**** BEGIN LOGGING AT Thu Mar 09 03:00:02 2017 Mar 09 08:24:35 hello Mar 09 08:25:17 rburton, ref. vlock: i have tried the most recent poky version (as instructed in quick start guide) Mar 09 08:26:24 and this is more 'friendly': it says: src/vlock/Makefile.am:9: warning: source file '../kbd_error.c' is in a subdirectory, but option 'subdir-objects' is disabled Mar 09 08:27:04 now, my older version does not say it, but looking in the build/src/vlock, indeed, the dir is empty just like it is in the newer upstream version Mar 09 08:27:37 now i have to find out how to enable this subdir-objects option ... Mar 09 08:50:54 if i add EXTRA_OECONF += "--subdir-objects" to the recipe, do_configure() complains that option is not recognized Mar 09 08:51:07 what is the proper way to add this option from the recipe? Mar 09 09:02:41 hmm, nightly-oe-selftest failing in do_populate_lic for binutils-cross-i586 Mar 09 09:02:42 cornel, I did just patch configure.ac in a similar case Mar 09 09:03:51 Heya Mar 09 09:04:46 Does anyone know why i can't run /sbin/depmod on yocto? i get: can't change directory to 'lib/modules/4-10-0-yocto-preempt-rt'" no such file or directory Mar 09 09:06:06 ant_work, thank you. i was afraid of this answer :) Mar 09 09:06:25 I don't know the sources... Mar 09 09:06:47 ant_work, that;s ok, i think i can do it Mar 09 09:07:05 rburton told me i can use bitbake -C unpack if i want to look at the sources Mar 09 09:07:32 cornel, if it is using autottols would be smthg like Mar 09 09:07:39 -AM_INIT_AUTOMAKE([-Wall foreign]) Mar 09 09:07:39 +AM_INIT_AUTOMAKE([-Wall foreign] [subdir-objects]) Mar 09 09:07:42 I have done a simple .bbappend to add a patch to a recipe (from my layer). When I build it, I get: ERROR: Can NOT get PRAUTO from remote PR service Mar 09 09:07:45 Any idea what could cause that? Mar 09 09:08:30 ant_work, found it, thank you very much Mar 09 09:30:20 rburton: congratz for nomination, I have something for you Mar 09 09:30:30 rburton: meta-gplv2/recipes-devtools/bison/bison_2.3.bb: Unable to get checksum for bison SRC_URI entry bison-2.3_m4.patch: file could not be found Mar 09 09:31:46 Hello, is this right way to disable gstreamer and make use of qtmultimedia from qtwebkit_git.bb. EXTRA_OECONF += "QMAKE_CFLAGS+= WEBKIT_CONFIG-=use_gstreamer WEBKIT_CONFIG-=use_gstreamer010 WEBKIT_CONFIG+=use_qt_multimedia" Mar 09 09:31:47 ? Mar 09 09:32:13 JaMa: bah. about to walk the dog, can you mail that so i don't forget? Mar 09 09:35:51 rburton: ok Mar 09 09:46:33 guys, I've tried to build a core-image-base with latest Yocto branch: morty. The goal is to check locale generation Mar 09 09:47:01 reading in the docs I've found: By default, all locales are generated. (http://www.yoctoproject.org/docs/1.8/mega-manual/mega-manual.html#var-GLIBC_GENERATE_LOCALES) Mar 09 09:47:34 I've tried to look inside the generated filesystem, path:"/usr/share/locale", I see only en_GB Mar 09 09:48:06 if executing the "locale" command (package: glibc-utils), default locale is POSIX Mar 09 09:48:39 looking to the features of the next Yocto 2.3 I've found this bug: https://bugzilla.yoctoproject.org/show_bug.cgi?id=9070 Mar 09 09:48:40 Bug 9070: enhancement, Medium, 2.3 M3, ross.burton, ACCEPTED , Add test that locales are working Mar 09 09:48:54 is it possibile that there is something wrong with locales? Mar 09 09:49:34 hi everyone Mar 09 09:50:27 would it be possible to ask bitbake to build a package completely 'static' Mar 09 09:51:54 (in order to execute the application on a system which is not updated) Mar 09 09:53:20 mythi: in refkit, we have a /usr/lib/os-release which contains strings like "distro-version-to-be-added-during-image-creation". The version under /etc is okay. Mar 09 09:54:02 Ah, there's a FIXME in refkit-image.bbclass.... Mar 09 09:58:24 pohly1: I implemented a PoC of the PATH trick we talked about: http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/wip-rss2&id=24cc50d836e0557188344a06bf2fb1f119f16bf0 Mar 09 10:00:51 RP: that looks very easy indeed. However, where it get's complicated is when ASSUME_PROVIDED is used for additional things, like libsdl. Mar 09 10:01:43 Will that be found when we call pkg-config from the native sysroot? Mar 09 10:02:31 My point is, the PATH trick is likely to be very useful, but we need a way to make it optional. Mar 09 10:02:46 Anyway, I'll play a bit with it. Mar 09 10:04:22 pohly1: I think we build sdl these days for qemu don't we? Mar 09 10:05:00 Hmm, I thought the default was to build with vncserver instead of SDL by default. Mar 09 10:05:27 pohly1: but you can change that with PACKACGECONFIG options and build sdl-native? Mar 09 10:05:37 pohly1: FWIW I've build core-image-sato-sdk with this Mar 09 10:05:48 pohly1: only have one fix in that branch to tzcode-native Mar 09 10:05:56 Poky indeed enables sdl, without ASSUME_PROVIDED. Mar 09 10:06:16 But other distros (refkit included) just use the OE-core default, which is sdl disabled. Mar 09 10:09:00 * RP wonders why the recipes-lsb4 libpng uses unlink instead of rm Mar 09 10:10:49 RP: don't forget the darn ncurses-native Mar 09 10:12:59 JaMa, about GPLv2...should we put all the kernels there? Mar 09 10:13:03 :) Mar 09 10:13:25 ant_work: when Linus moves the mainline kernel to GPLv3, sure Mar 09 10:14:43 he.., iirc he, Arnd and others are against Mar 09 10:14:54 Linus is *very* anti gplv3 Mar 09 10:15:03 he believes the markets should sort the DRM issue Mar 09 10:15:08 not licensing Mar 09 10:16:46 kanavin: I did say this knowing it won't happen for a ton of different reasons, not least dead copyright holders... Mar 09 10:19:01 JaMa: bison fix merged to meta-gplv2 thanks. I'll pull the other oe-core bit in