**** BEGIN LOGGING AT Sat Dec 29 02:59:58 2018 **** BEGIN LOGGING AT Sat Dec 29 19:18:43 2018 Dec 29 20:12:37 Hey all, I have enabled "CONFIG_MODULE_COMPRESS" in my Linux Kernel and included "kernel-modules" to get it to rootfs. But yocto still complains in package_qa. How can I include compressed kernel module into rootfs? **** ENDING LOGGING AT Sat Dec 29 20:17:59 2018 **** BEGIN LOGGING AT Tue Jan 01 01:50:10 2019 **** ENDING LOGGING AT Tue Jan 01 01:52:48 2019 **** BEGIN LOGGING AT Wed Jan 02 12:40:08 2019 Jan 02 13:08:04 New news from stackoverflow: Display webpages with QtWebEngine on embedded device - ERR_SSL_PROTOCOL_ERROR Jan 02 13:16:54 RP: do you know when ross would be back? Jan 02 13:17:16 kanavin: I am not sure Jan 02 13:17:28 kanavin: I'd assumed today but it would seem not Jan 02 13:18:20 RP: I have got virgl patchset to a working proof of concept by the way Jan 02 13:18:32 RP: I'll send it as RFC when people are mostly back from holidays Jan 02 13:18:55 kanavin: nice. That should be an interesting development, brings back something we use to do badly :) Jan 02 13:19:14 kanavin: it'll wipe out my build time gains from the toolchain cleanup ;-) Jan 02 13:19:20 RP: I have written down some thoughts here: https://bugzilla.yoctoproject.org/show_bug.cgi?id=12938#c1 Jan 02 13:19:20 Bug 12938: enhancement, Medium+, 2.7 M2, Martin.Jansa, NEW , virglrenderer+spice support in qemu-native with accelerated GL Jan 02 13:20:50 kanavin: hmm, some good questions there Jan 02 13:22:05 RP: yep, calling into host's /usr/lib/lib*GL* is probably less robust than writing to an X server socket Jan 02 13:22:23 RP: still, I think we are heading towards a reality where some kind of working GL is a 'must Jan 02 13:22:29 ' for writing UIs Jan 02 13:23:41 typically people build images and then transfer them to the NUC or even target boards for testing - using qemu would be less awkward Jan 02 13:24:53 kanavin: agreed. This is why we tried this a few years ago Jan 02 13:36:06 RP: I have fixed a few things in perl-sanity that were found last time, can you try this branch on the AB? http://git.yoctoproject.org/cgit.cgi/poky-contrib/log/?h=akanavin/perl-sanity Jan 02 13:41:16 kanavin: build triggered: https://autobuilder.yoctoproject.org/typhoon/#/builders/85/builds/78 Jan 02 13:49:30 RP: thanks, is that the 'quick' build? Jan 02 13:50:08 kanavin: yes Jan 02 13:50:26 kanavin: two builds running, one a -next and yours Jan 02 13:51:59 RP: I'd also like some package updates to get merged - ross had them in mut, and then nothing happened http://git.yoctoproject.org/cgit.cgi/poky-contrib/log/?h=akanavin/package-version-updates Jan 02 13:56:57 kanavin: sorry, they go lost in the holidays. Queued in -next now Jan 02 13:59:44 RP: thanks :) Jan 02 14:33:44 kanavin: few failures already :( Jan 02 14:34:35 kanavin: further than last time though I think? Jan 02 14:34:45 RP: yes, they are failures I haven't seen before :) Jan 02 14:39:28 kanavin: good and bad :) Jan 02 14:40:33 RP: is eclipse-plugin-oxygen failure unrelated? Jan 02 14:40:53 kanavin: yes, its failing everywhere. Waiting for people to return from vaction Jan 02 14:43:36 We must outlaw vacation Jan 02 14:43:45 at least the last two weeks of the year Jan 02 14:44:28 heh Jan 02 15:10:37 Crofton: +1 :) Jan 02 15:11:12 If everyone os going to take two weeks off together, lets do it when the weather is better! Jan 02 15:11:46 Crofton: its actually been quite good for mtbs here ;-) Jan 02 15:15:50 RP: this error seems specific to the build host - would you be able to share the WORKDIR of perl-native and dos2unix recipes? https://autobuilder.yoctoproject.org/typhoon/#/builders/64/builds/137 Jan 02 15:17:45 kanavin: you don't have ssh access to them? Jan 02 15:18:12 RP: I can try - I have access to auh machine, but not sure if I can access autobuilders Jan 02 15:18:51 kanavin: looks like not that one :( Jan 02 15:19:39 seems like I do not: Jan 02 15:19:41 