**** BEGIN LOGGING AT Tue Mar 12 02:59:57 2019 Mar 12 04:24:21 smurray: for tthis, I'd take a bitbake branch with the changes on Mar 12 04:24:45 smurray: I'm sure I've see a list of the commits needed somewhere Mar 12 04:25:56 RP: there's something like 7-8 commits, iirc. After poking around earlier I was figuring I could do a patch series against 1.40, but a branch works Mar 12 04:29:17 RP: so submit a pull request once I've got something, I guess? I don't have autobuilder access atm, should I get request that and do a run through via poky-contrib, perhaps? Mar 12 04:30:25 smurray: patch series works fine as long as its against bitbake and makes it clear they're all cherry picks and how clean the picks were Mar 12 04:31:10 smurray: happy to run such a series through the autobuilder but its mainly bitbake selftest we need to have passing in this case Mar 12 04:31:25 smurray: I would want to see the tests included/backported Mar 12 04:31:46 RP: okay, I'll work up something tomorrow, and make sure the tests are there (and pass ;) ) Mar 12 04:32:32 RP: thanks! Mar 12 08:07:53 re Mar 12 08:08:53 if i have a package that's built with own provided configure -> makefile, i still should import autotools to build it right? Mar 12 08:09:14 and just specify my options to configure in EXTRA_OECONF Mar 12 08:14:17 right, looks like it Mar 12 08:17:48 varjag: um, no Mar 12 08:18:01 varjag: you should import autotools if the package is autotoolized Mar 12 08:18:58 varjag: if the package is autotoolized AND brings a configure script in the sources, then importing autotools will discard that script and regenerate it, as far as i know Mar 12 08:19:27 but the real point is "is the package autotoolized", and not "does it bring a configure/Makeefile" Mar 12 08:21:49 hm.. it looked like autotools.bbclass would run oe_runconf if the configure is there anyway Mar 12 08:22:02 that just what i said, right? Mar 12 08:22:23 but the package is not autotoolized Mar 12 08:24:17 autotools.bbclass is really a little bit more than just configure; make; make install Mar 12 08:24:20 https://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/classes/autotools.bbclass#n149 Mar 12 08:26:29 yup, what i mean the package has no configure.ac or makefile.am but upstream-supplied configure, and wondered if autotools was still idiomatic way to build such packages in yocto Mar 12 08:26:30 ah ok, but this seems to be limited to the case where the source is autotoolized AND brings the autotools sources. where as shipping generated configure + Makefile will jump a couple of steps Mar 12 08:27:04 varjag: that sounds very much like the devs of that package use autotools internally, but do not push its sources. Mar 12 08:27:51 it's a bit more complicated than that, it is a compiler that has a sort of 2-stage/bootstrap build Mar 12 08:28:12 how I can add meta layers from the yocto esdk ? Mar 12 08:28:22 so i believe there's quite a lot of manual tuning done Mar 12 08:28:23 alicef: from? Mar 12 08:28:59 varjag: in that case, i'd probably use autotools.bbclass as a template, and pour everything into a custom .inc for that specific package. Mar 12 08:29:20 LetoThe2nd: from a openembedded layer for example Mar 12 08:29:24 mm.. makes sense Mar 12 08:29:31 is it even possible ? Mar 12 08:29:47 alicef: you mean you want to add a layer TO the esdk? Mar 12 08:29:57 yes Mar 12 08:30:33 alicef: that should work just like in any other situation too. check out the desired layer, add it to bblayers.conf Mar 12 08:31:01 oh ok sounds easy Mar 12 08:31:16 but no guarantees that it doesn't break the build if its something that brings invasive changes Mar 12 08:31:32 so this is more of a "technically easy" Mar 12 08:31:57 by the way there is a way to use devtool add but keep the recipe