**** BEGIN LOGGING AT Thu Jun 23 02:59:57 2016 **** BEGIN LOGGING AT Thu Jun 23 04:34:29 2016 Jun 23 05:42:28 it could also be python3 behavior causing some undeterministic behavior Jun 23 05:43:49 but I see this with 2.1 release as well so may be its not python3 Jun 23 06:05:17 bluelightning: Here is sigdump of task in question http://sprunge.us/abeM Jun 23 06:05:40 bitbake-dumpsig tmp-glibc/stamps/raspberrypi3-oe-linux-gnueabi/wpe-image/1.0-r0.do_image_rpi_sdimg.sigdata.d6c80580d044fe2d636f29b97ca7dfc7 Jun 23 06:06:32 as you can see the sigdata file has d6c80580d044fe2d636f29b97ca7dfc7 but when we dump it the hash comes out to be e8020179eb295100793da3d48b5b11b4 Jun 23 06:06:54 exactly as reported in http://hastebin.com/josusilopu.avrasm Jun 23 06:10:06 as you can see in dump do_image_rpi_sdimg variable uses DATETIME var Jun 23 06:11:40 and https://gist.github.com/16ec5593c9c847d9b1704691867fc3a1 fixes my problem Jun 23 06:40:17 hi everybody. I have a question about kernel and I am sure somebody here can help me out. Jun 23 06:41:17 I have received an SDK which is actually SPI driver that needs to be added into the kernel Jun 23 06:41:36 I added the source code under drivers/misc/SPI Jun 23 06:41:49 and expected that it would build out of the box :) Jun 23 06:41:53 but I get the following error Jun 23 06:42:07 Building modules, stage 2. MODPOST 2881 modules ERROR: "spi_new_device" [drivers/misc/SPI/spirom.ko] undefined! ERROR: "spi_busnum_to_master" [drivers/misc/SPI/spirom.ko] undefined! ERROR: "__spi_register_driver" [drivers/misc/SPI/spirom.ko] undefined! ERROR: "spi_async" [drivers/misc/SPI/spirom.ko] undefined! ERROR: "spi_finalize_current_message" [drivers/misc/SPI/spi_amd.ko] undefined! ERROR: "spi_register_master" [drivers/misc/SPI/spi_ Jun 23 06:42:25 here is more clear version: http://pastebin.com/zdRmaSn2 Jun 23 06:42:52 so, I guess it can't find the kernel modules but I have no idea what I should do next Jun 23 07:27:02 Is it possible to install multiple gcc versions in an sdk? Jun 23 08:18:06 morning Jun 23 09:05:09 good morning Jun 23 17:14:09 all: any linaro toolchain experts? having problems generating sdk from within yocto using meta-linaro-toolchain Jun 23 18:45:45 kergoth: how is e.g. local.conf.append.mel-dra7xx-evm included ? Jun 23 18:46:15 ah, that's handled by mentor's setup scripts. meta-mentor includes a setup-environment script which does some other bits above and beyond what oe-init-build-env does Jun 23 18:46:40 it's not ideal, but that mechanism lets us work around bsp issues from the distro until the upstream bsp resolves it, for example Jun 23 18:49:29 yep I see, cool Jun 23 18:52:52 dl9pf_: https://gist.github.com/kergoth/11f89fd365d78429e528 summarizes the differences Jun 23 18:52:57 if your'e curious Jun 23 18:55:42 * dl9pf_ looks - tnx for the link Jun 23 18:58:23 we're hoping to get more of that merged at some point here Jun 23 18:58:32 bluelightning: whatever happened with the tinfoiil changes to use the bitbake server? Jun 23 18:58:43 i'm assuming it never got merged and needed polish? Jun 23 19:27:22 kergoth: still working on them Jun 23 19:27:48 k, np, just figured i'd check, i remembered seeing it, and wasn't sure what the status was Jun 23 19:27:48 thanks Jun 23 19:28:18 kergoth: some of the recent bitbake changes triggered some pretty significant refactoring, one part (task execution within devtool) broke unfortunately and I haven't quite figured out how to fix it yet Jun 23 19:28:43 ahh, yes, that's not too surprising. lots of old topic branches probably got screwed by that :) Jun 23 19:28:45 * Crofton|work tries to get gnuradio building with qt5 Jun 23 19:47:42 kergoth (and others) you might have an opinion on this.. Jun 23 19:48:09 in the bitbake fetcher -- we recently had some customers that were trying to put the git mirrors (the download git directories) on a windows filesystem Jun 23 19:48:34 they found a patch where the states URL ends in '/'. The fetcher mirrorname then ends in a '/', which is translated to a '.' Jun 23 19:48:46 this is invalid on the filesystem and caused the mirror to not show up.. Jun 23 19:49:38 I've got a few simply patches to strip the '.' [after the translation] from a url MIRRORNAME (and corresponding in the git.py). Does this seem like a reasonable change? Jun 23 19:49:58 I'm also worried other fetchers, hg and perforce may do something similar -- but I don't have any specific tests for themt o check Jun 23 19:51:10 (I plan to send the changes to the bitbake-devel list, but I'm looking for feedack on the strategy first) Jun 23 19:53:34 (or is the recipe at fault here -- and 'don't do that' is the answer Jun 23 19:55:38 fray: '.' is invalid in the filesystem? last i checked periods work fine on a windows fs.. or is it just the trailing .? Jun 23 19:55:53 trailing '.' in a directory name is apparently invalid on windows.. Jun 23 19:56:05 i'd say we should just strip it off. trailing slashes in paths cause annoyances all over teh place. if you use basename, you'll get hte empty string, and dirname the entire rest of the path Jun 23 19:56:27 SRC_URI was git://dev.laptop.org/projects/soc-gtk/ Jun 23 19:56:35 * kergoth nods Jun 23 19:56:36 which created: dev.laptop.org.projects.soc-gtk. Jun 23 19:57:16 while NOT advised in the git URL, it does "work".. so stripping trailing '.' from the mirrorname/directory seems like the correct fix to me Jun 23 19:58:18 I'd strip it off right after we decode it into its components. we don't need it at that point, but the original url will still be used as is Jun 23 19:58:29 then we don't need an explicit '.' removal later Jun 23 19:58:31 but *shrug* Jun 23 20:02:09 so you are thinking it would be better to strip trailing '/' from the ud.path at the decode step.. Jun 23 20:02:29 only concern I have there is in the non file, non git cases.. perforce, cvs, etc.. do any of the require a traling '/'? Jun 23 20:02:56 i wouldn't change urldecode, just do it at the urldecode call site in the fetcher that needs it done Jun 23 20:03:03 * kergoth shrugs, could be missin gsomething though Jun 23 20:03:22 urldecode is in __init__.py isn't it? Jun 23 20:03:50 (I'm looking it up now BTW to figure out where it's actually called) Jun 23 20:05:00 the urldata_init (inside of the git.py) could be one place to do this.. Jun 23 20:05:39 ahh found it.. the actual decodeurl is in the __init__ of __init__.py, that then calls into the urldata_init per type Jun 23 20:06:26 so I could strip trailing '/' fro the ud.path in the git.py urldata_init and it'll come back and not cause a problem for the MIRRORNAME setup either.. Jun 23 20:06:31 that seems like a reasonable fix **** ENDING LOGGING AT Fri Jun 24 02:59:58 2016