due course Mar 09 10:20:51 RP: thanks Mar 09 10:21:29 RP: I'm testing your whitelist change for native tools in much bigger build and getting a lot of strange errors, will report back a bit later Mar 09 10:22:41 ant_work, i've applied the patch, and in the end vlock build directory was still empty :/ Mar 09 10:24:50 there must be some other reason then, check the logs Mar 09 10:26:07 ant_work, in configure log, everywhere i see that vlcok is enabled, and in the summary section vlock ends disabled Mar 09 10:26:23 RP: do you remember this? https://bugzilla.yoctoproject.org/show_bug.cgi?id=8083 Mar 09 10:26:25 Bug 8083: enhancement, Medium, 2.2 M3, bill.randle, RESOLVED FIXED, Add n32 test case to autobuilder nightly-multilib Mar 09 10:26:30 RP: I need a moment to talk about it :) Mar 09 10:27:48 ant_work, https://paste.fedoraproject.org/paste/BxxzPyf4DWQngJdyHg2~k15M1UNdIGYhyRLivL9gydE= Mar 09 10:28:28 dnf refuses to do populate_sdk, and I believe it is right: Mar 09 10:28:28 Error: Transaction check error: Mar 09 10:28:28 file /usr/include/bits/long-double.h conflicts between attempted installs of lib64-libc6-dev-2.25-r0.mips64 and lib32-libc6-dev-2.25-r0.mips32r2 Mar 09 10:29:30 smart happily installed both packages, but that is not right - only one of the headers was picked randomly, and the headers are *different* in those packages Mar 09 10:29:47 so what I'm getting at is that the autobuilder test is broken :) Mar 09 10:31:13 Does anyone know why i can't run /sbin/depmod on yocto? i get: can't change directory to 'lib/modules/4-10-0-yocto-preempt-rt'" no such file or directory Mar 09 10:33:07 cornel, pls check src/vlock/Makefile Mar 09 10:34:06 ant_work, in the unpacked dir? or in the build dir? Mar 09 10:36:39 you should check $B. W/out seeing the recipe I cannot imagine what's happening :) Mar 09 10:37:15 is it abandonware? Mar 09 10:37:33 kanavin: it does sound like its broken, yes Mar 09 10:37:50 kanavin: there are special header install functions to help handle cases like this Mar 09 10:38:07 kanavin: sounds like libc6-dev needs fixing Mar 09 10:38:36 JaMa: which version are you testing? the one I posted yesterday or the one I linked to today? Mar 09 10:38:43 RP: can you point me to an example of such header fixing? Mar 09 10:38:43 ant_work, good point, the recipe is still tainted (by me) Mar 09 10:39:05 JaMa: I'd try the one I linked to today as it works much better (core-image-sato-sdk and core-image-lsb build) Mar 09 10:39:42 kanavin: grep for oe_multilib_header Mar 09 10:39:56 kanavin: and see multilib_header.bbclass Mar 09 10:40:09 RP: thanks - I can honestly say I learned more about yocto in these three months than in the previous 1.5 years of doing package updates :D Mar 09 10:40:26 kanavin: I can imagine! Mar 09 10:40:41 kanavin: The variety does open your eyes a bit :) Mar 09 10:43:02 ant_work, removing my extraline from recipe did not help Mar 09 10:43:07 ant_work, this is the recipe Mar 09 10:43:12 https://paste.fedoraproject.org/paste/u8S31Z2o5iMQ8ddJKL-~7F5M1UNdIGYhyRLivL9gydE= Mar 09 10:47:26 cornel, so you have 'pam' in your DISTRO_FEATURES? Mar 09 10:47:55 hi all, i am wondering why i find headers in usr/include on my target rootfs, in manifest file i see some -dev packages like libreadline-dev. so anyone, how to find out why dev packages are installed to target image? they are not selected by IMAGE_INSTALL Mar 09 10:49:05 ant_work, one minute please, i have a test kbd build in progress Mar 09 10:50:16 afais there is a packageconfig option [pam] which does "--enable-vlock" Mar 09 10:52:04 ant_work, https://paste.fedoraproject.org/paste/5AyIIeedqp13RAJHbUjxxV5M1UNdIGYhyRLivL9gydE= Mar 09 10:55:16 RP: the one with yesterday with extra tools you mentioned in a reply Mar 09 10:55:25 RP: ok will refresh to newer version Mar 09 10:56:36 JaMa: the key one I was missing is env, once you have that the errors become more sane Mar 09 10:56:43 RP: I don't see newer version of the patch on ML Mar 09 10:56:51 Adding "false true uname test hostname nm objdump objcopy cmp printf Mar 09 10:56:52 env" gets this going much further. Mar 09 10:57:01 is this ^ what you meant as new version? Mar 09 10:57:10 JaMa: http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/wip-rss2&id=24cc50d836e0557188344a06bf2fb1f119f16bf0 is the newer one Mar 09 10:57:20 JaMa: not posted a new one on the ML yet Mar 09 10:57:28 cornel: once the configure warning is solved by adding [subdir-objects] option and you have vlock in the configure options it should just build... Mar 09 10:57:29 ok, thanks Mar 09 10:57:40 RP: some of my issues are probably due to icecc.bbclass Mar 09 10:57:50 cornel, I am not at my desk, can't check now, sorry Mar 09 10:57:58 JaMa: that probably doesn't help. I'm trying to get testimage to work now... Mar 09 10:59:01 ant_work, thank you very much. i honestly don't understand why this thing is not working. maybe rburton will want to take a look at this ... Mar 09 10:59:06 RP: btw what should happen when one of the tools doesn't exist on host? shouldn't we use this to ensure that all "usually" needed tools are available? Mar 09 10:59:17 RP: just like the sanity check does for just a few of them? Mar 09 10:59:27 brb, need to buy food Mar 09 11:00:03 JaMa: right now it will give a traceback if one is missing Mar 09 11:00:10 JaMa: We need to improve that ;-) Mar 09 11:01:35 * RP has testimage working now... Mar 09 11:02:43 RP: a similar issue with bashbug. It is a python script in bash recipe meant for reporting bugs in bash :) and it has build parameters (like machine, compiler, etc.) inside itself - so in multilib setting this is causing a clash Mar 09 11:02:56 RP: should we package it separately? not