if already present Mar 12 08:32:06 and not recreating it from new Mar 12 08:35:04 I think i can probably make a bash script that ovveride the recipe after importing it Mar 12 08:35:15 but i was thinking if there is a better way Mar 12 08:52:31 Hi o/ Mar 12 08:52:47 did any one had problem with PSEUDO during rootfs creation ? Mar 12 08:53:02 winotu: certainly. Mar 12 08:53:18 so, what good did that answer do you? ;-) Mar 12 08:53:23 LetoThe2nd: during package installation ? Mar 12 08:53:56 winotu: in case you did not get the clue: just ask about your *exact* problem instead of a chain of meta-questions, please. Mar 12 08:54:11 LetoThe2nd: i get it :] Mar 12 08:54:55 LetoThe2nd: when dnf tries to install packages during do_rootfs it shows error with lack of disk space Mar 12 08:55:42 and you are sure that you're not running out of space? specifically the image size that you might have specified? Mar 12 08:56:07 for now I narrowed it to dnf invokation under PSEUDO, and for now I don't have any clue where to find any further place to find the root cause Mar 12 08:56:40 LetoThe2nd: yes i checked /dev/shm /tmp and filesystem on which the rootfs is created/installed with dnf command Mar 12 08:57:17 LetoThe2nd: image size variables are not used during this phase Mar 12 08:58:05 i see. anything highly specific about the host you build on? or generally about your build? Mar 12 08:58:26 LetoThe2nd: or at least I can'd find in code any referrence to them when PSEUDO is started and rootfs is created before dnf is used for packages installation Mar 12 08:58:27 (just trying to gather information for those who might read along or backlog) Mar 12 08:58:46 LetoThe2nd: ext4, rh7 Mar 12 08:58:50 LetoThe2nd: any idea about devtool add ? Mar 12 08:59:00 alicef: no. if i had, i would have told. Mar 12 08:59:12 RP: do we test on RHEL7? Mar 12 08:59:14 LetoThe2nd: thanks Mar 12 09:03:59 winotu: has this already worked once? are you on master? Mar 12 09:16:50 LetoThe2nd: yes it worked for some time, i'm using rocko poky Mar 12 09:17:13 officialy Yocto supports Centos and Fedora Mar 12 09:19:21 winotu: i'm not saying that its unsupported per se, just that sometime things break through an update. and if its in the test matrix we have a chance that its automagically catched. Mar 12 09:20:10 aah ok Mar 12 09:20:11 winotu: but then the real question is: when did it stop working :) Mar 12 09:20:20 no poky update Mar 12 09:20:41 I just added new stuff to image and boom Mar 12 09:20:53 and which stuff did you add? Mar 12 09:21:03 (you did not mention that so far, btw.) Mar 12 09:21:07 tpm2 ex Mar 12 09:21:22 just more rpm packages for installation Mar 12 09:21:41 that shouldn't be a problem during rootfs packages installation Mar 12 09:22:03 well... "shouldn't" Mar 12 09:22:06 and the same issue I had some time ago without changes Mar 12 09:23:13 then I thought that /tmp was full Mar 12 09:23:27 and used differenet host to build Mar 12 09:24:55 and if you revert the change things work again? Mar 12 09:27:50 LetoThe2nd: not checked yet as I need new packages in image. So I focused how or where to increase size of rootfs where dnf install packages. Mar 12 09:28:34 winotu: hey, a couple of minutes ago you claimed to have verified that the size is not checked at that stage. so what now? Mar 12 09:28:55 LetoThe2nd: IMAGE_ROOTFS_SIZE and MAXSIZE Mar 12 09:29:05 yes, and? Mar 12 09:29:12 LetoThe2nd: i wrote that already checked that i'm in a corner Mar 12 09:29:32 seems that pseudo provides for dnf wrong that about free space Mar 12 09:29:46 or DNF has this info in some config file ? Mar 12 09:30:08 i'm just suggesting that by reverting the change you can verify that the change actually triggered the problem, instead of maybe some other, coidential