**** BEGIN LOGGING AT Wed Jan 20 02:59:59 2016 Jan 20 07:33:50 Morning! Is there a way to add source and patch file checksums to sstate cache? We have cases where packages are not rebuilt if source files are changed by multiple patches in git and the all increment to the same PR version number. The patch/source files lie in same git repo as the recipe so I guess git version numbers don't work either. Jan 20 08:04:27 mcfrisk, IIRC sstate does not use PR, but maybe you want to use prserv? Jan 20 08:04:52 and IIRC sstate hases localfiles Jan 20 08:08:40 fsdun: we merge patches one by one, and compile and test them. Two patches update same recipe and bumb PR. First got build and updated sstate. Second did not get built since PR was not updated and only source file in recipe changed. We are using BB_SIGNATURE_HANDLER ?= 'OEBasicHash'. Could not see changes to this in yocto master compared to our 1.7.1. Jan 20 08:10:15 prserv would generate a new PR for each unknown input Jan 20 08:10:31 good morning Jan 20 08:11:23 fsdun: yes, we know prserv could help but can't deploy it at the moment into the distributed build machinery. Jan 20 08:12:37 sstate has been aproved a lot since 1.7.1 Jan 20 08:23:35 well, make would detect from time stamps if file has changed and rebuild it. ccache would notice checksum change and rebuild it. bitbake with sstate cache doesn't rebuild if source file changes and recipe stays the same. sigh. Jan 20 08:24:39 It does. try to build alsa-state, change a file, build alsa-state Jan 20 08:56:54 fsdun: yeah, you are right. source file changes triggers rebuild. Our case is some other error in the chain/builds. Thanks! Jan 20 08:57:50 if you remove the second patch it doesn't need to rebuild because you already built it mcfrisk Jan 20 08:57:56 is this your case? Jan 20 10:34:37 Hi Jan 20 10:35:00 I was trying to create a valleyisland-64 image using the meta-intel layer, but I get the following error: Jan 20 10:35:01 http://pastebin.com/p9hdW83P Jan 20 10:35:05 Any idea? Jan 20 10:35:47 Spulit: try deleting /home/fraunhofer/yocto/poky/build/tmp and seeing if it happens again Jan 20 10:37:48 rburton: Let me try... Jan 20 12:03:23 hi, I have tried both genericx86-64 and intel-corei7-64 targets to build the xen-minimal-image Jan 20 12:03:31 the kern modules still fail to build Jan 20 12:03:33 http://pastebin.com/8Kbs55bW Jan 20 12:03:51 and I have tried e.g. Jan 20 12:03:52 PREFERRED_VERSION_linux-yocto ?= "4.1%" Jan 20 12:03:52 PREFERRED_VERSION_linux-yocto_intel-corei7-64 ?= "4.1%" Jan 20 12:04:05 but still no joy Jan 20 12:05:04 sorry I do not know the build system well enough to know where to look at debugging the kernel build Jan 20 12:57:54 well making checkuri work for tslib was more effort than i was expecting. i blame kergoth Jan 20 13:52:04 hallo :). How can i add some files from source-tree to a dev-package in a recipe? Jan 20 14:00:46 hallo :). How can i add some files from source-tree to a dev-package in a recipe? Jan 20 14:08:51 rburton, I think he's not activily maintaining tslib Jan 20 14:42:24 hi Jan 20 14:43:28 is there possibility to override build directory in jethro (we have old rygel recipe which works fine in 1.6 but in 2.0 some xml aren't generated properly and I assume it's because some error in Makefiles) Jan 20 14:43:49 which I would like to avoid hacking Jan 20 14:44:10 yocto 1.6 build was done directly in S dir Jan 20 14:44:22 in 2.0 there is build dir and I think this makes problem Jan 20 14:44:47 marek: you can probably set build dir in local conf Jan 20 14:45:30 it's only problem for one recipe Jan 20 14:45:51 it's build dir when building rygel recipe Jan 20 14:46:41 so in old one it was /temp/work/xxx/rygel/rygel-123 as source and now I have /temp/work/xxx/rygel/rygel-123/build after running configure Jan 20 14:46:49 marek: create a bbappend for the recipe and override it? Jan 20 14:47:08 I tried override S Jan 20 14:47:17 not sure if I can override also this behavior Jan 20 14:47:29 thus I'm asking if possible and also how ;) Jan 20 14:56:26 marek: I think you want to inherit autotools-brokensep (instead of autotools) so that the build is done in S, rather than a separate build directory