**** BEGIN LOGGING AT Tue Apr 25 03:00:03 2017 Apr 25 06:29:16 I have a problem with my bitbake core-image-sato, can someone help me please ? The following packages have unmet dependencies: Apr 25 06:29:16 packagegroup-core-ssh-dropbear : Depends: dropbear but it is not going to be installed Apr 25 06:29:16 E: Unable to correct problems, you have held broken packages. Apr 25 06:29:18 [log_check] WARNING: Debian package install does not support BAD_RECOMMENDATIONS Apr 25 07:49:21 I would like to do a 32 bit generic intel image but I have always many mistakes. Can someone help me please ? Apr 25 07:56:28 hamdyaea: you will have to explain what you did and what went wrong... Apr 25 08:02:18 jku : I try to create a genex86 32 image, rpm package, openssh , python, Apr 25 08:09:57 hello, how do I prevent do_image_ubi from executing? I get error Error: max_leb_cnt too low (4072 needed). I only need an sdcard image anyway. Apr 25 08:10:30 hamdyaea packagegroup-core-ssh-dropbear is conflicting with open ssh sins openssh and dropbear are both ssh servers Apr 25 08:11:37 this is my local.conf: https://pastebin.com/CJGRzp94 Apr 25 08:12:30 MarcWe : OK I will remove the openssh package Apr 25 08:12:31 eduardas_m you need to set the MKUBIFS_ARGS -c to a more high number in you device config file Apr 25 08:13:10 MarcWe: what if I dont want to make a ubifs in the first place? Apr 25 08:13:42 I only care about emmc and sdcard Apr 25 08:14:17 but by default the recipes also try to build ubifs Apr 25 08:14:23 eduardas_m: that is somthing i can't anser for you. i soved the issue with setting the -c to 12000 Apr 25 08:15:01 MarcWe: will try, thank you Apr 25 08:16:49 eduardas_m: i personally would look at the bitbake -e of your build, and then inspect the modifications of IMAGE_TYPES. hopefully this tells you where the ubifs gets inserted. Apr 25 08:17:15 I'm still also a bit of a noob :P Apr 25 08:19:06 jku, RP: I'm seeing a problem with the DISTRO_FEATURE removal for native recipes. My example is swtpm-wrappers, which inherits native.bbclass. Apr 25 08:19:33 native_virtclass_handler in native.bbclass checks whether PN ends in -native. Apr 25 08:19:48 Always has (well, since 2011). Apr 25 08:20:43 But that means that swtpm-wrappers uses different DISTRO_FEATURES than other native recipes, and that breaks when using the usrmerge DISTRO_FEATURES because paths are not the same anymore. Apr 25 08:21:40 pohly: it would use different distro_features to the target but it should see the same distro features as all other native recipes Apr 25 08:22:17 RP: the current code returns before changing DISTRO_FEATURES because of the PN check. Apr 25 08:22:21 That looks like a bug to me, Apr 25 08:22:54 Should the DISTRO_FEATURE cleanup be done before the return? Apr 25 08:23:24 I.e. unconditionally, in any recipe inheriting native.bbclass? Apr 25 08:23:26 pohly: no, since distro_features is only changed if the recipe endswith -native ? Apr 25 08:23:53 pohly: are we talking about a recipe which is native but isn't called xxx-native ? Apr 25 08:24:05 RP: yes, exactly that. Apr 25 08:24:12 pohly: don't do that Apr 25 08:24:37 pohly: there are other places which make assumptions about naming Apr 25 08:25:07 RP: is that documented somewhere? If not, shouldn't there be some kind of warning? Apr 25 08:25:23 I wasn't aware of that restriction/. Apr 25 08:25:30 pohly: the bitbake virtclass code is tricky and that handler can be called with target context recipes which is why the guard is there, we can't relax that Apr 25 08:25:38 pohly: how exactly would you detect this? Apr 25 08:25:56 pohly: I'm open to ideas but its hard :( Apr 25 08:26:31 pohly: It could be the issues with the handlers being called out of context may be fixed these days. I know I did make a number of fixes in that area, not sure I ever got them all though Apr 25 08:26:34 I don't know. If there's no