**** BEGIN LOGGING AT Thu Jul 12 03:00:01 2018 Jul 12 06:16:48 New news from stackoverflow: Which yocto is best for me Jul 12 07:44:12 good morning Jul 12 07:49:20 hi mckoan Jul 12 08:05:45 hi guys Jul 12 08:06:19 I have a problem with a build which in an older poky version used to work (2.1) Jul 12 08:06:31 in 2.5 I am facing a circular dependency Jul 12 08:06:48 can anyone have a look at the following scenario? Jul 12 08:07:17 recipe 1: Jul 12 08:07:22 addtask deploy after do_compile before do_build Jul 12 08:07:29 do_compile[depends] += "recipe2:do_configure" Jul 12 08:07:34 recipe2: Jul 12 08:07:41 do_compile[depends] += "recipe1:do_populate_sysroot" Jul 12 08:08:05 is there any way I can tweak this in poly 2.5 ? Jul 12 08:08:08 poky* Jul 12 08:35:59 Alexp: do you _need_ the before do_build condition? Jul 12 10:17:30 New news from stackoverflow: Enable secure boot in SAMA5d3_xplained [on hold] Jul 12 10:25:23 npm question: I am porting my nodejs/npm based bitbake recipes from morty to rocko; the package task fails as the install task creates symlink in image/usr/lib/node_modules that points back to the package directory. Did anyone work around this already ? Jul 12 10:30:02 hi cornel, I am not sure I understand your question; what is I need is recipe1 to gather some data from recipe2, then recipe2 will be compiled using the output from recipe1 sysroot Jul 12 10:32:05 so maybe yo don't need that condition Jul 12 10:36:57 addtask deploy after do_compile before do_build ? I could try to remove by even so, not sure if the circular dependecy issue won't occur Jul 12 10:42:29 Hi! Does somebody use a valgrind? Jul 12 10:42:42 I mean on Yocto Jul 12 10:42:52 should just work Jul 12 10:43:48 * astrunin[m] sent a long message: < https://matrix.org/_matrix/media/v1/download/matrix.org/IbascZcYpyPUaSATJDMhcQUz > Jul 12 10:44:39 i'd ask the valgrind channel about that Jul 12 10:45:00 maybe valgrind-on-mips is just not that useful Jul 12 10:45:55 I read changlogs and it looks like this version supports MIPS64 Jul 12 10:46:22 unfortunately, I can't find any Valgrind's channels Jul 12 10:46:37 to ask them Jul 12 10:47:54 may be the problem is about stripped ld-2.26.so? Jul 12 10:49:00 how can I disable stripping for ld? Jul 12 10:49:21 Another proof https://github.com/habitat-sh/core-plans/issues/850 Jul 12 10:50:40 I think it may help : INHIBIT_PACKAGE_STRIP = "1" Jul 12 11:03:45 oh, it wasn't my case Jul 12 11:04:38 I don't know what's wrong with valgrind, but it doesn't work for me Jul 12 11:13:42 'There is an IRC channel for Valgrind developers: #valgrind-dev at irc.freenode.net.' Jul 12 11:13:53 or irc at http://valgrind.org/support/mailing_lists.html Jul 12 11:14:34 erm, mailing list at http://valgrind.org/support/mailing_lists.html Jul 12 11:15:01 it says quite clearly invalid instruction, so its probably just that your code is using instructions that the valgrind interpretter doesn't know Jul 12 11:17:12 https://sourceware.org/git/?p=valgrind.git;a=blob;f=NEWS;h=2e13004e41563e3ecd855c94ee10af7378367068;hb=HEAD suggests that lots of mips stuff was added in git but isn't yet released Jul 12 12:03:38 How can I install 2 version of the same package with opkg ? I want to install some packages in a specific folder and when it's done, move the folder to the correct directory Jul 12 12:11:54 > it says quite clearly invalid instruction, so its probably just that your code is using instructions that the valgrind interpretter doesn't know Jul 12 12:11:54 nope. I use standard 'ls" linux utility to check valgrind (it is described in valgrind's readme). I don't use "my code" to check it. I will try to ask in their channel Jul 12 12:41:03 i getting an error for a strange reason Jul 12 12:41:28 when running bitbake -c menuconfig virtual/linux ... Jul 12 12:41:45 if i do nothing in the menu but exit, i get no error. Jul 12 12:42:16 if i drop down one level (e.g., drivers), then exit, i get no error Jul 12 12:43:11 if i drill down two levels, to driver/real time clock, BUT DO NOT CHANGE ANYTHING, just exit, then i get this: https://paste.fedoraproject.org/paste/StcrHoJrcbxrX~aTH4pO3A Jul 12 12:43:42 this sounds like a bug Jul 12 12:44:16 anyone heard