Jan 20 14:59:21 joshuagl: thanks that was it Jan 20 14:59:38 np Jan 20 15:05:55 <_4urele_> hi everyone Jan 20 15:24:41 rburton: hehe Jan 20 15:24:52 tslib is in 'apply bugfix patches when they show up' mode, unless someone wants to take over Jan 20 15:24:59 i don't even have hardware to test the damn thing anymore Jan 20 16:40:36 kergoth: yeah not really blaming you, found a bug in the fetcher Jan 20 16:40:44 just that tslib was the first one to hit it Jan 20 16:42:10 i figured :) Jan 20 17:05:47 anybody know how one could run a custom driver installation script properly during the build of an image? Jan 20 17:07:18 karobar: have a recipe that executes the script, or a recipe that performs the same actions? Jan 20 17:09:03 CTtpollard: I've run the script from a recipe but it doesn't seem to be able to find the /usr/src/kernel directory on the image rootfs Jan 20 17:10:34 CTtpollard: I've got my recipe depending on virtual/kernel, which I thought would mean that the kernel headers would be there Jan 20 17:11:05 I would expect so Jan 20 17:36:41 I've got an Intel Edison running yocto with systemd-timesyncd and it appears that it will set the clock backwards when correcting time, is there a way to configure timesyncd to not do that? Jan 20 17:37:45 Isn't there always a risk of that when time is corrected? Software can't make the assumption that time will never go backwards.. Jan 20 17:37:48 does anyone have any experiences with timesyncd doing that? I couldn't find anything on google, at least not much in the configuration department... seems like the other option is to switch to a different ntp daemon (ntpd, chrony, etc) Jan 20 17:38:24 chrony (and I believe ntpd) will try not to set the clock backwards, but rather slow or speed up the system clock Jan 20 17:39:42 ah, interesting Jan 20 17:40:07 Ultimately, the problem we're trying to solve is taking high frequency readings (60hz) for energy usage and we need time to be relatively stable Jan 20 17:40:38 also, we're running over a cellular link, so ntp responses aren't very timely... Jan 20 17:41:47 admittedly i know very little about time in this context, but wouldn't you be able to leverage the monotonic clock rather than the system one? or perhaps both in concert if you need to record system clock values? Jan 20 17:41:52 * kergoth shrugs, wanders off for a bit Jan 20 17:42:48 kergoth: hmm... what's the monotonic clock? I'll look into that Jan 20 17:43:31 nerdboy: good; we are happy for patches Jan 20 17:43:40 google for posix monotonic clock Jan 20 17:43:51 might not be appropriate given what you're doing, but worth checking into Jan 20 17:43:53 thanks kergoth Jan 20 17:43:56 np Jan 20 17:44:46 it's basically a monotonically increasing value since the boot of the system. since it's not wall time, it's not subject to ntp corrections, so it's ideal for measuring elapsed time and the like Jan 20 17:45:03 * kergoth gets food Jan 20 18:42:28 did someone noticed strange black magic in gdb's PACKAGECONFIG? unlike all other recipes this one shows PACKAGECONFIG as python function in bitbake -e Jan 20 19:06:41 JaMa: it's just bitbake acting up due to the use of a packageconfig named 'python', since we use the 'python' flag to indicate whether a function is python or shell. it should be only emitting python functions if both 'func' *and* 'python' are set Jan 20 19:10:49 kergoth: interesting, thanks Jan 20 19:11:09 now level 2, sometimes it shows it with "python" also in the body: Jan 20 19:11:14 python PACKAGECONFIG () { Jan 20 19:11:14 python} Jan 20 19:12:15 of course it does Jan 20 19:12:24 that's the value of the PACKAGECONFIG variable :) Jan 20 19:12:36 a function is just a key/value pair like any other variable, just flagged as a function Jan 20 19:12:41 the function body is the value of the variable Jan 20 19:12:47 which in turn works unexpectedly in 'if [ -n "${@base_contains('PACKAGECONFIG', 'python', 'python', '', d)}" ]; then' Jan 20 19:13:00 ? Jan 20 19:13:04 that will still work just fine Jan 20 19:13:21 it's a display bug in bitbake -e, that's all Jan 20 19:13:29 but "python" wasn't added to PACKAGECONFIG in that build (at least not what bitbake -e shows in history Jan 