 so that people can fetch them from the download cache instead of cloning them Sep 30 18:59:14 svn repositories are cached in the download directory in other, trickier ways Sep 30 19:00:04 Actually, bitbake implicitly creates tarballs for svn repos Sep 30 19:00:19 ah, yeah, had forgotten it did that too Sep 30 19:00:29 Or at least it used to do that. Don't know about the current version. Sep 30 19:01:02 it also stores a checkout of the repo in the download dir to help generate the tarballs Sep 30 19:01:06 but i'm a bit murky on the details Sep 30 19:06:22 bluelightning: do you know if it's correct that the metadata cache/ simply stores a hash of raw file contents and reparses files when it changes? Sep 30 19:06:51 seems to work that way from some quick experimentation, but thought i'd confirm. made it seem a lot less magical to me, so i could submit a small doc. patch. Sep 30 19:06:55 Ulfalizer: there are several things in cache/, one of which is the recipe cache Sep 30 19:07:25 yeah, saw some "dependency cache" in the bitbake output too Sep 30 19:07:37 I'm seeing `ERROR: local variable 'value' referenced before assignment` on my automated build machine, with only the task (openjdk-8-native_102b14.bb:do_fetch) given on the next error line. Is there a way to get more information on the source of the error? Sep 30 19:07:55 well, AFAIK the dependency cache and the recipe cache would be the same thing Sep 30 19:08:26 alright... maybe i could keep it vagueish and only mention the recipe reparsing bit Sep 30 19:08:44 like, say, a line number for something? Sep 30 19:09:48 fishey1: which version of the build system are you using? are you using poky or separate bitbake + oe-core? Sep 30 19:10:03 bluelightning: poky. latest master. Sep 30 19:10:41 hmm... someone else reported this but I was unable to track down what was causing it there either Sep 30 19:10:44 well, latest from 6 hours ago Sep 30 19:11:19 I couldn't find *any* code where we refer to a variable "value" without first assigning it Sep 30 19:11:31 of course it must be there somewhere... Sep 30 19:14:22 fishey1: one thing you could try is 'bitbake -e openjdk-8-native' and searching for 'value' in the output. all the function bodies are listed near the end. Sep 30 19:14:31 that might find it if it's from the metadata Sep 30 19:14:49 maybe \ can be used to search for it as a word Sep 30 19:15:02 or save it and do grep -w Sep 30 19:15:07 Ulfalizer: ok, I'll try that out. Sep 30 19:15:39 could start tracing from the body of do_fetch Sep 30 19:16:24 * Ulfalizer just added some docs to the reference manual to mention that bitbake -e includes the function bodies Sep 30 19:16:29 that's pretty handy to know Sep 30 19:50:39 hmz... i wonder if the cache just looks at the access time or something like that Sep 30 19:50:56 * Ulfalizer vagueifies some more Sep 30 19:57:59 looks like it checks mtime and file checksums Sep 30 22:40:56 in what cases would 'raise bb.build.FuncFailed("msg")' be preferred over bb.fatal("msg")? Sep 30 22:41:31 looks like the argument to FuncFailed() was originally meant to be the name of the function, only no one uses it like that. still called 'name'. :S Sep 30 22:41:59 Doesn't bb.fatal internally throw an exception? Is that FuncFailed or a different one? Sep 30 22:42:44 it's BBHandledException(), which seems to mean "we've already shown the user the information they need, so don't show them any more" Sep 30 22:43:36 maybe FuncFailed() is for functions in [pre/postfuncs] and similar, when you don't want the entire task to fail, if bb.fatal() would do that Sep 30 22:43:42 that seems weird though... Sep 30 22:55:48 looks like something somewhere might throw a FuncFailed when you use bb.fatal() too, because you get "Function failed:" in the output, and that's what FuncFailed adds too. Sep 30 22:56:23 the *original* intent was for the function/task to error via whatever appropriate means, bb.fatal, whatever, and funcfailed was what you'd catch if you were calling exec_func/exec_task. that is, it's what those functions raise, not what metadata functions should be raising Sep 30 22:56:27 it didn't end up being used that way Sep 30 22:56:34 but there's really never a reason to raise it yourself Sep 30 23:03:41 ok, thanks for the clarification Sep 30 23:04:40 kergoth: mind if i quote that on the OE-core mailing list? Sep 30 23:05:04 fine with me **** ENDING LOGGING AT Sat Oct 01 02:59:58 2016