of this problem and/or know a workaround/fix? Jul 12 12:56:19 don't know if it helps, but I always jump into the build directory of the kernel and call make ARCH=arm menuconfig directly...? Jul 12 13:02:16 svogl: i am not sure where that build directory exists and when i have access to it. Jul 12 13:02:32 svogl: are you talking about when using devtool? Jul 12 13:03:13 no, i normally don't use devtool Jul 12 13:03:19 the whole magic happens here: Jul 12 13:03:30 Storage/swupdate-test/sources/poky/build-hw-test-image/tmp/work/imx6ul_var_dart-fslc-linux-gnueabi/linux-variscite/4.1.15-r0/ Jul 12 13:03:47 (plus a leading slash) Jul 12 13:04:21 agreed it's there now, but when does that directory first get created? Jul 12 13:04:39 when you do a bitbake virtual/kernel Jul 12 13:04:54 it is created when fetching, under temp/ you find all scripts and log outputs Jul 12 13:05:04 it is deleted when the package has been built ok Jul 12 13:06:14 you might want to add your kernel to RM_WORK_EXCLUDE in local.conf Jul 12 13:06:25 if you want to keep the intermediates and play aroud Jul 12 13:06:30 if you want to keep the intermediates and play around Jul 12 13:06:50 that will prevent it from being removed when the build succeeds? Jul 12 13:06:59 xactly Jul 12 13:07:16 have all my devel projects in there Jul 12 13:10:10 ok, thanks for the pointers, svogl Jul 12 13:11:49 svogl: is it true that the defconfig file and .config file are the exact same formats? Jul 12 13:13:00 yes, you can copy the .config out after running your menuconfig Jul 12 13:13:21 exactly. thanks. Jul 12 13:18:03 New news from stackoverflow: Yocto Project SDK Plug-in does not install for Eclipse Oxygen Jul 12 13:18:34 is there a bitbake command to "clean" the working directories? Jul 12 13:20:55 or does one just rm -fr /tmp/work ? Jul 12 13:21:23 bitbake -c clean Jul 12 13:48:09 New news from stackoverflow: Error support.c no such file in libmpsse for Yocto Pocky Jul 12 13:58:43 Hi, did anyone have experienced Postgresql package ? i add it to my build but there is some bin program that i don't have on my built image ( createdb, psql...) Jul 12 14:00:13 vladzouth: use oe-pkgdata-util list-pkgs -p [recipe] to see what packages are built by the recipe. you probably need to add a tools package Jul 12 14:07:16 rburton: i don't see what is exactly a tools package? is there a functionnality of bitbake because when i run the command, i see nothing on the output? Jul 12 14:12:56 vladzouth: check the postgres recipe, see what packages it generates. Jul 12 14:16:36 zibri: When i check my package image on tmp dir, i see that missing package on it. But these are not install in the rootfs. How to explain that? Jul 12 14:18:15 New news from stackoverflow: Yocto Project SDK Plug-in does not install for Eclipse Oxygen Jul 12 14:32:11 vladzouth: check the packages-split directory Jul 12 14:33:38 all directory under ${WORKDIR}/packages-split will be packaged as separate packages, so if the files you want are in one of them, make sure that package is also installed to your image Jul 12 14:34:19 (or.. just check the recipe, where all this is specified) Jul 12 14:43:05 i am trying to get yocto to use my own defconfig file and failing. here's what i've done: ... Jul 12 14:44:08 ...this is all under sources/meta-variscite-fslc/recipes-kernel/linux ... Jul 12 14:44:32 1) created a files folder and added my_defconfig (the config file which i want to be used) Jul 12 14:45:39 2) added a linux-variscite_4.1.15.bbappend: https://paste.fedoraproject.org/paste/ZRm-mCBLlRRBjyVWI-GI~Q Jul 12 14:47:24 yet i've verified that the config that is used by "bitbake -c manuconfig virtual/linux" is not the one in my files folder Jul 12 14:47:46 so where is "bitbake -c manuconfig virtual/linux" getting the defconfig from?!?!? Jul 12 14:49:49 svogl: do you have any idea on this? Jul 12 14:50:26 @yates - as the package name says - it is virtual Jul 12 14:50:48 @yates - I guess your kernel is called linux-variscite Jul 12 14:51:28 svogl: i thought, via the magic of yocto, that virtual/linux is translated to linux-variscite. not true?!?!? Jul 12 14:53:05 regardless, bitbake -c menuonfig linux-variscite yields the same