happening. Mar 12 09:30:11 *provide wrong information about free space Mar 12 09:30:41 LetoThe2nd: it's possible as I have seen in past similiar issue Mar 12 09:31:48 for now looking in the bbclass es and checking all the python code narrowed me to usage of dnf under pseudo Mar 12 09:49:47 is it common to have fetch fails from sourceforge repos, or is it just me Mar 12 09:53:14 varjag: create PREMIRROS repo on Your local host and forget for sometime about any external communication Mar 12 10:14:41 winotu: great idea Mar 12 11:03:35 LetoThe2nd: i checked building on different host machine ( spec is the same for host env and hw) is it possible that some FS error on host shows? Mar 12 11:04:23 winotu: i'm no intimate knower of pseudo, so, i can't tell. Mar 12 11:04:31 RP: ^^^^ maybe you. Mar 12 11:05:49 LetoThe2nd: I didn't include important sentence, and it works on that separate machine. Same changes same package size for installation. Mar 12 11:07:34 winotu: tbh, you didn't include important bits and pieces for several times, and therefore i am not really inclined to spend much more effort here. my gut feeling is that you have indirectly triggered some strange behavious, maybe even bug. maybe its related to different package versions of the host. Mar 12 11:07:50 like i said, maybe RP can tell more, i cannot. Mar 12 11:07:55 (and its lunch now, YAY!) Mar 12 11:13:11 LetoThe2nd: I don't want to to take Your time, I'm just looking for hints/tips where to look. If I could find the information easly in code I wouldn't ask here for them. And as I already stated for me it's some PSEUDO issue when dnf is exectued. Mar 12 11:14:49 when do_rootfs is run PSEUDO is used and python run subprocess to run dnf installation of package in specified rootfs. Mar 12 11:17:13 if the WORKDIR/rootfs is not an special filesystem created under PSEUDO, one and only issue can come from pseudo db about used filesystem which is used then by dnf for free space calculation Mar 12 11:18:51 maybe my pseudo binary is built somehow incorrectly I will check that Mar 12 11:37:04 Is this an appropriate place to ask a quick question for using Yocto, not developing? Mar 12 11:38:20 jrypkec: yes Mar 12 11:40:12 I'm trying to find a list of what packages are available via rdepends Mar 12 11:42:13 QA is complaining about missing libudev. Theres a bunch of bindings availble but not the actual package on http://layers.openembedded.org/layerindex/branch/master/recipes/?q=libudev Mar 12 11:42:51 So I'm guessing udev is generally available somewhere Mar 12 11:44:44 jrypkec: The search you're using gives you recipe names, not package names Mar 12 11:44:52 RDEPENDS expects package names. Mar 12 11:45:45 in openembedded-core, the udev recipe provides the libudev package, see http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-core/udev/eudev_3.2.7.bb?h=master#n68 Mar 12 11:46:08 ok. Is there a way to print the rdepends possible values? Mar 12 11:46:11 Providing a searchable index of packages is not always easy, because package names might be generated at runtime Mar 12 11:46:57 E.g., in recipes that provide language-specific packages, there's no line PACKAGES += "package-${lang}" for every language, instead there's a rule that generates the packages according to a regular expression while packaging. Mar 12 11:47:29 The easiest way to get a list of available packages for your system is to check your package feeds (where exactly those live depends on which package manager you're configured to use) Mar 12 11:50:17 ok thanks Mar 12 12:18:18 rburton: we have a problem. Something is wrong with the autobuilder. Looks like hanging fetches :( Mar 12 12:18:31 urgh Mar 12 12:19:07 rburton: I suspect a hung build/slow fetch somewhere which is blocking everything else with a lock