**** BEGIN LOGGING AT Fri Feb 08 02:59:59 2013 Feb 08 08:09:39 good morning Feb 08 09:12:52 morning all Feb 08 09:29:11 Am I the only one to feel it's a bit odd to publish the Yocto test reports in xlsx format? Feb 08 09:30:01 no, you're not the only one Feb 08 09:39:22 Zagor: please feel free to reply requesting a more reasonable format Feb 08 09:39:31 FWIW I also agree Feb 08 10:29:15 bluelightning: thanks for R* in buildhistory Feb 08 10:30:11 interesting when I move some patch referenced in SRC_URI and move it from "files" to "BPN" bitbake shows WARNING: Unable to get checksum for foo SRC_URI entry that.path Feb 08 10:30:20 but then it finds it and builds foo correctly Feb 08 10:31:22 JaMa: hmm... that is unexpected Feb 08 10:32:13 yes and wasn't doing that before I think Feb 08 10:32:27 trying to reparse without cache Feb 08 10:33:16 Warning is gone after removing tmp-eglibc/cache/default-eglibc/ Feb 08 10:34:32 maybe it always worked like this (cannot detect that some files were moved in directories) but because there were PR bumps it was invalidating parse cache for such recipe Feb 08 10:42:12 I guess I can see how it could happen Feb 08 10:42:44 it only actually works out the path to files during parsing; the checksumming of files is a separate process that always gets run Feb 08 10:43:49 hm, i wish qemu had an option to not boot until a vnc had connected Feb 08 11:25:06 sameo: got a recipe for us to review yet? :) Feb 08 11:27:28 rburton: I wish :) But the merge window being around the corner, I'm busy adding buses to the kernel... Feb 08 11:27:46 ha Feb 08 11:29:00 rburton: https://lkml.org/lkml/2013/2/7/480 Feb 08 11:29:45 merge window rush! Feb 08 11:35:27 or maybe I will ask here ;D Feb 08 11:35:35 Funny thing I found. conf/site.conf has DISTRO_FEATURES += "ptest" and eglibc-initial fails. bitbake -e tells: DISTRO_FEATURES="ptest pulseaudio sysvinit" but when I remove my += then I have proper list. Feb 08 11:35:42 ideas? Feb 08 11:36:54 if I set DISTRO_FEATURES_INITMAN = "ptest sysvinit" then DISTRO_FEATURES gets proper value Feb 08 11:37:04 wat Feb 08 11:39:04 but with D_F_I set like that build is not possible as this var is wrong place for it (its used in DEPENDS) Feb 08 11:40:04 what happens with DISTRO_FEATURES_append in site.conf? Feb 08 11:41:06 it gets proper value Feb 08 12:20:30 hey ppl! anybody managing /mirrors/sources directory ? Feb 08 12:46:30 If I am having trouble with generating a u-boot image should I seek help here, on the mailing list of somewhere like stackoverflow?? Feb 08 14:46:57 .nick gbr Feb 08 15:30:28 Please let me know if you are on the bridge for the 1.4 M3 release readiness meeting. Thanks! Feb 08 15:30:35 Song_Liu: Sean here. Feb 08 15:30:50 Song_Liu: Laurentiu Palcu joined Feb 08 15:31:51 Cristian Iorga joined Feb 08 15:35:15 Paul Eggleton joined Feb 08 15:35:28 https://wiki.yoctoproject.org/wiki/Yocto_Project_v1.4_Status#Milestone_2.263 Feb 08 15:35:58 * abesis-dev wonders why oe could not download linux-mainline for Freescale? Feb 08 15:38:01 abesis-dev: because plenty of freescale drivers are not in mainline kernel. Feb 08 15:39:23 Zagor: So, should I give up trying ? what would be to best wat to have a rootfs without kernel ? Feb 08 15:39:53 wat=way Feb 08 15:40:53 https://wiki.yoctoproject.org/wiki/Yocto_1.4_Milestone_Test_Report#Yocto_1.4_M3_RC2_Build_Test_Report Feb 08 15:42:34 abesis-dev: I'm not sure, I never tried that :-) Feb 08 15:44:26 bluelightning: We can't forget about running your script that produces the list of supported and tested distros before we release :) Feb 08 15:44:54 Zagor: thanx anyway. Feb 08 15:45:03 scottrif: right... that would assume that the list has been updated in the sanity config though Feb 08 15:45:12 I'm not sure that has yet occurred Feb 08 15:45:22 bluelightning: yes - I guess we need to ensure that Feb 08 15:45:25 * abesis-dev still wonders where is "git2_github.com.Freescale.linux-mainline.git.tar.gz" file? Feb 08 15:45:39 with VIRTUAL-RUNTIME how do I specify an override, for example if you see the patch I have just posted to the mailing list Feb 08 15:45:56 I have provided VIRTUAL-RUNTIME_network_manager = "connman" Feb 08 15:46:20 but init-ifupdown still makes it into my image regardless, through packagegroup-core-boot Feb 08 15:46:52 abesis-dev: typically in your downloads/git2/ directory Feb 08 15:48:01 abesis-dev: sorry, your downloads/ directory. the git2 subdir is for the actual repos. Feb 08 15:48:18 Zogor: Actually, I took a look at the log file and my problem is bitbake could not download this file. Feb 08 15:48:34 It is looking for the tar.gz file at "http://downloads.yoctoproject.org/mirror/sources/" Feb 08 15:49:03 but it is not there, that is because bitbake stops with and error message. Feb 08 15:49:53 which freescale meta layer are you using? Feb 08 15:50:19 ok, I had to set it in the distro features Feb 08 15:50:29 sorry, not distro features, distro conf Feb 08 15:51:46 Zagor: https://github.com/Freescale/fsl-community-bsp-platform Feb 08 15:52:32 Zagor: I tried another one at https://github.com/Freescale/meta-fsl-arm, yesterday with same problems Feb 08 15:52:52 both gives axcatly the same error. Feb 08 15:53:05 axcatly=exactly Feb 08 15:59:29 abesis-dev: not sure why you get that. unfortunately I have to run. good luck. Feb 08 16:00:00 Hi, I'm trying to do offline builds of 1.1 based system but git based packages like sat-solver are failing. I'm following https://wiki.yoctoproject.org/wiki/How_do_I#Q:_How_do_I_create_my_own_source_download_mirror_.3F but any other hints what to try? Feb 08 16:59:30 i'm clearing having a slow day. what's the difference between def foo(): and python foo() {} in a bbclass? Feb 08 16:59:38 i suspect it's scope Feb 08 17:02:01 dfe'd functions are callable from python directly, they go into the python namespace Feb 08 17:02:10 python foo() {} is just a variable value with the python flag set Feb 08 17:02:17 you have to use exec_func and the like to run those Feb 08 17:02:22 ah Feb 08 17:02:30 I rather dislike the distinction, myself, as it causes confusion -- too many ways of doing things Feb 08 17:02:33 bitbake is mental :) Feb 08 17:02:35 but.. we're stuck with it for now Feb 08 17:02:44 yes, looking at other classes doesn't make it clear in the slightest Feb 08 17:02:44 the def syntax wa shoehorned in later Feb 08 17:02:49 it didn't exist originally Feb 08 17:03:02 so if i've got a python populate_packages_prepend that wants to call another function, it should be a def function Feb 08 17:03:06 (the other function) Feb 08 17:03:37 either way will technically work, but exec_func can't get a value returned from the function being executed, so it's more limited Feb 08 17:03:49 on the plus side, you don't have to go messing with globals() and the like to run them Feb 08 17:03:55 basically, Feb 08 17:04:00 python populate_packages_prepend () { Feb 08 17:04:02 the main purpose of def was to allow things like Feb 08 17:04:02 if oe.utils.contains ('DISTRO_FEATURES', 'systemd', True, False, d): Feb 08 17:04:04 systemd_populate_packages (d) Feb 08 17:04:06 } Feb 08 17:04:08 FOO = "${@bar()}" Feb 08 17:04:11 then i'd have def systemd_populate_packages(d): Feb 08 17:04:16 yeah, exactly Feb 08 17:04:47 the meaning of python populate_packages_prepend just became clear, and i think i need a bath Feb 08 17:05:39 heh, you'r ejust concatenating a block of python to the existing populate_packages task Feb 08 17:05:56 yeah Feb 08 17:06:03 * rburton scrubs Feb 08 17:06:10 you might be able to use packagefuncs for othat sort of thing, but i'm not sure if its the exact place ou'd want it Feb 08 17:06:25 that sort of explicit hoook is far superior to prepend/append, really.. Feb 08 17:06:40 * kergoth thinks _append/_prepend suck and need to be eliminated except in special cases or use of overrides Feb 08 17:06:41 turns out the inherit ${@base_contains....