way to detect it, at least it should be documented. Apr 25 08:26:44 pohly: happy to see documentation... Apr 25 08:27:31 pohly: grep the codebase for "endswith('-native')" to see what I mean about assumptions about cross/nativesdk/native :/ Apr 25 08:28:07 ERROR: core-image-sato-1.0-r0 do_image_wic: Function failed: do_image_wic (log file is located at /home/hamdy/poky/build/tmp/work/genericx86-poky-linux/core-image-sato/1.0-r0/temp/log.do_image_wic.4615) Apr 25 08:28:07 ERROR: Logfile of failure stored in: /home/hamdy/poky/build/tmp/work/genericx86-poky-linux/core-image-sato/1.0-r0/temp/log.do_image_wic.4615 Apr 25 08:29:52 Error: exec_cmd: cp /home/hamdy/poky/build/tmp/work/genericx86-poky-linux/core-image-sato/1.0-r0/core-image-sato-1.0/hddimg/EFI/BOOT/* /home/hamdy/poky/build/tmp/work/genericx86-poky-linux/core-image-sato/1.0-r0/deploy-core-image-sato-image-complete/core-image-sato-genericx86-20170425081143/build/hdd/boot/EFI Apr 25 08:29:57 Can someone help me please ? Apr 25 08:33:35 RP: the documentation for native.bbclass actually describes the recipe naming ("Create a myrecipe-native.bb that inherits the native class"). It's just not clear that this is more than an example and that things will break when not following that example. Apr 25 08:33:58 Not sure whether it is worth adding a warning. Apr 25 08:53:50 pohly: Probably worth just an extra comment mentioning it? Apr 25 08:54:20 RP: Is the same true for the nativesdk- prefix? Apr 25 08:59:01 is there a simple way to bulid all the packages that would make up an image, but not actually assemble the image? Apr 25 09:02:46 phako[m]: what would be the useccase? Apr 25 09:12:12 assembling various images. with common and disjunct subsets of packages. and no image with a superset of all of them Apr 25 09:12:41 so currently we build an artificial superset image, just to throw it away Apr 25 09:13:00 phako[m]: why not just build them in a row? where its possible, stuff should be reused. Apr 25 09:13:21 pohly: yes Apr 25 11:04:19 hello! I have a recipe overwrite from local layer, it overwrites curl, and when do_populate_sysroot it complains about a file conflict in the sysroot. What is the proper way to tell bitbake that file is provided by curl-mine instead upstream curl? Apr 25 11:06:12 zumbi: don't call it something else Apr 25 11:06:33 call it curl and use layer/recipe preference to pick yours instead of the oe-core one Apr 25 11:06:43 or, don't write a new recipe, just use a bbappend Apr 25 11:10:23 rburton: I have no control over the layers, it happens to be setup like that Apr 25 11:10:54 if the local layer doesn't call the recipe curl_[version].bb then it is Doing It Wrong Apr 25 11:11:21 if it is actually called eg curl-mine then it should RPROVIDES curl for compatibility Apr 25 11:11:36 I think it renames the package, because it uses different version Apr 25 11:11:58 selecting a different version is such a common operation it's directly supported in bb file naming Apr 25 11:14:36 Is it possible to add software like python, nano etc to already build yocto image ? Apr 25 11:15:22 hamdyaea: yes, tmp/deploy/ contains a package feed for use with dpkg/rpm/opkg/whatever you picked Apr 25 11:15:40 (if you actually have runtime package management to utilize that) Apr 25 11:19:47 rburton: Can you tell me more how to use it or do you know a web site where I can read informations about it ? Apr 25 11:20:20 the documentation? :) Apr 25 11:20:40 and by default your images can't use it anyways. Apr 25 11:21:30 rburton: Yes but I don't find any info about it Apr 25 11:21:34 the easy way for development purposes is to run a httpd of some sort on your build server sharing tmp/deploy/rpm (eg), set PACKAGE_FEED_URIS to the right url, and rebuild an image Apr 25 11:21:45 then you can use dnf etc to install stuff Apr 25 11:22:00 *assuming* you've enabled package-management in IMAGE_FEATURES