Mar 12 12:19:38 rburton: I tried asking halstead before I collasped with jetlag but its not looking well this morning :( Mar 12 12:20:23 this is a nightmare scenario with me travelling :( Mar 12 12:23:04 * RP ponders what to do about it Mar 12 12:28:09 LetoThe2nd: I maybe will have some finding, apparently host / dir is full, which I'm not using for any building and apparently dnf through pseudo uses wrong root dir for checking free space. I'm testing this on different host with similiar scenario. Mar 12 12:33:30 winotu: hum. i kinda doubt that. but of course, its possible. Mar 12 12:46:23 winotu: oh that was fixed Mar 12 12:46:58 rburton: pyro in that case. Mar 12 12:47:19 rburton: which poky ? Mar 12 12:47:33 definitely master Mar 12 12:47:36 trying to find the commit now Mar 12 12:47:45 i'm using 2.4 rocko Mar 12 12:48:30 ah rocko, my bad. Mar 12 12:53:04 kanavin might remember what the fix was but iirc he is on holiday now Mar 12 13:06:59 Hi! I'm trying to install/enable a wpa-supplicant systemd service from another recipe by adding `SYSTEMD_SERVICE_${PN} = "wpa_supplicant@wlan0.service"` but I'm getting the error: SYSTEMD_SERVICE_... value wpa_supplicant@wlan0.service does not exist. Any ideas? I already tried adding wpa-supplicant as DEPENDS and RDEPENDS Mar 12 13:18:17 Is there an easy way to have Yocto build both the aarch64 and aarch toolchains? Mar 12 13:18:18 We have a mixed SoC that has some of both cores. Mar 12 13:26:23 Quazil: you mean armv 32bit ? and arm 64 bit ? Mar 12 13:26:46 Quazil: and probably You would like to use multilib in one image ? Mar 12 13:27:45 Quazil: if yes then https://www.yoctoproject.org/docs/latest/mega-manual/mega-manual.html#combining-multiple-versions-library-files-into-one-image Mar 12 13:44:12 Hi guys, I'm trying to do some easy thing in yocto but somehow it doesn't work as I would expect Mar 12 13:44:39 I'm trying to link a program with asound lib, includes are in /usr/include and lib is in /usr/lib Mar 12 13:44:57 however it doesn't seem to compile out of the box when I'm trying to do things as follow : Mar 12 13:45:33 ${CC} alsa_tester.c -o alsa_tester -lasound Mar 12 13:45:48 I added -I/usr/include which improved a little bit Mar 12 13:46:04 but then it said that it doesn't have the libasound Mar 12 13:46:24 so I added -L/usr/lib which didn't fix the issue. Mar 12 13:46:44 I think there's something obviously wrong with what I'm doing... Mar 12 13:46:46 yacar_: You definitely don't want to be pulling in host includes & libraries during the build Mar 12 13:47:22 You need to add the relevant packages to DEPENDS in your recipe Mar 12 13:48:14 Oh ok I see, I'll try that :) Mar 12 13:48:44 I think it's 'alsa-lib' for libasound but I'm not 100% sure Mar 12 13:49:22 The idea with Yocto Project/OE is that we isolate the build from the host so you don't accidentally link against things that aren't built for the hardware you're targeting Mar 12 13:49:42 yep that's the one, I added it to my image recipe but didn't in my package recipe Mar 12 13:51:01 On recent branches you'll find that recipes only have access to things which are listed as dependencies. That's to stop inconsistent builds when things like autoconf enable/disable features based on what libs are found Mar 12 13:52:11 winotu: Not multi-lib; the Linux image is all aarch64 but we have a real-time core that is 32bit. Mar 12 13:52:11 Maybe I want to build two SDKs? Mar 12 13:54:54 Quazil: You can build own toolchain to build sw for that 32 bit core or use built toolchain to compile for that core binary file and own separate recipe for that Mar 12 13:55:13 Quazil: or You just need SDK for the core sw ? Mar 12 13:57:21 New news from stackoverflow: OpenGL: Mesa3D Offscreen Software Rendering Performance Issues Mar 12 13:57:53 