} trick upsets bitbake's caching Feb 08 17:06:50 hmm, interesting Feb 08 17:09:08 make a change in the then-included class, and bitbake doesn't see it Feb 08 17:09:20 changing the class is easier than chasing that bug... Feb 08 17:09:30 especially as i was having trouble making a minimal test case Feb 08 17:10:30 ah, so it may be the config caching bits that are used to decide whether to reparse the recipes that's the problem? we've hit similar bugs with things like adding tasks in classes not being spotted Feb 08 17:10:44 oh what Feb 08 17:10:46 | //usrusr//binbin//ldld:: cannotcannot findfind --ldbusldbus--11 Feb 08 17:10:58 haha Feb 08 17:11:16 looks like non-line-buffered output from multiple processes being mixed or something Feb 08 17:11:19 weird Feb 08 17:11:28 yeah, very weird Feb 08 17:12:21 looks like rp's dbus change though Feb 08 17:16:18 halstead: ping Feb 08 17:18:38 rburton: which dbus change? Feb 08 17:18:58 RP: your dbus x11 native thing? Feb 08 17:19:07 probably a stale .la from a rebuild though Feb 08 17:21:16 RP: Thoughts on deprecating SkipPackage in favor of a variable like BB_SKIPPED for the long term with bitbake? being able to parse skipped recipes could have value, for example, to use bitbake -e. It'd just end up setting 'skipped' in the recipe cache data either way. Feb 08 17:22:11 kergoth: I suspect there are issues where we don't have a coherent datastore to expand with some of them Feb 08 17:22:41 kergoth: I know I tried something like this to simplify code once before and it didn't end well. I don't remember the details though :/ Feb 08 17:23:26 Hmm, okay, so it's less simple than expected, but I still think there'd be value. Might add it as a backburner task if I get bored to see how badly it explodes Feb 08 17:24:48 scottrif, pong. Feb 08 17:25:02 halstead: going private.... Feb 08 17:27:40 rburton: how does that cause the error though? Feb 08 17:29:03 RP: old .la files refering to dbus-x11 from before the rebuild that didn't get replaced? Feb 08 17:29:38 rburton: hurry up and get that B!=S fix ;-) Feb 08 17:31:10 kergoth: for the populate_packages_prepend case, we now have variables you can add the function names to Feb 08 17:31:34 kergoth: I've not done a big search/replace yet but the infrastructure is there and I started on some like the kernel Feb 08 17:31:47 makes the profiles easier to read too Feb 08 17:35:01 * kergoth discovers there's no way to tell the cooker to unlock its lockfile without ending the process it's running in Feb 08 17:35:07 RP: nice Feb 08 17:35:10 RP: that's good to hear Feb 08 17:37:14 aw crap, one difference between def and python is that def doesn't have magic interpretation of ${D} Feb 08 17:37:21 which is why my "just move the code" isn't working Feb 08 17:37:48 rburton: you need d as a parameter to the function Feb 08 17:38:17 the old code was happily doing "${D}" + filename Feb 08 17:38:24 and tada! the right paths Feb 08 17:38:37 switching to d.getVar("D", True) now Feb 08 17:58:47 rburton: that is horrible, python code shouldn't rely on that Feb 08 18:08:39 RP: maybe we should consider deprecating that at some point, both for python functions and inline python. there are better ways to go, and would make it consistent with def'd functions Feb 08 18:08:53 could have a script check for all cases where expansion of a python function causes a change Feb 08 18:15:26 kergoth: yes, something to think about for sure Feb 08 19:28:39 how to enforce a recipe to be run do_clean if I build one other recipe? Feb 08 20:24:39 ftonello: you have updates to that recipe? Feb 08 22:14:48 otavio, I sent an e-mail about imx28evk related to the 1.4_M3 release. **** ENDING LOGGING AT Sat Feb 09 02:59:57 2013