sure how to deal with it Mar 09 11:03:21 kanavin: where is that installed to? Mar 09 11:03:28 RP: /usr/bin Mar 09 11:03:43 kanavin: that would be handled by normal multilib conflict resolution then? Mar 09 11:04:00 RP: no because it's a python script. multilib resolution only handles ELF binaries Mar 09 11:04:39 kanavin: I'm sure rpm5 never cared. But I'd package it separately, we don't really need that on an embedded system Mar 09 11:05:21 RP: it never cared because it never attempted to install both bash and lib32-bash Mar 09 11:05:38 (and that is because they were both named bash... long story :) Mar 09 11:05:46 kanavin: ah, ok Mar 09 11:06:49 RP: I'll split it to a separate package Mar 09 11:35:10 hi, where can I find the kernel sources of the kernel that is used in my image? Mar 09 12:00:04 rburton, RP: I need to update the buildset-config on the new cluster, please don't queue any more builds Mar 09 12:03:38 rburton, RP: done, queue at leisure Mar 09 12:39:57 patchwork is down? Mar 09 12:40:30 getting 500 when trying to see a patch or a series Mar 09 12:41:13 meta-xilinx is using the linux-xlnx 4.6 kernel, how would I go about chaing that to linux-xlnx 4.8? Mar 09 12:41:20 changing Mar 09 12:42:14 yourfate: linux-xlnx master is using 4.9, if thats ok you can use linux-xlnx-dev to build master Mar 09 12:42:24 ah Mar 09 12:42:35 so I have to set that in the local.conf somehow? Mar 09 12:42:50 yourfate: PREFERRED_PROVIDER_virtual/kernel = "linux-xlnx-dev" Mar 09 12:44:26 nrossi: hi, do you know what's the difference between linux-xlnx and linux-digilent ? Mar 09 12:44:46 nrossi: thanks, trying that Mar 09 12:45:16 yohboy: linux-xlnx is the development kernel from Xilinx, linux-digilent is Digilents fork (maybe of the linux-xlnx kernel at some version?) Mar 09 12:45:40 nrossi: ok, so what's the best one to use ? Mar 09 12:47:26 yohboy: imo neither ;), best to use the mainline kernel or linux-yocto (which is the mainline kernel if using meta-xilinx zynq). Mar 09 12:48:56 but if I need stuff like axi dma? Mar 09 12:49:00 ist that in that kernel as well? Mar 09 12:49:10 yourfate: axi dma is in the mainline kernel Mar 09 12:49:17 oh Mar 09 12:49:29 b/c it wasn't in the 4.6 linux-xlnx as far as I can tell Mar 09 12:50:09 nrossi: so I'll try with linux-yocto :) also, digilent and xilinx both using meta-xilinx layers (some difference), where is the true meta-xilinx layers ? which one I should use ? Mar 09 12:50:30 yourfate: i think more support for axi dma variants was added into the mainline kernel in 4.6? but i am not fully aware tbh Mar 09 12:50:50 yohboy: the one i maintain is always available on git.yoctoproject.org Mar 09 12:51:00 yohboy: anything else is a mirror or a fork. Mar 09 12:52:16 b/c I'm trying to access axi from yocto, and I'm a bit lost :d Mar 09 12:53:33 nrossi: ok I understand ! I'll use the one you maintain so ! thank you for helping me to understand yocto :) Mar 09 12:54:10 yohboy: if you are unsure about layer locations/sources etc. the http://layers.openembedded.org/ is a great resouce :) Mar 09 12:54:19 yourfate: what exactly are you trying to do with the axi dma? Mar 09 12:54:32 I need a fairly high performance interface between linux and the fpga Mar 09 12:54:46 to stream some data over network into the fpga and vice versa Mar 09 12:55:10 yourfate: writing your network->fpga as a kernel driver/module or userspace? Mar 09 12:55:45 network -> userspace, where some processing happens etc Mar 09 12:55:52 then userspace -> fpga somehow Mar 09 12:56:57 the board is a digillent zybo btw ^^ Mar 09 12:57:02 nrossi: nice thank you ! to be sure I understand, for my zybo, the layers I need are: meta-xilinx, meta-openembedded, poky ? it's a good start ? I want to do things correctly :) Mar 09 12:58:33 yohboy: you only need meta-xilinx and poky (or oe-core), but sounds like you have what you need anyways :) Mar 09 13:02:44 nrossi: yes but I think it is not because things work that they are done well, as long as I do not understand all the links, I can not make sure :( So what is the meta-openembedded for ? you know ? Mar 09 13:03:15 afaik that's the layer that provides userspace package recipies Mar 09 13:03:27 nrossi: I think I have so much useless recipes/layers, because I fork a digilent project Mar 09 13:03:29 at least that's where I got stuff like stress, zsh, iperf Mar 09 13:03:31 etc Mar 09 13:04:15 so I need this one also ? Mar 09 13:04:20 yourfate: im not sure exactly how you would do it (last time i looked at dma from userspace was writing a axi dma driver using uio ;)) but the dmabuf kernel api is where you want to look for using the dma engine Mar 09 13:04:57 yohboy: if your not using any recipes from meta-oe, then you don't need it no Mar 09 13:05:21 nrossi: it make sense Mar 09 13:06:03 yourfate: http://lxr.free-electrons.com/source/Documentation/dma-buf-sharing.txt << have a good read of that :) Mar 09 13:06:49 will do Mar 09 13:06:52 thanks! Mar 09 13:09:50 what's uio? Mar 09 13:10:35 yourfate: thats a can of worms, it allows for userspace io access, essentially exposing memory mapped registers of bus devices to a userspace application Mar 09 13:11:09 hmm, so your suggest way of interacting with fpga fabric from linux would be dma buffer sharing? Mar 09 13:11:51 How much do we care about glibc info files? Mar 09 13:12:04 I need about 60Mbit/s bidirectional + some controll stuf like setting a few multiplexers Mar 09 13:12:36 yourfate: well it seemed like you were interested in that method. There are plenty of different ways to handle memory passing to devices. It really does depend on how you want to design it. Mar 09 13:13:40 I don't really know which method would suit me