20 19:13:38 there is no history of PACKAGECONFIG variable, that is Jan 20 19:13:55 let me try to change the varflag to see if it will show the history Jan 20 19:14:02 don't know what to tell you, if python is in teh body bitbake -e shows, then something is adding python to PACKAGECONFIG Jan 20 19:14:15 the body is the value, same as a KEY="value" form, just formatted differently Jan 20 19:15:12 heh meta-qt5/recipes-devtools/gdb/gdb_%.bbappend:1 Jan 20 19:15:21 # [defaultval] "python" Jan 20 19:15:28 ok, that explains it, thanks Jan 20 19:15:31 np Jan 20 19:17:40 latest gdb also doesnt build for target on x86 Jan 20 19:17:46 on musl atleast Jan 20 19:18:28 see https://github.com/kraj/openembedded-core/commit/435ade46174a1099ff00d86387ebd57b68592c67 Jan 20 19:34:24 morning all Jan 20 21:13:08 I seek advice from the wizards! Jan 20 21:13:40 I wish to perform a do_install_append, conditional upon the state of a PACKAGECONFIG flag. Jan 20 21:13:46 Is this acceptable magic? Jan 20 21:24:25 wyrm: you'd do that by making the do_install_append unconditional and using this within the body: if ${@bb.utils.contains('PACKAGECONFIG', 'featurename', 'true', 'false', d)}; then ... Jan 20 21:59:29 Is there a way to download a file in the SRC_URI and not have the unpacker unpack it? Jan 20 22:01:44 Cardoe: as far as Im concerned the do_unpack is also called from the fetcher code, you'd have to modify it manually for it to stop doing that Jan 20 22:02:54 its somewhere on bitbake/lib/ Jan 20 22:03:33 aehs29: thanks. I'll look. Jan 20 22:08:08 Cardoe: I think if you add ;unpack=0 that will disable that Jan 20 22:08:27 (to the item in SRC_URI) Jan 20 22:09:41 ah thank you Jan 20 22:59:39 i have added command and output in pastebin link Jan 20 22:59:40 http://pastebin.com/8Kbs55bW Jan 20 23:05:39 running this command Jan 20 23:05:44 gst-launch-1.0 imxv4l2src device=/dev/video0 ! imxipuvideotransform ! imxipuvideosink framebuffer=/dev/fb1 Jan 20 23:05:59 ERROR: Pipeline doesn't want to pause. ERROR: from element /GstPipeline:pipeline0/GstImxIpuVideoSink:imxipuvideosink0: could not open /dev/fb1: No such device or address Additional debug info: ../src/common/blitter_video_sink.c(743): gst_imx_blitter_video_sink_open_framebuffer_device (): /GstPipeline:pipeline0/GstImxIpuVideoSink:imxipuvideosink0 Setting pipeline to NULL ... Jan 20 23:38:48 Hi, anyone familiar with "A stop job is running for..." when powering down? The service relates to a script run on startup using update-rc.d Jan 20 23:39:32 generic unhelpful message. I wish ot said _which_ thing was stuck. Jan 20 23:42:34 otavio_: stay tuned and hopefully we can make it work Jan 20 23:43:04 the missing audio mux is in there now... Jan 20 23:43:21 not exactly what i need, but what the hey... Jan 20 23:51:09 fledermaus, are you referring to my query? Jan 20 23:51:37 geotheory_: yes. Jan 20 23:52:45 (guess who's been staring at similar on and off for the last few hours) Jan 20 23:53:12 The thing stuck is boot_script.service, which I presume relates to the boot_script.sh script I run on startup. That sets several python processes running with e.g. "python script.py &". Why ot btw? Jan 20 23:53:41 s/ot/it/ Jan 20 23:54:04 nerdboy: :) if possible, see if it has not been sent for next. If it has we can backport the commits ;-) Jan 20 23:54:20 if you know which processes it's waiting for then the answer is probably that they aren't shutting down in response to a TERM signal Jan 20 23:54:54 script is added to daemon with "update-rc.d boot_script.sh defaults 99" so I gather should be first to shut down Jan 21 00:19:05 If boot_script.sh is run on startup by update-rc.d ("update-rc.d boot_script.sh defaults 99") and calls other scripts (e.g. "python script.py &") that are ongoing, does boot_script.service end after it finishes, or only when its child processes also end? Jan 21 00:32:38 depends on how the service is declared, I think Jan 21 00:43:28 grrr... pppd is *not* creating /etc/ppp/resolv.conf despite that I have 'usepeerdns', and I definitely see it receiving primary & secondary DNS addresses Jan 21 00:49:22 fledermaus, I've described how I've declared it **** ENDING LOGGING AT Thu Jan 21 02:59:58 2016