result - my_defconfig is not being utilized. (i just tried it) Jul 12 14:53:11 menuonfig Jul 12 14:53:21 menuconfig! sheesh.. Jul 12 14:53:47 hm, I always use the concrete package names Jul 12 14:54:07 n.b. that the bb files can be expanded by bbappend files Jul 12 14:54:14 propably there's one for your board variant Jul 12 14:54:23 with the defconfig youre looking for Jul 12 14:54:24 svogl: yes, did you see my fpaste above? Jul 12 14:54:49 @yates saw that, yes - guessed your board from that Jul 12 14:55:48 i've lost you. linux-variscite is our board, yes. what more are you trying to say? Jul 12 14:56:36 linux-variscite_4.1.15.bb is the main recipe in the directory i mentioned above Jul 12 14:57:07 @yates your defconfig should be in the subdir where the bb file is Jul 12 14:57:48 hey uh, I'm running into multilib errors with python on rocko 2.4.x, are those known ? Jul 12 14:57:51 Running transaction test Jul 12 14:57:54 Error: Transaction check error: Jul 12 14:57:54 this kind of stuff Jul 12 14:57:57 file /usr/bin/python3.5m-config from install of lib32-python3-core-3.5.3-r1.0.armv7ahf_neon conflicts with file from package python3-core-3.5.3-r1.0.aarch64 Jul 12 14:57:59 file /usr/bin/g-ir-annotation-tool conflicts between attempted installs of lib32-gobject-introspection-1.52.1-r0.armv7ahf_neon and gobject-introspection-1.52.1-r0.aarch64 Jul 12 14:58:01 @yates BUT, there could be a linux-variscite_4.%.bbappend file with modifications, if there are more board variants ( we have that here with our toradex modules) Jul 12 14:58:30 Crofton|work: zeddiii ^ Jul 12 15:00:38 if a .bb file contains a function xyz(), and a .bbappend file also contains that function, does the .bbappend version override the .bb version? Jul 12 15:01:39 Marex: i thought we'd put python-config into the dev package for that reason Jul 12 15:01:45 yates: yes Jul 12 15:02:05 Marex: re g-i, i guess we need to split the package up into tools and library Jul 12 15:02:57 rburton_: I think I'm installing the dev packages too Jul 12 15:03:18 +IMAGE_FEATURES += "dev-pkgs tools-sdk tools-debug tools-profile debug-tweaks ssh-server-openssh" Jul 12 15:03:21 yep Jul 12 15:03:38 Marex: sure but those files are kind of expected to conflict Jul 12 15:04:10 rburton_: so I should try building without the debug / dev stuff and see what breaks ? Jul 12 15:04:31 Marex: well i'd say the py one is a bug and i thought we fixed that already, and g-i needs repackaging a bit Jul 12 15:04:42 RP: had to drop from triage, need to go to dentist Jul 12 15:07:03 rburton_: are you sure it's fixed in rocko ? Jul 12 15:07:16 rburton_: maybe it's fixed in master only Jul 12 15:15:35 zibri: In the recipe file: http://cgit.openembedded.org/meta-openembedded/tree/meta-oe/recipes-support/postgresql/postgresql.inc?h=krogoth , there is a variable FILES_${PN}-client which includes the program file. Do you know how to enable it? Jul 12 15:27:18 in this bbappend, where is the source file for the "cp" command in do_preconfigure_prepend() going to come from? https://paste.fedoraproject.org/paste/ZRm-mCBLlRRBjyVWI-GI~Q **** BEGIN LOGGING AT Thu Jul 12 15:33:40 2018 Jul 12 15:34:54 sjolley: bugzilla is still down, will / are you investigating that? Jul 12 15:36:19 vmeson: its back Jul 12 15:36:41 halstead: any idea why bugzilla disappeared just now for a few minutes? Jul 12 15:37:38 a netfart. Jul 12 15:37:43 RP: thx, I waited at least 3 minutes... maybe it's slow intel server booting! :) Jul 12 15:37:44 RP: there are a few jobs that might cause that but they should not be running right now. Jul 12 15:39:20 halstead: we were loading it a bit during triage but it did seem to disappear for a while... Jul 12 15:45:48 RP, Were you seeing timeout messages in the browser or getting blank white pages? Jul 12 15:46:31 halstead: timeout connection issues I think, I didn't leave the window open for more than 30s Jul 12 15:46:37 vmeson: ^^^ ? Jul 12 15:47:22 according to the manual, WORKDIR would be somewhere over in tmp/work/..., so i would expect the file in x/files/my_defconfig to not be copied. in fact i would expect an error, but i'm not getitng an error Jul 12 15:47:26 Or just super slow to load pages? Jul 12 