alexander@alexander-box:~/development/poky/meta$ ssh akanavin@autobuilder.yoctoproject.org Jan 02 15:19:41 akanavin@autobuilder.yoctoproject.org: Permission denied (publickey). Jan 02 15:24:23 kanavin: https://autobuilder.yocto.io/pub/dos2unix.tgz Jan 02 15:24:28 kanavin: https://autobuilder.yocto.io/pub/perl-native.tgz Jan 02 15:24:39 RP: cheers Jan 02 15:26:22 RP: seems like shebang length exceeded issue Jan 02 15:26:29 it's 127 right? Jan 02 15:29:17 kanavin: yes, I think so Jan 02 15:29:28 kanavin: we did have to work around that in the past from what I recall Jan 02 15:54:10 RP: " The perl upgrade continues to be blocked on the fallout from meta-openembedded " - I have patches :) Jan 02 15:55:22 kanavin: I meant Jen's work - or did you fix that? Jan 02 15:56:09 RP: I didn't; our patches are actually independent. Jan 02 15:56:46 kanavin: right. I was kind of leaning to upgrading, then looking at the cross switch Jan 02 15:57:24 RP: depends on who gets his patchset to a working state first I guess :) Jan 02 15:57:51 kanavin: :) Jan 02 15:58:05 the idea of perl-cross is that there's a bit of pain right now, but future updates should be much easier Jan 02 15:58:16 if we stay with existing recipe, it will be lots of pain every time Jan 02 15:58:39 kanavin: I get it, one just seemed much closer than the other. That of course can change Jan 02 16:26:14 RP: what is the repo with autobuilder configurations? Jan 02 16:26:35 kanavin: yocto-autobuilder-helper ? Jan 02 16:27:18 RP: ah, config.json? I somehow assumed those were in separate files Jan 02 16:30:02 kanavin: yes Jan 02 16:30:16 kanavin: its simple enough not to need separate files any more :) Jan 02 16:38:47 New news from stackoverflow: Adding iptables to yocto causes image do_rootfs to fail due to wrong kernel version Jan 02 18:23:53 HNY to all Jan 02 18:25:05 RP: I have runit (another init system) to contribute, I was wondering if oe-core is right place to maintain it or should it have layer of its own or maybe part of meta-oe Jan 02 18:37:48 khem: same to you khem Jan 02 18:37:48 * armpit hmm, Honey to all.. khem you are sweet too Jan 02 18:40:44 In order to set some kernel CONFIG_xxx option, is there a bitbake variable one can set, or should one put them in a .cfg file and add it to SRC_URI ? Jan 02 18:49:58 kroon: the latter, assuming the krenel recipe supports fragments. not all do, so it depends on your MACHINE Jan 02 18:52:16 kergoth, ok, thanks Jan 02 18:53:29 linux-yocto supports them, and a few others on a case by case basis. you can assume it works for the former, for the latter you'll have toe xamine the kernel recipe Jan 02 19:10:04 still, it would be nice to be able to just write a .bbappend instead of .bbappend+.cfg in this case Jan 02 19:17:32 I'm sure you could roll your own metadata driven config file adjuster, but how exactly you inject it depends on the recipe Jan 02 19:18:36 if the recipe uses a file://defconfig, it'd be trivial to inject the changes in a do_patch postfunc or do_configure prefunc, but otehrwise not so much.. the way the config becomes .config varies with recipe Jan 02 19:20:48 yeah, I was thinking that specifically for linux-yocto* maybe it makes sense to add support for this Jan 02 20:13:37 is there an elegant way to avoid .netrc and still have passwords Jan 02 20:13:43 flying around Jan 02 20:15:19 khem: a keyring API, e.g. libsecret Jan 02 20:15:59 unfortunately not all tools in poky enable support for that, e.g. the subversion binary won't have it due to missing dependencies Jan 02 20:16:31 Plus you'd have to somehow get a DBus session bus into the build environment for that to work, which probably means fiddling with BB_ENV_EXTRAWHITE Jan 02 20:51:14 * derRichard is a amazed by the horrid complexity of kernel .config management in yocto ;-\ Jan 02 20:58:42 yeah, trying to securely pass passwords into the underlying fetch tools sounds non-trivial and fetcher-by-fetcher.. Jan 02 21:04:30 It's