Apr 25 11:22:55 rburton: So i just share the folder /tmp/deploy/rpm or deb and then I can access it and use dpkg etc.. ? Apr 25 11:23:03 yes Apr 25 11:23:10 PACKAGE_FEED_URIS will write the configuration for you Apr 25 11:23:19 I'm trying to switch from a init.d sytem to systemd. but when i boot the device its complaining that it cant find the init not syncing: No working init found. Apr 25 11:23:37 * LetoThe2nd hands rburton todays "runtime package managemant FTW support badge" Apr 25 11:24:05 rburton: Ok I will search about it Apr 25 11:27:28 MarcWe: always wise to wipe tmp when changing just in case. sstate will ensure it rebuilds in seconds. Apr 25 11:27:30 rburton: If I want to install a package who is not in tmp/deploy I can just download some files in it and then install it ? Apr 25 11:27:39 hamdyaea: bitbake it first then Apr 25 11:27:54 you can only install what is in deploy, so you'll need to build it Apr 25 11:30:57 rburton: OK thanks, one other question : can I use a share folder ? I mean I have a NAS in the network with a shared folder or I need to use only http protocole ? Apr 25 11:32:25 hamdyaea: sure, if you mount the shared drive on your target. Apr 25 11:33:05 (http is easier though as there's no mounting required) Apr 25 11:39:45 rburton: the target and share will be in the same network so it's ok. Thanks for the good infos Apr 25 11:40:26 hamdyaea: you probably want to look at http://www.yoctoproject.org/docs/latest/dev-manual/dev-manual.html#using-runtime-package-management Apr 25 11:45:21 kanavin: I will look at it Apr 25 11:58:40 hamdyaea: python -m SimpleHTTPServer works fine, if you are just serving packages locally for yourself. Apr 25 12:11:25 hello Apr 25 12:11:39 is it possible to know which layers are used for an image ? Apr 25 12:12:31 or should I look at the packages contained in the image and then check which layer provides them ? Apr 25 12:19:44 JoiF: OK I look at it Apr 25 12:24:56 top22: yes - I think the layer information is not preserved into the packages or the image Apr 25 12:35:50 Hello, is it possible to define a variable which i can share between different recipes? Apr 25 13:06:37 If my yocto image run with rpm packages, can I instell centos or any rpm public repository to install softwares with smart ? Apr 25 13:07:17 Or a fedora repository ? Apr 25 13:13:20 hamdyaea: no Apr 25 13:13:23 in a nutshell, no Apr 25 13:16:24 * LetoThe2nd still thinks that most people here totally overrate and misinterpret package management in the context of oe images. Apr 25 13:17:04 hamdyaea: that will work about as well as trying to install fedora packages on redhat, or suse packages on centos Apr 25 13:17:34 neverpanic: nope Apr 25 13:18:35 i am totally convinced that it will work worse. Apr 25 13:19:26 well, that already doesn't work very well, but I agree it will probably be even worse on a Yocto-built image. Apr 25 13:19:52 because in the context of desktop/server distributions distributions, you at least have *similar* APIs. something that just totally does *NOT* apply to custom distros like the things we are doing here :) Apr 25 13:19:56 I don't think it makes sense to distinguish different levels of 'broken' here, though. Apr 25 13:20:16 * mmh Apr 25 13:20:23 neverpanic: agreed there :) Apr 25 13:20:57 neverpanic: i am just extremely reluctant to give examples here where people already have preexisting assumptions of "worksforme" Apr 25 13:21:25 as that would only lead to even more runtim package management brokenness we have to deal with Apr 25 13:39:27 python do_something(var) { Apr 25 13:39:40 is it possible to define a python function with a parameter Apr 25 13:39:49 and call this function from an other function inside the recipe Apr 25 13:44:06 hello, i'm trying to build image-minimal on archlinux and I have a problem with openssl: /usr/bin/ld: libcrypto.a(sha1-x86_64.o): relocation