**** BEGIN LOGGING AT Tue Jul 04 03:00:01 2017 Jul 04 05:36:04 hello, does anyone have experience running Yocto builds on Ryzen machines? do you encounter issues (segfaults) like the users running parallel compilations on Gentoo? https://www.phoronix.com/scan.php?page=news_item&px=Ryzen-Compiler-Issues Jul 04 06:44:56 hello ! May I ask you a question ? I have an "useradd" recipe that also install "/home/myuser" directory. In one of my other recipes I have to install "home/myuser/userfolder". Running bitbake generates a conflict error between the two recipes 'cause both try to install /home/myuser. Which is the best way to fix this ? Jul 04 07:06:20 fberg: create a recipe that takes care of the user creation, then have both depend on it. Jul 04 07:10:02 ok ! thank you LetoThe2nd Jul 04 07:27:59 Leto, as you suggested me yesterday (in order to disable some modules to autoload at boot) I have created a new recipe in my meta-mylayer/recipes-kernel/linux/linux-raspberrypi_4.9.bbappend Jul 04 07:28:10 in these recipes I have added the following lines: Jul 04 07:28:29 # Include kernel configuration fragment Jul 04 07:28:29 FILESEXTRAPATHS_prepend := "${THISDIR}/files:" Jul 04 07:28:29 SRC_URL += " file://myconfig.cfg" Jul 04 07:29:18 then in the same folder of the recipe I have added the "files" folder with the myconfig.cfg file Jul 04 07:30:23 after that i run: bitbake -f -c kernel_configme linux-raspberrypi Jul 04 07:31:18 by the way if I examine the .config the changes of myconfig.cfg are not present Jul 04 07:31:55 I've tried to google it but I can't still ind a solution Jul 04 07:31:56 fberg: sorry, but regarding kernel configuration fragments i can't comment further than to look at http://www.yoctoproject.org/docs/current/kernel-dev/kernel-dev.html#changing-the-configuration Jul 04 07:32:08 as i don't use them, i favor monolithic defconfig files Jul 04 07:32:54 respectively, http://www.yoctoproject.org/docs/2.3/dev-manual/dev-manual.html#configuring-the-kernel Jul 04 07:33:05 Ok ! thank you Leto Jul 04 10:14:05 khem: awesome Jul 04 11:31:27 rburton: was there a plan WRT meson & rocko? Jul 04 11:33:07 anyway, there might be a libinput release without any autotools support within rocko timeframe Jul 04 11:33:25 heh Jul 04 11:33:30 so i was looking at meson again last week Jul 04 11:33:37 small issue is gtk-doc totally fails Jul 04 11:33:41 new 1.8 is the last with autotools and they release quite often Jul 04 11:33:43 fancy leaning meson :) Jul 04 11:33:48 erm, learning Jul 04 11:34:20 libinput might be doable as i doubt it has gtk-doc or g-i stuff Jul 04 11:34:24 my test was json-glib which has both Jul 04 11:34:28 right Jul 04 11:36:47 well that was annoying Jul 04 11:37:17 jku: feel welcome to look at meson in meta-oe and the patches in poky-contrib:ross/meson Jul 04 11:38:23 okay I might Jul 04 13:03:42 khem: ppc just passed. just ovmf in lsb builds left... Jul 04 13:09:29 rburton: is ovmf also due to hardening ? Jul 04 13:09:38 I was thinking it was due to glibc 2.26 Jul 04 13:12:01 no still failing Jul 04 13:12:29 'Bad definition for symbol ''@0 or unsupported symbol type. For example, absolute and undefined symbols are not supported.' Jul 04 13:14:25 is the pie stuff making new symbols that the custom tooling doesn't handle? Jul 04 13:14:38 yeah same error i see Jul 04 13:14:48 but I thought it was glibc 2.26 Jul 04 13:15:02 ok I will take a look at it Jul 04 13:15:16 i think it needs to pass more no-pie arguments to the gcc it uses to compile before GenFw Jul 04 13:15:20 