Quazil: https://www.openembedded.org/wiki/Adding_a_secondary_toolchain Mar 12 13:58:05 Awesome, thank you Mar 12 14:01:44 paulbarker: that was it thanks ! Mar 12 14:02:03 I've another problem with the install rule but at least now it's compiling :) Mar 12 14:32:38 kanavin: If You could point me to fix with dnf/pseudo issue I would be really gratefull. Seems that dnf checks / root on host instead rootfs directory for installing packages. Mar 12 15:17:44 Question for the group: Is the main point of BBMULTICONFIG to allow you to build recipes under multiple settings of DISTRO? Mar 12 15:26:14 jtrimbl3, is allows for multiple images to be built with dependences between the images Mar 12 15:26:59 Isn't an "image" a special type of recipe? So why couldn't one "image" depend on another "image" without multiconfig? Mar 12 15:27:37 jtrimbl3: it can, but only if they share the same overall configuration, machine, libc, etc Mar 12 15:27:57 someone may well want to build images for multiple machines in a single run, i.e. for a container vs the hardware, a musl initramfs vs glibc rootfs, etc Mar 12 15:28:12 Ah fun, daadaf7fd7d5566741817e720a0a8c88768a7f28 in kernel-meta broke the qemux86/x86_64 RRECOMMENDS line by dropping the old sound driver Mar 12 15:28:14 It seems like multiconfig becomes necessary when you need to change something like "distro policy" Mar 12 15:28:15 thats where multconfig deps would be helpful Mar 12 15:28:26 or machine, or libc, or basically anything set in global configuration Mar 12 15:28:27 and since it's a recs not a depends, nothing failed. Mar 12 15:29:11 OT, but am I correct in remembering that if you need to combine multiple hashes into a single new one, you use xor? Mar 12 15:31:02 Hey guys - I'm trying to figure out how devtool, recipetool, and bitbake are passing info back and forth so I'm breaking out into pdb Mar 12 15:31:35 the problem is that devtool, bitbake, or *something* is redirecting std out, so I don't get any character echo in PDB Mar 12 15:31:40 ... so I'm typing blind. Mar 12 15:31:51 Anyone know how I can wrench stdout back for that use case? Mar 12 15:32:18 depends on context. Mar 12 15:32:30 I've got something that appends a base class to all recipes via INHERIT+= in my distro policy, but I want to generate a separate "installer" image (which depends on my main image) that doesn't do the INHERIT thing. It seems like the only way to do that is a multiconfig with different "DISTRO=" values, right? Mar 12 15:32:33 are you debugging inside a task, event handler, what? Mar 12 15:32:50 jtrimbl3: or just change your class to have a way to disable its functionality with a variable.. Mar 12 15:33:02 I placed some pdb.set_trace() statements within, for instance, recipeutils.py Mar 12 15:33:19 so I could see what happens when, say, I run 'devtool upgrade xmodmap' Mar 12 15:33:40 you should really just send messages to an appropriate logger instance and let it sort out how to display it. Mar 12 15:33:46 I'd like to pick apart "rd" and see what it holds Mar 12 15:33:49 all the tools use python logging Mar 12 15:34:02 @kergoth: But then it seems like I'd need to build some recipes with two different values of the enable/disable variable Mar 12 15:34:03 K, I'll just have to get more familiar with where that stuff is logged. Mar 12 15:34:06 Thanks kergoth Mar 12 15:34:26 jtrimbl3: or set the variable appropriately in some other context with an override, sure Mar 12 15:34:35 jtrimbl3: FOO_pn-myinstallerimage = "0" Mar 12 15:34:36 or whatever Mar 12 15:34:47 you'll have to have a way to express when it's obeyed and when it isn't regardless Mar 12 15:34:52 the only question is how you want to make it happen Mar 12 15:35:01 and where that knowledge is stored Mar 12 15:35:31 kergoth: Oooh I like