R_X86_64_PC32 against undefined symbol `OPENSSL_ia32cap_P' can not be used when making a shared object; recompile with -fPIC Apr 25 13:44:35 but I I'm under the impression that -fPIC is passed to gcc nevertheless Apr 25 13:47:21 HUnk: sure, def myfunction(param): ... and then just call it from whatever python do_something task you have Apr 25 13:48:27 If my yocto image run with rpm packages, can I instell centos, Fedora, or any rpm public repository to install softwares with smart ? Apr 25 13:52:32 hamdyaea: no Apr 25 13:53:10 just like you can't install suse packages on fedora, and they both use rpm Apr 25 13:55:36 rburton: Can I find a list somewhere of all package possibilties in yocto ? Apr 25 13:55:53 hamdyaea: Did you miss the discussion above where we explained that this is not a good idea? Apr 25 13:56:21 neverpanic: Yes I missed it sorry Apr 25 13:56:28 hamdyaea: https://layers.openembedded.org/layerindex/branch/master/recipes/ Apr 25 13:56:29 hamdyaea: https://layers.openembedded.org/layerindex/branch/master/recipes/ has a list of a lot of recipes Apr 25 13:56:45 thanks Apr 25 14:32:13 seebs: any thoughts on http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/wip-rss2&id=3457c165e63efd09161397e0fdcfff0c7cfc0663 ? Apr 25 14:37:51 All the rpm packages in tmp/deploy are already installed in my image when I bitbake ?? Or they are options that I can install or no ? Apr 25 14:39:01 hamdyaea: those are all the packages that got generated alon the build process. and everything that is on your target is also there, but not everything that is there is also on the target. Apr 25 14:39:23 hmm. It's not insane, but it seems like it might be Too Aggressive. It's gonna close *every* client with no pending requests, and that seems like it might be too expensive. Apr 25 14:39:37 I was thinking of trying to identify the "oldest" client or something. Apr 25 14:42:37 LetoThe2nd: everything that is on your target is also there, but not everything that is there is also on the target. | I don't understand, it's there or no ? Apr 25 14:43:19 hamdyaea: your target contains a subset of the packages in tmp/deploy Apr 25 14:43:39 e.g. there will be debug symbols in tmp/deploy, but those will usually not be installed on your target Apr 25 14:44:32 hamdyaea: if something is on the target, it is also there. Apr 25 14:44:41 neverpanic: OK I see now, so there are all package already on my image and some others Apr 25 14:44:52 but the reverse is not true. Apr 25 14:44:54 hamdyaea: if you just do bitbake myimage then deploy will only contain the output from the *recipes* that went into the image. not every package, eg dbg symbols, will be in there. if you want to build more packages just bitbake those too, ie bitbake samba. Apr 25 14:45:02 won't go in your image, but will be built Apr 25 14:45:58 and can I create myself some rpm packages for yocto ? Apr 25 14:46:10 yes Apr 25 14:46:44 ok I see Apr 25 14:46:49 try to get away from the rpm mindset. Apr 25 14:47:05 and from "the yocto" mindset. Apr 25 14:47:32 to be correct: "you can create packages for your custom distribution". Apr 25 14:48:00 because its really something completely different. Apr 25 14:49:24 So If I want to add a new package, I can look in temp/Deploy if it's there I can add it with smart. If It's not there I can add it with layers and by modifying local.conf, If it not in thoses two places I have to create the package myself Apr 25 14:51:19 while that suffices for a first try and error approach, its not really correct again. Apr 25 14:51:56 because once you add a layer to the build it is best to consider everything you had before to be incompatible. Apr 25 14:52:10 RP: Thinking about it more, there is a small but non-zero chance of data loss in the case where a client wrote a message *between* the select and the close. Apr 25 14:52:24 If we process any messages, the loop can