ovmf build is quite wierd Jul 04 13:15:25 *very* Jul 04 13:15:44 right Jul 04 13:16:47 thanks khem Jul 04 13:18:32 that also means glibc 2.26 is green Jul 04 13:20:54 I wonder why ovmf is is in OE-core though its more appropriate in meta-intel layer Jul 04 13:22:48 khem: i honestly can't remember Jul 04 13:23:03 is it not usable on arm64 stuff too? Jul 04 13:24:18 arm builds dont build much of edk2 from oe-core, so in theory it might be in practice it is not Jul 04 13:33:56 hello guys ! I have a bbclass that has a function that do the following: echo "" >> ${IMAGE_SYSCTL_CONF} && echo "vm.min_free_kbytes = 8192" >> ${IMAGE_SYSCTL_CONF} Jul 04 13:34:35 since I have to add a line in ${IMAGE_SYSCTL_CONF} and ${IMAGE_SYSCTL_CONF} is defined in the bbclass, how can I add this lin within another recipe ? Jul 04 13:34:51 normally I would use a bbappend but I do not think it's that the case Jul 04 13:36:42 you can do it via same bbclass or via a port processing function for image Jul 04 13:37:09 above class is appending to it Jul 04 13:37:20 so order wont matter Jul 04 13:40:10 ty khem. what do you intend for a port processing function for image ? Jul 04 13:40:25 I'm really not an expert Jul 04 13:42:01 RP: it seems much better however it seems to have avoid parallel builds across configs. Intended? Jul 04 13:42:07 I prefer not touching the bbclass Jul 04 14:06:36 otavio: it sounds like those configs don't have matching parallel elements Jul 04 14:10:49 RP: I will remove it all here and start over to check Jul 04 14:11:41 RP: is there something I can do to help to finish this patch? Jul 04 14:16:24 otavio: also keep in mind "setscene" from sstate happens up front, I guess what you might be seeing is that if sstate artefacts aren't available, it will build them for reach build Jul 04 14:16:42 otavio: with the current design of bitbake its very hard to avoid that Jul 04 14:17:10 otavio: as I said yesterday, we probably need to rewrite it with new events for bitbake, version bumps and so on Jul 04 15:00:50 does anybody know why if in my recipe i add Jul 04 15:01:07 my_postprocess_function() { Jul 04 15:01:07 echo "hello" > ${IMAGE_ROOTFS}/hello.txt Jul 04 15:01:07 } Jul 04 15:01:07 ROOTFS_POSTPROCESS_COMMAND += "my_postprocess_function; " Jul 04 15:01:24 and the bitbake core-image-minimal -f -c do_rootfs Jul 04 15:01:33 the hello.txt is never generated Jul 04 15:05:40 fberg: which recipe are you adding that to? Jul 04 15:06:04 one of a custom layer Jul 04 15:06:17 so, not the core-image-minimal recipe then? Jul 04 15:06:25 nope blue Jul 04 15:06:34 that won't work I'm afraid Jul 04 15:06:44 I'm amazed Jul 04 15:06:47 :) Jul 04 15:07:00 you can only do something like that in the image recipe or at the global level (or a class inherited in either place) Jul 04 15:07:36 I have to add some configuration to /etc/sysctl.conf Jul 04 15:07:44 however, you can do postinstall functions in individual recipes, that's probably what you want Jul 04 15:07:53 but sysctl.conf is whiped aout by a bbclass Jul 04 15:08:02 pkg_postinst_ Jul 04 15:08:07 at ROOTFS_POSTPROCESS function Jul 04 15:08:54 so I am completely confused Jul 04 15:09:16 wiped out by which bbclass? Jul 04 15:09:22 thank you blue I'll give it a try Jul 04 15:09:46 from sdcard_image-rpi.bbclass Jul 04 15:10:12 http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#new-recipe-post-installation-scripts Jul 04 15:10:29 rpi_generate_sysctl_config() { Jul 04 15:10:30 # systemd sysctl config Jul 04 15:10:30 test -d ${IMAGE_ROOTFS}${sysconfdir}/sysctl.d && \ Jul 04 15:10:30 echo "vm.min_free_kbytes = 8192" > ${IMAGE_ROOTFS}${sysconfdir}/sysctl.d/rpi-vm.conf Jul 04 15:10:30 # sysv sysctl config Jul 04 15:10:30 IMAGE_SYSCTL_CONF="${IMAGE_ROOTFS}${sysconfdir}/sysctl.conf" Jul 04 15:10:30 test -e ${IMAGE_ROOTFS}${sysconfdir}/sysctl.conf && \ Jul 04 15:10:31 sed -e "/vm.min_free_kbytes/d" -i ${IMAGE_SYSCTL_CONF} Jul 04 15:10:31 echo "" >> ${IMAGE_SYSCTL_CONF} && echo "vm.min_free_kbytes = 8192" >> ${IMAGE_SYSCTL_CONF} Jul 04 15:10:32 } Jul 04 15:10:55 that's not wiping it out, that adds to it Jul 04 15:11:39 echo "vm.min_free_kbytes = 8192" > ${IMAGE_ROOTFS}${sysconfdir}/sysctl.d/rpi-vm.conf Jul 04 15:11:50 sysctl.d not sysctl.conf Jul 04 15:12:03 or rather, a file in sysctl.d Jul 04 15:12:03 the ">" does not clear the file before writing to it ? Jul 04 15:12:10 yes, it does Jul 04 15:12:14 it's not the same file ;) Jul 04 15:12:42 yes I know, but the meta-raspberry guys seems paricularly evil to me Jul 04 15:13:16 cause if i set a variable in sysctl.conf but not in rpi-vm.conf Jul 04 15:13:25 has no effetc on systcl Jul 04 15:13:35 so what you want is to change the vm.min_free_kbytes value? Jul 04 15:13:58 I have to simply add other two configurations for ipv4 Jul 04 15:14:14 eg net.ipv4.ip_nonlocal_bind=1 Jul 04 15:14:37 then why bother about this? you can either add your own sysctl.d file or modify sysctl.conf, neither of those will clash with the above code Jul 04 15:16:20 but what about adding: echo "net.ipv4.ip_nonlocal_bind=1" > ${D}${systemd_unitdir}/sysctl.d/rpi-vm.conf Jul 04 15:17:00 to the install section of my recipe ? Jul 04 15:17:38 will it take effects or rpi_generate_sysctl_config() will wipe it out during the do_rootfs task ? Jul 04 15:24:18 fberg: just use a different file in sysctl.d Jul 04 15:24:30 fberg: all files in that directory get read and acted upon Jul 04 15:24:51 yep Jul 04 15:25:08 giving it a try now Jul 04 15:25:13 ty blue Jul 04 15:52:35 rburton: root caused the ovmf issue and fixed, patch is on mailing list as well as pull branch Jul 04 15:52:46 yeah just saw, i'll refire that builder now Jul 04 15:53:34 ovmf build system is a hairball Jul 04 15:53:58 thanks for digging in Jul 04 15:54:17 cool. Jul 04 15:54:48 thats good. hopefully we will have no more surprises Jul 04 15:55:37 and once its merged in, i will send glibc 2.26rc pull Jul 04 15:55:57 it has got one run so I dont expect more surprises Jul 04 16:02:31 https://bugzilla.yoctoproject.org/show_bug.cgi?id=11751 makes interesting reading - xz compression in rpm saving space but increasing build time Jul 04 16:02:32 Bug 11751: normal, Undecided, ---, leonardo.sandoval.gonzalez, NEW , Performance regressions detected on core-image-sato and virtual/kernel at 2.4M1 rc1 Jul 04 16:54:50 Hi, my recipe is a Makefile based recipe with the main makefile which calls Makefile in subdirectories redefining CFLAGS on the fly. Apparently, it does not work. Could it work ? or Any Idea about where to look for some clue ? Jul 04 18:36:09 hey guys I'm having trouble compiling glib on yocto. It compiled just fine in ubuntu (using gcc) but on yocto (Nitrogen6x) it says it is missing gio/gio.h. Here's a pastebin of my makefile https://pastebin.com/0h5vH83P. I have confirmed that my distro has gio/gio.h Jul 04 18:45:11 There is no variable to setup sharing of prserv.sqlite3, right ? Jul 04 18:46:39 anyway I fucked up by removing cache/ without realizing it was in there... Jul 04 18:47:06 You can start it manually, or if you have a serialization point you can