best / would be easiest to implement. Mar 09 13:14:28 the fpga part ist monitoring lvds data streams at the gpio, and will bundle them into data frames that should go to a pc via network. Mar 09 13:14:43 and the other direction as well :D Mar 09 13:15:57 yourfate: if its just a stream of data, then maybe a dma is what you are after, and you could just expose the stream to userspace via a chardev. Mar 09 13:16:42 ah, so I'd have an interface in /dev Mar 09 13:16:56 yourfate: and expose your control stuff as a register set, and just access it via uio Mar 09 13:16:59 yourfate: yep Mar 09 13:17:24 right, and how would I go about that? Mar 09 13:17:33 yourfate: should be easy to do as an external kernel module, there are plenty of chardev kernel module examples around :) Mar 09 13:19:16 right, I'll look into it. thanks Mar 09 13:20:58 also the register set uio thing, you said earlier that's a can of worms? :) Mar 09 13:22:32 yourfate: oh no i just mean the topic is quite complex, device tree config, etc. Mar 09 13:23:05 right, those device tree souces should be generated by the xilinx sdk, so I've heard Mar 09 13:23:29 yourfate: yer.... about that.... doing anything custom is beyond the xilinx sdk device tree generator :P Mar 09 13:23:43 oh Mar 09 13:23:55 I think this project might be a lot more complex than anticipated Mar 09 13:24:17 initially I thought I'd include some xlinix libs in a c++ project and then I could talk to the axi :D Mar 09 13:24:19 yourfate: fpga stuff is never clean cut... Mar 09 13:24:56 I'm a fpga dev, that side is usually makes sense to me, but I've not really meddled in the linux kernel before Mar 09 13:26:39 JaMa: I have a much nicer patch now: http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/wip-rss2&id=89172b0c49a0d4826ba02f387f69bbddb3c86931 Mar 09 13:27:10 This one is user extensible which helps :) Mar 09 13:27:21 then, what's the difference between 2 conf : zybo-zynq7 and zybo-linux-bd-zynq7 ? this is for the same board (zybo) ? Mar 09 13:27:24 yourfate: its worth learning the complexities, its very useful to have proper integration with the kernel for managing things. Mar 09 13:29:52 yohboy: as you are aware, the zynq is configurable. One board has many permutations of possible jobs, the 'zybo-zynq7' is an unconfigured fpga, where as the zybo-linux-bd-zynq7 is the zybo configured with the 'linux-bd' reference design of digilents. Mar 09 13:31:08 nrossi: nice, ok, thank you ! Mar 09 13:32:24 RP: my icecc issue might be caused by readlink which you already added in newer version, testing it now Mar 09 13:33:54 RP: can you shortly explain why bb.utils.setup_native_bindir function needs to be in bitbake and cannot be e.g. annonymous python in oe-core? Mar 09 13:34:34 but I like this new version much more already :) Mar 09 13:34:41 JaMa: in the original patch it had to be. It could probably be in oe-core now Mar 09 13:35:16 JaMa: when things were in layer.conf, we hadn't parsed anything in meta yet basically Mar 09 13:35:53 OK Mar 09 13:36:25 so now only to add bbwarn when srctool = which(path, tool) is empty and it's almost perfect :) Mar 09 13:36:46 JaMa: yes, although I just spotted a bug which makes be wonder how this works at all! :) Mar 09 13:37:24 FWIW adding readlink to the original version allowed me to progress a bit, now it's building qtwayland :) Mar 09 13:38:16 JaMa: cool, I think the set there now should be quite good. I'll move more of the code around and then queue up a test on the autobuilder Mar 09 13:38:37 JaMa: I think I'm preferring this to Saur's blacklist approach though Mar 09 13:39:19 me too, thanks for working on it Mar 09 13:40:29 Hopefully this might also replace sanity.bbclass:SANITY_REQUIRED_UTILITIES once proper error message is shown Mar 09 13:41:03 nrossi: I just saw the xilinx kernel already contains an axi dma driver: http://www.wiki.xilinx.com/DMA+Drivers+-+Soft+IPs Mar 09 13:41:43 I was promoting to use simple minimalistic chroots/containers/nspawn/whatever for OE builds, but this is 2nd best thing on a way towards determinism Mar 09 13:41:47 yourfate: yep, thats what i was talking about earlier, http://lxr.free-electrons.com/source/drivers/dma/xilinx/xilinx_dma.c Mar 09 13:42:41 so the kernel module I'd write would be between that driver and the /dev device then? Mar 09 13:43:24 JaMa: right, it doesn't replace that but it should at least make things more consistent Mar 09 13:43:24 RP: another candidate ninja: line 26: getconf: command not found Mar 09 13:46:44 nrossi: I want to boot on SD card, I understand the process you describe in http://git.yoctoproject.org/cgit/cgit.cgi/meta-xilinx/tree/README.booting.md, but what about bitstream ? Should I put it in the first partition, and it will be load alone before booting ? Mar 09 13:47:31 JaMa: A better version again (getconf added, moved from bitbake, error if required utils not present) Mar 09 13:47:33 http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/wip-rss2&id=15a4d93dab2765e51eb971711e956019ffbf73de Mar 09 13:47:35 yohboy: depends, need to load it before the kernel boots? if so then yes first partition and use u-boot to load it Mar 09 13:48:35 yourfate: yep, but you use that driver via the kernel dma api. Mar 09 13:48:52 nrossi: it's better if it's load before the kernel boots, isn't it ? else we can load it after boot ? Mar 09 13:49:59 yohboy: the problem is there is no right answer, some designs need to have it loaded in order to boot the kernel where as others don't need the bitstream loaded until after the kernel has loaded. So its up to you :) Mar 09 13:52:36 RP: nice Mar 09 13:52:53 anyone noticed qemu-native recently depending on MACHINE variable? Mar 09 13:52:54 