take a while to get to some clients, which means that the select data can be stale. Apr 25 14:52:30 seebs: I did test it and it seems to work FWIW in my test case where things previously hung Apr 25 14:52:36 hamdyaea: i really, really suggest to stick with adding everything during a monolithic build. add everything you want to IMAGE_INSTALL, then build. Apr 25 14:52:37 yeah. Apr 25 14:52:52 hamdyaea: so you can always see if your dependencies are right, and you can reproduce it. Apr 25 14:53:18 LetoThe2nd: Yes I see Apr 25 14:53:26 I would expect it to work most of the time. But that's the other reason I wanted to close relatively-few clients; I want to reduce risk. Apr 25 14:53:35 seebs: presumably any client sending data when the connection was lost would resend that data? Apr 25 14:54:06 seebs: I couldn't find anything to indicate client age/activity in the current data structures :/ Apr 25 14:54:30 seebs: you can probably guess I started getting hanging builds again ;-) Apr 25 14:55:50 if the connection goes down so that the write fails, it'll resend. Apr 25 14:55:58 if it was a FASTOP and the send completed, it won't. Apr 25 14:56:37 So there's not a huge window there, but there's a non-zero window. Apr 25 15:01:19 seebs: is there a way to stop new writes but read any data left in the fd? Apr 25 15:01:50 I don't *think* so. Apr 25 15:02:44 seebs: so the only real way to do this would be to send a "reconnect" event to clients? Apr 25 15:03:13 (without data loss) Apr 25 15:03:41 Well, we could turn off the "fastop" thing, but that would definitely affect performance. Although... I don't know how MUCH. Apr 25 15:04:05 seebs: I quite like fastop Apr 25 15:04:13 Hmm. We could attempt a non-blocking read or something... Apr 25 15:04:29 I do too, although it's probably got to be made smarter because it does lead to potential data loss on server crashes. Apr 25 15:04:30 seebs: that could still race Apr 25 15:04:55 yeah. Apr 25 15:06:41 seebs: I guess clients need to maintain a queue of all events since the last reply from the server and if the connection is lost without hearing from the server, replay them? Apr 25 15:07:56 ... probably. something like, anyway. hmm. Apr 25 15:08:13 i hate that a lot because it breaks what was probably the largest single performance improvement in pseudo. Apr 25 15:08:19 which was "stop doing allocation for every message" Apr 25 15:08:28 i'ma have to ponder this a bit. Apr 25 15:31:49 hello, has anybodyhave an issue building lzop with gcc 6.x ? Apr 25 15:32:15 version 1.03 of lzop seem to fail: checking whether your compiler passes the ACC conformance test... FAILED Apr 25 15:45:21 rburton: Figured few issues with libnsl and samba/autofs etc. changes in glibc 2.26 and fixed them have updated contrib branch kraj/master now with them. Apr 25 15:45:42 next cycle when its available please consider that branch Apr 25 15:46:08 will do Apr 25 15:46:13 khem: can you fix elfutils in meta-gpl2? Apr 25 15:46:36 a number of the test stages failed in my last gcc7 run, just fired a run with less other changes to see if its gcc7 or something else Apr 25 15:49:52 rburton: it will be a challange to keep fixing these ancient packages with new gcc Apr 25 15:50:08 I havent added meta-gplv2 to the testing mix yet Apr 25 15:50:22 will do that once the oe-core proper is clean with gcc7 Apr 25 15:50:26 khem: its just elfutils, from memory its just fallthrough comments Apr 25 15:50:35 rburton: ok thats easy to fix Apr 25 17:51:29 hi, I'm having an issue building fido due to a patch in poky recipes-core failing to apply: http://paste.debian.net/929288/ Apr 25 17:53:19 units/systemd-udevd.service.in isn't the same as whats in the patch... but I didn't have this issue a few weeks ago, and the offending patch was committed in 2016, so I'm not sure what else might have changed... **** ENDING LOGGING AT Wed Apr 26 03:00:00 2017