15:47:40 I'm don Jul 12 15:47:48 it's a net attack from china! Jul 12 15:48:06 't see any load issues. I need to check on the network. Jul 12 15:50:10 halstead: fair enough, just thought I'd mention since multiple people saw it Jul 12 15:50:13 halstead: I got a blank browser tab, I retried serveral times over 3+ minutes. Jul 12 15:51:13 vmeson, That indicates an issue with the reverse proxy or the perl workers. Thanks! Jul 12 15:52:21 halstead: k, thanks for keeping YP BZ going with such a high uptime. We only noticed because this happened during the bug review meeting. Jul 12 17:35:55 Has anyone had luck building a bootloader for a Picozed 7030 and FMCv2 carrier? The only working image I can find comes from a pre-built binary (WindRiver Pulsar). Jul 12 17:47:06 tgoodwin: try mainline U-Boot, you'll need to supply ps7_init files though Jul 12 17:50:44 marex-cloud: been down that route via meta-xilinx's rocko branch and replaced the ps7_init files with ones from the Vivado SDK (just in case...). The resulting build just has flashing PS LEDs on the FMC carrier, no console output at all. Jul 12 17:51:30 I mean mainline U-Boot, not the garbage Xilinx provides Jul 12 17:51:33 marex-cloud: I then tried switching to rel-v2018.2 to see if I could get any farther. After correcting a typo in the machine definition, it's the same result. Jul 12 17:52:04 marex-cloud: Right, the preferred provider for that machine (picozed-zynq7) is u-boot (vs u-boot-xlnx). Jul 12 17:52:29 v2018.07 or u-boot/master? Jul 12 17:52:56 Rocko, so 2017.09. Jul 12 17:53:09 Apparently the meta-xilinx started patching standard recipes in v2018.2 and forcing wrong package versions, what a crap Jul 12 17:53:13 I saw that yesterday Jul 12 17:53:46 2017.09 is ancient, try master or 2018.07 Jul 12 17:54:06 Yeah, I noticed over in meta-xilinx-tools they torched their xilinx-bootbin class but still reference it in the readme (vs. 2017.4) Jul 12 17:54:56 U-Boot can generate the boot.bin with mkimage ... Jul 12 18:10:22 marex-cloud: right. I'm fairly confident they're using the one generated from the u-boot process. I'm letting a clean build start right now just in case all my hacking at this threw something off. Jul 12 18:11:39 tgoodwin: they use fsbl Jul 12 18:11:56 tgoodwin: I'd just build U-Boot by hand to avoid the overhead Jul 12 19:07:30 marex-cloud: I suppose what I mean is that the way the recipe works is to take the boot.in from u-boot. Unless I'm misreading something, it's SPL+FSBL. Jul 12 19:14:16 tgoodwin: you don't need fsbl with mainline U-Boot Jul 12 19:16:54 I pulled that comment from their comments. Jul 12 19:17:24 (from the xilinx layer, even though the machine def'n is pointed back at u-boot) Jul 12 19:19:26 Anyway, comment still stands though. Per the build artifacts, it's using mainline u-boot's boot.bin and spl. Jul 12 20:10:27 marex-cloud: Thanks for your help. Unfortunately I'm still stuck even with 2018.07 by hand. I added my ps7_init files to board/xilinx/zynq/zynq-picozed, ran 'make zynq_picozed_defconfig && make'. That's with 2018.07 by hand. The resulting boot-up is similar: no console output, flashing PS LEDs on the FMC carrier. Jul 12 20:11:40 Huh, apparently Guido stepped down from being the Python BDFL Jul 12 20:19:03 JPEW: I can understand that email... Jul 12 20:19:34 RP: Don't go getting any ideas ;) Jul 12 20:21:02 On my hash equivalence server changes, I have some changes for bitbake mixed in with oe-core changes... they can't really be separated at this point, so should I push the patches to both lists? Not sure on the edicate there. Jul 12 20:21:29 I was only going to push them as RFCs.... I think I've getting to the point where I need some feedback Jul 12 20:21:45 JPEW: worst case you change bitbake and then change the min version reqs in OE but there are usually ways to avoid this Jul 12 20:22:04 JPEW: For an RFC you could just acknowledge there is an issue there Jul 12 20:22:51 RP: Ya that was my plan.... which mailing list would you suggest? Jul 12 20:24:02 The bitbake changes are more invasive, but the oe-core changes are more of the actual "make this thing work" substance. Jul 12 20:25:18 JPEW: oe-core and bitbake? Jul 12 20:25:28 