that Mar 12 15:35:50 could set it in your distro policy or wherever is most appropraite for your situation Mar 12 15:36:09 But I can have different values of OVERRIDES for different recipes? Mar 12 15:37:18 er, I can build the same recipe multiple times with different values of OVERRIDES? Mar 12 15:37:31 so that one time it will get built with MY_VARIABLE=foo Mar 12 15:37:48 and the second time it will get built with MY_VARIABLE=bar (by setting MY_VARIABLE_myspecialimage=bar)? Mar 12 15:38:35 OVERRIDES always includes 'pn-${PN}' Mar 12 15:38:45 so you can override a variable in any recipe from the config metadata by using it Mar 12 15:38:49 FOO_pn-my-recipe-name Mar 12 15:39:12 you could do that, but you'd have to use multiple bitbake invocations Mar 12 15:39:17 unless you use multiconfig, of course Mar 12 15:39:31 ok so sounds like i might be best off using multiconfig Mar 12 15:40:14 if you want to build multiple configs in a singl ebitbake run, that's what it's for Mar 12 15:40:34 as a first cut I just set up multiconfig (with only one "config" in the multiconfig/) directory and now it's saying: "Uninative selected but not configured correctly, please set UNINATIVE_CHECKSUM[x86_64]" Mar 12 15:40:55 * Tartarus shoots off a patch Mar 12 15:40:59 though that's not necessary in my non-multiconfig case Mar 12 15:42:56 seems that normally UNINATIVE_CHECKSUM gets set from poky/meta/conf/distro/include/yocto-uninative.inc -- not sure why that's not being picked up now... Mar 12 15:54:16 seems like I have to set a "DISTRO" in my local.conf even though my goal is to change DISTRO in each of my multiconfigs Mar 12 15:58:23 zeddii: did you fail to push something? WARNING: linux-yocto-5.0+gitAUTOINC+8fbd119bd9_37c8f2a3df-r0 do_fetch: Failed to fetch URL git://git.yoctoproject.org/linux-yocto.git;name=machine;branch=v5.0/standard/base Mar 12 16:12:47 jtrimbl3: remember local.conf *is* one of the multiconfigs Mar 12 16:12:59 you have local.conf + any multiconfigs set Mar 12 16:13:25 i've a host/container thing going on, local.conf is host and multiconfig/foo.conf is the guests Mar 12 16:37:13 hey guys, just wondering, what the difference between normal and extensible SDK ? Mar 12 17:02:50 rburton: the repo looks corrupted in the shared downloads dir. Have wiped it and letting it start aain Mar 12 18:16:38 rburton: Thanks -- I hadn't realized that. The world is starting to make sense now. Mar 12 18:16:59 Also important to set TMPDIR in my "multiconfig/whatever.conf" or else things get very confused Mar 12 19:20:15 jtrimbl3, multiconfig is also in the mega manual Mar 12 19:44:09 Can I set a separate value of BBMASK inside my "multiconfig/my_config.conf" ? Mar 12 19:47:03 haven't tried it. I would expect that behavior to be supported Mar 12 19:47:48 separate question -- is it possible to have a separate "bblayers.conf" in a separate multiconfig? Mar 12 20:05:50 jtrimbl3: No, not as it works today Mar 12 20:06:36 jtrimbl3: BBMASK would work Mar 12 20:08:52 hmm, https://layers.openembedded.org/layerindex/branch/master/layer/meta-flatpak/ is getting a little old, wonder if anyone is working on that anymore Mar 12 21:52:11 Hi, I'm trying to use the CRI-O recipe in meta-virtualization with rocko. The official layer doesn't support rocko, but I've managed to easily backport the runc layer. The CRI-O layer keeps failing with "attempting to install package into read-only GOROOT" errors, however. Mar 12 21:54:10 s/layer/recipe/g, sorry Mar 13 02:47:32 If I'd like to override a bitbake variable with the "logical and" of two values that might appear in OVERRIDES, how would I do that? Mar 13 02:48:42 E.g. I want to SRC_URI_append if mender-uboot AND a particular value of MACHINE are in OVERRIDES **** ENDING LOGGING AT Wed Mar 13 02:59:57 2019