nrossi: so ok, I have to make a choice so. Another question that I ask myself is: assume I already have one bootable SD card, with one bitstream, and the system works well, if I only want to change the hdf, I have to rebuild the image ? because the bitstream will change, and the DTB too. That's not clear in my mind Mar 09 13:53:21 I'm waiting for some current builds to finish before I can check why, but maybe someone already seen and fixed it.. Mar 09 13:53:48 JaMa: hopefully not my patches that did that :| Mar 09 13:53:53 ERROR: qemu-native different signature for task do_configure.sigdata between qemux86 and qemux86copy Mar 09 13:53:56 basehash changed from b90be5a6f6e98fd67b0e57390eb68c23 to 84041688e10287e17c29cd7dea712c51 Mar 09 13:54:00 Variable MACHINE value changed from 'qemux86' to 'qemux86copy' Mar 09 13:54:20 causes bunch of gtk related recipes to be also MACHINE specific because they depend on qemu-native for postinst Mar 09 13:55:49 JaMa: I've not seen that on our autobuilder tests and probably would have... Mar 09 13:56:18 btw is git://git.pokylinux.org/poky-contrib the same as git://git.yoctoproject.org/poky-contrib-archive ? Mar 09 13:56:48 I wasn't seeing RP's commit (I don't use poky repos) and noticed that my git remotes still point to pokylinux.org Mar 09 13:57:49 JaMa: git.pokylinux.org == git.yoctoproject.org Mar 09 13:58:19 hmm I still don't see 15a4d93dab2765e51eb971711e956019ffbf73de in the repo Mar 09 13:58:19 JaMa: 99.9% sure its the same server Mar 09 13:58:24 yohboy: depending on how you build your SD image then yes changes the bitstream should cause the image to be rebuilt (same if you were to change the dtb/kernel/etc). Mar 09 13:58:30 JaMa: I updated again, sorry Mar 09 13:58:39 will download from cgit then Mar 09 13:58:42 JaMa: http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/wip-rss2&id=786b118ae3b436f1ff67399e1a3c37ae3602af6b Mar 09 13:59:05 JaMa: starting to fix function names, write a commit message and so on ready for AB testing Mar 09 13:59:24 ok, fetched now Mar 09 14:00:38 nrossi: I've looked at an example rpoject that uses the linux/dmaengine, dma-mapping (that's the kernal dma api I assume) stuff as well as the linux/amba/xilinx_dma. using that stuff it looks like he writes to dma from userspace, wihtout his own kernel module Mar 09 14:02:12 yourfate: give it a go, :). As far as I am aware it should work fine from userspace since that is what the Xilinx HLS tools are supposed to do. Mar 09 14:04:45 nrossi: actually, i'm student and my project is to provide a system on linux on a zybo, where we can easily change hdf, kernel driver etc, remotely. So maybe the SD card is not the best solution. Maybe TFTP booting method will allow me to change the configuration remotely, without access to the zybo ? Mar 09 14:05:07 what's the advantage of writing a kernel module then? Mar 09 14:05:28 yohboy: yep thats the power of u-boot, you should be able to fully PXE-like boot a board from a remote server Mar 09 14:06:24 yourfate: no copying of memory from user to kernel to user. But for a 60Mb throughput that is likely not really problematic Mar 09 14:06:25 I found that the default settings for the zybo-linx-bd thing does not write a uEnv.txt, so you'll have to configure uboot manually Mar 09 14:06:50 nrossi: that so nice, I need to try it ! I can load the bitstream also remotely with uboot ? even if it is not in SD card ? :) Mar 09 14:07:09 yohboy: you can load the bitstream even from running linux Mar 09 14:07:23 yohboy: thats actually how i test on my zybo. Have u-boot spl and u-boot.img in QSPI and net boot the bitstream/etc Mar 09 14:08:02 nrossi: i only use the SD card for rootfs when they are too large for ramdisk :) Mar 09 14:08:13 yohboy: ^ (talking to myself :|) Mar 09 14:08:30 RP: flex seems to be new requirement, on the box where I'm usually building it wasn't installed Mar 09 14:09:31 nrossi: ok, that seems nice ! I'm not sure to understand the QSPI method yep, I have to find some doc to read :) Mar 09 14:09:52 JaMa: I definitely got errors when it was missing Mar 09 14:10:02 yourfate: I can do that after the boot so ? Mar 09 14:10:08 yes Mar 09 14:10:14 there is a device names /dev/xilsomething Mar 09 14:10:14 JaMa: I think that was from binutils or gcc Mar 09 14:10:25 and you can just do Mar 09 14:10:34 cat bitstream.bin > /dev/xilsomething Mar 09 14:10:43 do flash the fpga Mar 09 14:10:43 meta/recipes-devtools/binutils/binutils.inc:DEPENDS = "flex-native bison-native zlib-native gnu-config-native autoconf-native" Mar 09 14:10:44 to Mar 09 14:11:24 yourfate: yohboy: /dev/xdevcfg, but thats in the linux-xlnx kernel only. The fpga_manager stuff in mainline does not provide that loading mechanism Mar 09 14:11:30 RP: I think binutils and gcc were using flex-native, because I was seeing them failing to build with flex-2.6.3 in oe-core Mar 09 14:11:57 right, but afaik the default kernel used is the linux-xlnx Mar 09 14:12:04 when using that machine Mar 09 14:12:09 yourfate: nice thank you, so I just have to send the bitstream by ssh with scp for example, then use your method ? Mar 09 14:12:31 yohboy: if you mount the rootfs over network anyways you can just copy that into the folder Mar 09 14:12:42 or scp it over there or something, right Mar 09 14:12:49 yourfate: yep, but just so you guys were aware, regarding the previous discussion around which kernel to use Mar 09 14:12:59 JaMa: I just know I saw build failures without it :/ Mar 09 14:13:01 nrossi: maybe the linux-xlnx kernel will be easier for me ? what are the advantage to use the linux-yocto kernel ? Mar 09 14:13:28 yourfate: yes you're right Mar 09 14:16:13 yohboy: the problem is Xilinx has been slow