Sure, I can do that. Jul 12 20:28:43 i followed the instructions for customizing the kernel here: https://www.yoctoproject.org/docs/2.6/kernel-dev/kernel-dev.html#changing-the-configuration Jul 12 20:29:19 then i did a bitbake virtual/kernel -c menuconfig, but the customization was not in the configuration. Jul 12 20:30:04 -c menuconfig should apply those customatizations before running the ncurses menuconfig thingie, right? Jul 12 20:30:48 i had previously bitbaked the kernel successfully Jul 12 20:32:07 tgoodwin: talk to monstr at #u-boot tomorrow (GMT) Jul 12 20:32:15 tgoodwin: he should be able to help you with that Jul 12 20:33:40 also note that the .bbappend file and customization .cfg file are in another layer (our board's layer). that should still be picked up when building the linux-variscite recipe if it's in my build/confi/bblayers.conf, right? Jul 12 20:52:04 i tested the generated kernel and my intended customization is indeed not present Jul 12 20:52:12 zgrep CONFIG_RTC_DRV_ISL1208 /proc/config.gz Jul 12 21:30:17 i followed the instructions for customizing the kernel here and it's just not working: https://www.yoctoproject.org/docs/2.6/kernel-dev/kernel-dev.html#changing-the-configuration Jul 12 21:31:37 if, after making the .bbppend and .cfg files as explained above, i bitbake linux-variscite, then look at the sources/poky/build-hw-test-image/tmp/work/imx6ul_var_dart-fslc-linux-gnueabi/linux-variscite/4.1.15-r0/build/.config file, the configuration is not changed Jul 12 21:32:07 i've also rebuilt the entire kernel and tested on the actual device and found it actually hasn't changed Jul 12 21:32:23 could someone give me a clue please? Jul 12 21:33:07 was there a known problem with this on morty? Jul 12 21:39:32 not all kernel recipes support configuration fragments, so verify that first Jul 12 21:43:05 kergoth: how do you check? Jul 12 21:45:10 if it inherits kernel-yocto it supports them, other than that it's only specific kernel recipes. check for a dependency on kern-tools-native in the recipe Jul 12 22:07:48 those keywords would be in either a .bb or a .inc? if so, then nope, my kernel recipe doesn't have them Jul 12 22:08:22 there is a ./poky/meta/recipes-kernel/linux/linux-yocto.inc:55:inherit kernel-yocto Jul 12 22:13:10 if the kernel recipe either requires linux-yocto or inherits kernel-yocto, or manually implements fragment support with kern-tools-native, then a fragment will work, otherwise you'll have to either override the defconfig entirely or implement the fragment supports Jul 12 22:13:12 s/s$/ Jul 12 22:13:38 most vendor kernels just give you a defconfig and expect you to use it directly or write your own from scratch Jul 12 22:13:46 there's an open issue in bugzilla about handling it consistently for all kernels, but i don't know if that's made any progress Jul 12 22:13:54 which is a pain and why the fragment stuff exists, but try convincing the vendors that Jul 12 22:14:10 we've bbappended bsp kernel recipes to add fragments handling repeatedly, but it's not always easy, depending on where and when they create .config from the defcnfig Jul 12 22:14:15 (we being mentor) Jul 12 22:18:42 couldn't i just put a "REQUIRE += "kern-tools-native"" in my .bbappend??? Jul 12 22:20:04 yeah, i like the idea of the fragment implementation Jul 12 22:20:28 no. depending on it isn't enough Jul 12 22:20:37 also there's no such thing as REQUIRE Jul 12 22:20:54 huh? Jul 12 22:20:59 but even if it DEPEDNS on it, it has to run the tool kern-tools-native provides to merge the fragments into .config after .lconfig is created but before it's used Jul 12 22:21:00 https://www.yoctoproject.org/docs/1.6/bitbake-user-manual/bitbake-user-manual.html#require-inclusion Jul 12 22:21:10 that's how you include a .bb or .inc Jul 12 22:21:15 that's not how you depend on another recipe Jul 12 22:21:26 its 'require' or 'include' for the former, DEPENDS for the latter Jul 12 22:21:34 both of which are documented at that link, read it more Jul 12 22:59:24 so it sounds like this isn't feasible without doing several modifications Jul 12 22:59:44 i'll just find a way to jam my own defconfig in. **** ENDING LOGGING AT Fri Jul 13 03:00:01 2018