treat its prserv.sqlite3 as master copy and distribute it to other machines Jul 04 18:47:45 I guess the "normal" setup would be to start a PRserv separately somewhere on the network and make your bitbake instance connect to it via TCP Jul 04 19:40:58 khem: ovmf fixed :) Jul 04 19:51:36 halstead: any idea why a google search with site:bugzilla.yoctoproject.org fail does not return any results? Jul 04 19:54:29 lsandov: not sure why but we all Google not to index. https://bugzilla.yoctoproject.org/robots.txt Jul 04 19:54:41 Ask not all. Jul 04 19:58:14 halstead: I see. bugzilla has its own search engine so there is no a strong need, I just found strange that google did not know nothing about it. tks Jul 04 19:59:46 lsandov: if you'd like it changed open a big for it and I'll gather comments. I think it might be because sometimes confidential information gets into bugs and we want to hide it. Jul 04 20:02:48 halstead: I am not sure if there are strong reasons to open it. Let me first ping rburton Jul 04 20:02:59 halstead I think its better for project to have its issues searchable in www Jul 04 20:03:08 helps with adoption Jul 04 20:06:20 onoffon: good point Jul 04 20:07:02 i'd like to search bugzilla and error reporting service with google. problem is that google crawling errors.yp would kill it... Jul 04 20:07:07 Makes sense to me. Just need a bug for it and some time for comments before we make any changes Jul 04 20:07:29 rburton error.yp may wait Jul 04 20:08:14 rburton we can work on giving Google hints for indexing errors if that's desirable. Depends on which pages are useful to index Jul 04 20:08:44 the pages with the error logs on... indexing that right now would be a disaster Jul 04 20:08:53 need a rewrite with a proper full text indexer Jul 04 20:09:02 I don't think the logs themselves are worth indexing but maybe the top level pages. Jul 04 20:09:03 and selective cutting of the log Jul 04 20:09:19 halstead: literally last week i wanted to search the errors database for a specific error message Jul 04 20:09:41 errors.yp can do that, but it took about three minutes for the search to return a page :) Jul 04 20:09:46 Hrmmm. Would that have been in the logfile. Jul 04 20:10:05 We can look into upgrading that system too. Jul 04 20:11:44 test Jul 04 20:19:38 rburton: I've been keeping an updated u-boot recipe locally to test for regressions. It seems good ... do you want to put it under MUT to antecipate problems which can be merged upstream? it is on rc3 now Jul 04 20:20:02 yeah could do Jul 04 20:20:04 send it over Jul 04 20:56:47 neverpanic: yes, obviously the most generic is a tcp server - but it's not that practical when you're occasionally building offline (train trips, etc) Jul 04 20:57:37 in fact, given the distributed nature of the thing, prserv would greatly benefit from being git-backed Jul 04 20:58:08 no need for a permanent server, and reconciled with the rest of the world when need be Jul 04 21:34:21 can we easily seed a prserv from an sstate ? Forging a dump to import, I guess ? Jul 04 21:56:30 rburton: good; will send it tomorrow Jul 04 23:21:01 rburton error reports could be improved by just sending 10 lines context around large logs e.g. if webkit fails the log.do_compile itself it in tens of megabytes Jul 04 23:21:38 I think it would do a great thing for project if our sources/patches were indexed and appeared on top in search engines Jul 04 23:21:49 same goes for mailing list discussions Jul 04 23:21:51 and wiki **** ENDING LOGGING AT Wed Jul 05 03:00:03 2017