to finish upstreaming everything for zynq. The two main missing pieces are the devcfg (for loading fpga) and QSPI flash. For the fpga loading fpga_manager is available in the mainline kernel however it is not exposed to userspace yet, so you would need to load the bistream via a kernel module. Otherwise linux-yocto Mar 09 14:16:13 is better for all the same reasons as the mainline kernel being better. It really is up to you to decide what works best for you. Mar 09 14:17:02 while we are at it, yocto vs pentalinux? Mar 09 14:17:16 I think yocto is more reliable Mar 09 14:17:38 I tried getting the petalinux toolchain to work, but at some point I rage-rm -rf'd it all Mar 09 14:17:40 and used yocto Mar 09 14:17:50 so would be nice to know what I'm missing Mar 09 14:18:10 yourfate: 'peta'linux :), and petalinux is now just tooling and a layer for yocto so. Yocto/OE :) Mar 09 14:18:26 oh, right Mar 09 14:18:50 JaMa: seems we have builders without flex too. Will try disabling (ditto unlzma) Mar 09 14:19:25 I think i've learned more in an hour here than from readling a day's worth of xilinx docs... Mar 09 14:19:44 I learn more here than 2 weeks alone Mar 09 14:19:48 really thank you Mar 09 14:20:26 yourfate: yohboy: helps your talking to the guy that started Yocto/OE support from within Xilinx :) Mar 09 14:20:34 :D Mar 09 14:20:40 nice to meet ppl like you on irc Mar 09 14:21:01 I agree :) how long do you work on this project ? Mar 09 14:23:13 nrossi: it (qemu-native) seems to be at least partially caused by your changes, it now depends on BLUEZ variable which in my case happens to be MACHINE specific Mar 09 14:24:10 Hi everyone, is there a way to clean the temporary files better than clean and cleansstate ? I modified a patch file but bitbake can't see it modified, it stays on the old version and always gives me the old error... Mar 09 14:24:17 JaMa: Ah, that would explain it. Mar 09 14:25:45 JaMa: But that shouldn't cause the qemu do_configure to change unless you have bluez enabled in the packageconfig? Mar 09 14:27:30 nrossi: give me sec to check -e output, now I'm testing to set BLUEZ_class-native explicitly to bluez5 Mar 09 14:28:46 nrossi: List of dependencies for variable PACKAGECONFIG[bluez] is {'BLUEZ'} Mar 09 14:28:51 Hey, i have a kernel module that has to be installed but it doesnt work, when doing modprobe modulenname i get : could not insert 'modulename' : exec format error Mar 09 14:28:59 Does anyone have an idea what could be done? Mar 09 14:29:12 nrossi: List of dependencies for variable PACKAGECONFIG is {'PACKAGECONFIG[kvm]', 'PACKAGECONFIG[nss]', 'PACKAGECONFIG[gnutls]', 'PACKAGECONFIG[vnc-png]', 'PACKAGECONFIG[xfs]', 'PACKAGECONFIG[fdt]', 'PACKAGECONFIG[gtk+]', 'PACKAGECONFIG[gcrypt]', 'PACKAGECONFIG[libiscsi]', 'PACKAGECONFIG[uuid]', 'PACKAGECONFIG[libcurl]', 'PACKAGECONFIG[ssh2]', 'PACKAGECONFIG[libcap-ng]', 'PACKAGECONFIG[curses]', Mar 09 14:29:18 'PACKAGECONFIG[xen]', 'PACKAGECONFIG[bzip2]', 'PACKAGECONFIG[numa]', 'PACKAGECONFIG[aio]', 'PACKAGECONFIG[sdl]', 'PACKAGECONFIG[vnc-jpeg]', 'PACKAGECONFIG[alsa]', 'PACKAGECONFIG[lzo]', 'PACKAGECONFIG[vnc-sasl]', 'PACKAGECONFIG[bluez]', 'PACKAGECONFIG[libusb]', 'PACKAGECONFIG[virtfs]', 'PACKAGECONFIG[glx]', 'PACKAGECONFIG[nettle]'} Mar 09 14:29:31 nrossi: so it gets the dependency on BLUEZ variable with or without bluez actually enabled Mar 09 14:30:16 nrossi: I don't mind forcing BLUEZ_class-native in our setup, so take this just as FYI if someone else hits this issue as well Mar 09 14:30:29 hopefully not many people need to support 5 different bluez stacks :) Mar 09 14:30:32 maka_: that sounds like your kernel module was compiled for the wrong architecture Mar 09 14:31:05 yourfate: that could be, it also said something about the version magic being different Mar 09 14:31:35 Would there be a fix for that ? besides --force-vermagic Mar 09 14:32:08 I don't know about versionmagic Mar 09 14:33:14 JaMa: i guess this is just a limitation with having distro options populating -native configs :( Mar 09 14:36:47 JaMa: would expanding out the packageconfig to be [bluez4] = ...,bluez4 and [bluez5] = ....,bluez5 with the same config options cleanly solve this issue? Mar 09 14:43:58 nrossi: that would help if ${BLUEZ} variable gets used only in target specific PACKAGECONFIG (default value) Mar 09 14:44:26 nrossi: unfortunately it isn't clear how BLUEZ is supposed to be used Mar 09 14:44:44 nrossi: half oe-core recipes are using it to pick between 2 PACKAGECONFIGs, e.g. PACKAGECONFIG ??= "${@bb.utils.contains('DISTRO_FEATURES', 'bluetooth', '${BLUEZ}', '', d)} \ Mar 09 14:45:13 nrossi: in 2nd half it's used to set DEPENDS and RDEPENDS like you did in qemu Mar 09 14:45:45 JaMa: thats why I did it that way, since QEMU is not picky on the version. Mar 09 14:48:37 nrossi: only as long as you consider bluez[45] as the only options for BLUEZ Mar 09 14:49:09 bluedroid, blueangel and other weird bluez implementations might not be enough even for qemu Mar 09 14:54:33 RP: FYI chromium build might be doing something weird with PATH, even with ld symlink correctly in place I still see | collect2: fatal error: cannot find 'ld' Mar 09 14:54:44 RP: will report back when I have more details Mar 09 14:56:41 nrossi: if I want to have the kernel dma api, I need to activate it in the linux-xlnx menuconfig? Mar 09 14:58:16 JaMa: is it looking for ld.gold? Mar 09 14:58:29 JaMa: I had to add ld.bfd after a similar experience Mar 09 15:00:10 it says just ld, but I have gold enabled, so adding ld.gold could help, thank trying.. Mar 09 15:00:43 nrossi: again me ^^ assume I have u-boot in a partition of my SD card, my zybo boot on it, I need to be connected by USB to see the serial console ? or can connect me to the zybo remotely ? then can I load the kernel, rootfs and DTB into the SD card , to be able to boot on SD card, and so have much memory size ? Where it is load by default ? Mar 09 15:02:11 yohboy: you can have wic build an sd card for you Mar 09 15:03:24 yourfate: yes, but assume I can't access the SD card, it is on my zybo, and the only access I have on the zybo is remotely, not sure wic will work ? Mar 09 15:03:46 ah, wic will create a image locally, which oyu have to dd to the sd card Mar 09 15:04:27 yes, so I can't use it Mar 09 15:05:51 right Mar 09 15:06:24 I wonder if: a user start the zybo (nothing on it, just u-boot I think), connect to the server tftp, then load the file Mar 09 15:06:34 yohboy: U-Boot has a netconsole feature, maybe thats what you are after? http://git.denx.de/?p=u-boot.git;a=blob;f=doc/README.NetConsole;h=af7fc6043a3f3f02497f7527342b18551402be9d;hb=HEAD Mar 09 15:06:55 then the zybo could boot on SD card Mar 09 15:07:06 some part missing in my head ^ Mar 09 15:08:21 nrossi: I changed the kernel to linux-xlnx-dev in meta-xilinx/conf/machine/zybo-linux-bd-zynq7.conf, but it fails with the following error: ERROR: linux-xlnx-dev-4.6+-xilinx-dev+gitAUTOINC+095872028d-r0 do_kernel_version_sanity_check: Package Version (4.6+-xilinx-dev+gitAUTOINC+095872028d) does not match of kernel being built (4.9). Please update the PV variable to match the kernel source. Mar 09 15:08:32 nrossi: maybe it could help, if user can't access zybo for example, I'll read it Mar 09 15:09:38 yourfate: Go edit the .bb file and set the LINUX_VERSION = "4.9", there was a patch for oe to handle disabling that check, need to set it up for linux-xlnx-dev Mar 09 15:10:33 o/ Mar 09 15:11:27 yourfate: Or set "KERNEL_VERSION_SANITY_SKIP = "1"" in your local.conf... Thats the magic variable that needs to be setup :) Mar 09 15:11:41 right, I changed the .bb file, seems like the build started now Mar 09 15:14:19 nrossi: I think it is what I need, just have to understand how to use it, thank you :) Mar 09 15:19:23 https://autobuilder.yocto.io/builders/nightly-non-gpl3/builds/184/steps/BuildImages/logs/stdio - ick Mar 09 15:19:49 I think we need to turn off man generation for that as its clearly wrong anyway Mar 09 15:26:36 nrossi: now it fails with Computing transaction...error: Can't install kernel-devsrc-1.0-r0@zybo_linux_bd_zynq7: no package provides /bin/awk Mar 09 15:27:26 yourfate: hmmmm what version of meta-xilinx/oe-core are you using? Mar 09 15:28:05 afaik i've checked out the morty branch Mar 09 15:28:31 55c835c73cc41b6fc809c941c295d62a612e49e0 Mar 09 15:28:31 yourfate: for both meta-xilinx and oe-core? Mar 09 15:28:32 is the commit Mar 09 15:28:34 yes Mar 09 15:30:03 commit 55c835c73cc41b6fc809c941c295d62a612e49e0 of meta-xilinx Mar 09 15:30:35 do you know what is the layer/recipe which provide u-boot ? Mar 09 15:30:52 yourfate: they are the same sha, for a oe commit :P Mar 09 15:31:02 right Mar 09 15:31:04 yohboy: u-boot is in oe-core/meta Mar 09 15:31:04 sec Mar 09 15:31:14 9304e43528faab0221ef35a3a129a438715c52b2 Mar 09 15:31:16 is meta xilinx Mar 09 15:31:23 copying from vm to vm sometimes tricks you Mar 09 15:31:26 nrossi ty :) Mar 09 15:31:33 or maybe it's a sha1 collision :P Mar 09 15:32:11 yourfate: ... :P, your on the right commits. What distro/distro version you using? Mar 09 15:32:28 ubuntu 16.10 :/ Mar 09 15:32:33 i know it warns me about the version Mar 09 15:32:38 but it has worked fine so far Mar 09 15:33:03 like, it built fine before switching to linux-xlnx-dev Mar 09 15:34:49 yourfate: give me a bit to check, im in the middle of emails and stuff ;) Mar 09 15:35:49 no problem. I'll be here for another 30 mins I think Mar 09 15:59:44 nrossi: I'm going to clock out for today. if you find anything just ping me, my bouncer will tell me. thanks for the input, you've been really helpful :D Mar 09 16:00:02 yourfate: okay Mar 09 16:55:43 halstead: are you around? Mar 09 17:30:03 lamego, I'm here now. Mar 09 17:44:36 rburton: I'm looking at selftest breakage. Strange, all tests have passed in my env. Mar 09 17:45:45 ed2: fun isn't it. have a look at what merged into next in case anything there might have broke it Mar 09 17:46:25 rburton: yep, will do Mar 09 18:10:23 anybody else getting nss fail? Mar 09 18:11:07 it worked with meta-printing on krogoth, fails on morty Mar 09 18:11:26 big gcc bump between those two Mar 09 18:12:30 i did have to fix some python stuff for morty (which works now) Mar 09 18:19:59 mr_science: nss is in oe-core Mar 09 18:20:13 if it failed lot of us would have noticed Mar 09 18:21:27 doesn't like either gold or bfd here... Mar 09 18:22:52 i switched off the hardened/pax kernel but it's still a hardened toolchain/userland Mar 09 18:47:22 what errors do you see. Mar 09 18:47:29 pie may be issue Mar 09 19:22:19 Hi Mar 09 19:23:07 I want subscribe to gnome outreachy Mar 09 19:25:18 so nobody's seen this error except me? Mar 09 19:25:22 error adding symbols: Archive has no index; run ranlib to add one Mar 09 19:25:54 on libfreebl.a in nss build Mar 09 19:26:19 is there any guideline to start contribute to this project? Mar 09 19:28:05 what the heck is a 'gnome outreachy' and why would you come to #yocto for it? Mar 09 19:29:25 trinity: that's great! if you could send an email to jeffrey.osier-mixon@intel.com - he should be able to help you get involved with outreachy Mar 09 19:30:18 trinity: admittedly we don't have great newcomers documentation right now but someone is working on it Mar 09 19:31:27 bluelightning: jefro mentioned that but didn't give details Mar 09 19:31:43 what's the story? Mar 09 19:31:48 nerdboy: stephano is the one working on that FYI Mar 09 19:32:10 (he volunteered to at OEDAM) Mar 09 19:32:22 trinity: my advice would be, run through the getting started guide and if that all works well see if you can try to do something more interesting (e.g. build an application you're interested in / target a device you have) Mar 09 19:33:53 nerdboy: I'll be adding info to the wiki in the coming weeks and proposing some changes to bugzilla to the group. Mar 09 19:34:39 * nerdboy still trying to figure out the outreachy connection... Mar 09 19:36:00 nerdboy: jefro https://wiki.gnome.org/Outreachy/2017/MayAugust#Yocto_Project Mar 09 19:36:25 https://www.yoctoproject.org/outreachy-internship Mar 09 19:41:24 that's right, so you guys are buying an intern? Mar 09 19:41:53 * nerdboy was also thinking Gentoo should spend some money on that... Mar 09 19:49:07 you'll have to check with jefro for details, I'm not sure. Mar 09 19:59:00 iirc, that's what the outreachy lady said at gsoc mentor mtg Mar 09 19:59:31 as in you need to bring your own money to "rent" an outreachy intern Mar 09 20:04:00 nerdboy: the term "fund" is perhaps more appropriate, but yes it's a paid internship programme Mar 09 20:25:21 rburton: can you, please? https://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/log/?h=akanavin/dnf-rpm4 Mar 09 20:25:29 the AB, I mean :) Mar 09 20:26:12 kanavin: you got me just as i was about to go watch Expanse, good timing Mar 09 20:26:43 kanavin: running on the main AB now Mar 09 20:26:46 now i'm off :) Mar 09 20:27:29 cheers! Mar 09 20:34:56 rburton: cdrtools-native failed to build when the build is run from test case: https://gist.github.com/bartosh/0b540ab65f1ecee80f31e20545f490e7 Mar 09 20:35:21 rburton: i run the build manually and it was built just fine. Mar 09 20:39:12 Is anyone else seeing the boost-1.63 libboost_numpy.so.1.63.0 installed but not shipped QA issue? Mar 09 20:39:35 fischerm, ^^^ Mar 09 20:56:22 Ah, looks like I should have searched my mail first. appears there is a patch already, but it isn't merged yet. Trying it out now. Mar 09 21:08:46 seems to have avoided the issue for me Mar 09 21:28:59 Hi folks, I have a board with requires a few binaries to be flashed in order to work, I'd like to make a recipe to fetch the tarball that contains them and deploy them in the deploy dir but without installing them in the rootfs. Any hint on how I could achieve that? Mar 09 21:29:49 inherit deploy, write a do_deploy task, make sure you pay attention to the difference between DEPLOY_DIR and DEPLOYDIR Mar 09 21:30:20 (if you don't, your recipe will work as long as sstate is not being used, and once it is, your files will just vanish) Mar 09 21:31:39 okay, I'll double check. I already have a task that inherit deploy with a do_deploy, but some stuff are also installed in the rootfs and I'm trying to figure out why Mar 09 21:32:00 I guess it's because I also have a do_install task Mar 09 21:32:43 possibly, yes. You don't need that with deploy recipes Mar 09 21:40:35 hum... in that case I get "Cannot install package mypackage" during do_rootfs Mar 09 21:43:27 my recipe looks like this: http://pastebin.com/NqEFMZJf Mar 09 21:46:03 I include this recipe in my IMAGE_INSTALL, don't know if I should do that or not, but I'd like to have these binaries in the deploy dir when the image is built Mar 09 21:53:26 jmesmon: thanks for testing Mar 09 22:04:03 yes, I can prove the sstate cleanup script guilty of the "nfs" problems! :) Mar 09 22:04:08 halstead: ^^^ :) Mar 09 22:05:00 AFK for a short bit. Mar 09 22:05:41 halstead: will send email Mar 09 22:13:44 good one RP Mar 09 22:14:10 moto-timo: its a bug that has been causing problems for *years* :) Mar 09 22:14:19 I remember Mar 09 23:08:39 RP: does atime work over nfs? struggling to see why it would be deleting a file the build needs and still have an atime of more than 10 days Mar 09 23:09:57 rburton: Not sure. It does at least prove where the issue is though Mar 09 23:10:39 RP: unable to find tunctl? http://pastebin.com/yLn091DX Mar 09 23:11:03 ant_home: probably needs fixing Mar 09 23:11:13 PN is defaultpkgname :/ Mar 09 23:11:46 pls try with a normal qemu image Mar 09 23:12:06 ant_home: you're asking me to try? Mar 09 23:12:26 yes, maybe is some local change here Mar 09 23:12:58 it tried to run tunctl of host (password was required) Mar 09 23:13:36 ant_home: Its after 11pm at night here, I've been working since 8am trying to make the PATH whitelist stuff work and I'm worn out. My local setup works but uses preconfigured tap devices. I'm sure I can break my workspace to test this but it is a bit of an ask :/ Mar 09 23:13:38 RP: absolutely. at least, schedule it to run when there are no builds maybe. Mar 09 23:14:14 RP: sleep. I ask now because you were touching host tools all day long :) Mar 09 23:14:15 rburton: or just -name *.tgz and whatever else, miss the tmp files Mar 09 23:15:09 rburton: I do worry that its finding other files it shouldn't and causing mystery eSDK failures Mar 09 23:15:46 RP: actually yes, ensuring it can't falsely hit a tmp file would help Mar 09 23:16:09 RP: how many files does it delete on each run? Mar 09 23:18:22 rburton: since I enabled it the other day, 99399 files Mar 09 23:18:35 jesus Mar 09 23:18:56 how many files does it not delete? :) Mar 09 23:19:31 rburton: the ab sstate cache is huge Mar 09 23:19:40 apparently so Mar 09 23:20:14 curious if there's a race on file appearing and stat entries being sane Mar 09 23:20:26 but using -name should stop that problem Mar 09 23:20:48 anyway g'night Mar 09 23:22:25 rburton: 'night! Mar 09 23:22:39 rburton: I did try and get a number of files but its a slow command **** ENDING LOGGING AT Fri Mar 10 03:00:03 2017