possible though, I've done it before with git, but gave up on SVN when I would have had to package more dependencies Jan 02 21:07:07 derRichard: its actually way better than it use to be :/ Jan 02 21:10:21 RP: i'm migrating from a different build system to yocto and the old .config approach was "copy xy-linux.config for machine XY from directory configs/ to kernel's .config". Jan 02 21:10:33 doing this in yocto seems to be more complicated that expected Jan 02 21:11:12 *than Jan 02 21:11:41 i thought KBUILD_DEFCONFIG_machine = ... is the way to go Jan 02 21:11:51 but it expects the kernel config in the kernel tree Jan 02 21:21:39 derRichard: file://defconfig in SRC_URI can do that iirc Jan 02 21:22:06 derRichard: then machineA/defconfig machineB/defconfig Jan 02 21:23:17 RP: aaah!! file://defconfig will search for a file named defconfig in ? Jan 02 21:24:00 search the docs for FILESPATH Jan 02 21:24:05 that's how file:// does its thing Jan 02 21:24:22 kergoth: ok, FILESPATH is what i need to understand. thx Jan 02 21:24:50 derRichard: what kergoth said Jan 02 21:25:02 yes, you can specify the defconfig that way. I tried it out last month and it works Jan 02 21:25:32 now the yocto-kernel-cache takes a bit of a learning curve :) Jan 02 21:28:25 RP: just sent out the (yocto-kernel-cache) config patches for qemuarma15 based system. Since the v4.19 kernel recipe will need the SHA from that to work, would you like me to wait for it to be accepted to send out the qemuarm update patches or send out now as an RFC (to get eyeballs on them earlier)? Jan 02 21:29:27 kergoth: hmm, i'm reading https://www.yoctoproject.org/docs/latest/ref-manual/ref-manual.html#var-FILESPATH Jan 02 21:29:35 but it does not mention the machine name Jan 02 21:29:50 or is the magic hidden in FILESEXTRAPATHS? Jan 02 21:30:01 you add it in the kernel recipe Jan 02 21:30:07 neverpanic: svn is not used so thats not going to be an issue Jan 02 21:30:11 at least, that is what I did Jan 02 21:30:43 jonmason: ok, this should work Jan 02 21:30:47 let me give it a try Jan 02 21:30:52 derRichard: hmm, that's rather sparse. no, it's the base_set_filespath function that adds the subdir based on OVERRIDES to the search path for each entry Jan 02 21:31:00 i'm surprised the docs are that sparse about it Jan 02 21:32:31 kergoth: TBH, i like yocto a lot but the docs are very sparse in general. they talk alot about obvious things but not about the neat tricks you can do and why/how they work. Jan 02 21:32:54 aehs29, you have any jtag/minized skills? Jan 02 21:33:33 derRichard: where we run into sparsesness of docs, could you file a bug please? Unfortunately its not obvious to us where this sparseness is :/ Jan 02 21:33:48 this particular case is a serious gap, though. OVERRIDES is key to the whole project and was critical in the original design of bitbake. the entries in overrides are used for individual vars throughout the metadata, config file inclusions (though manually), and pulling files/patches/config files Jan 02 21:33:51 RP: sure! Jan 02 21:33:58 yeah, a bug is a great idea Jan 02 21:34:45 i'm more or less back from vacation and have time ;) Jan 02 21:34:52 kergoth: we definitely need to fix it Jan 02 21:34:56 being able to define metadata in such a way that more specific information/configuration is used in preference to less specific is one of the original key concepts to how this all works Jan 02 21:35:25 kergoth: doesn't mean our tech writer knows that :/ Jan 02 21:35:41 i absolutely *suck* at acquiring beginner's mind or reading docs like a beginner, it's really really helpful for new users to submit bugs about docs so those of us with more experience know what isn't obvious Jan 02 21:36:03 kergoth, you are not alone :) Jan 02 21:36:08 indeed Jan 02 21:36:22 kergoth: happens all the time to me too. when you code something yourself everything is clear/easy :) Jan 02 21:37:04 derRichard: we've been doing this too long ;-) Jan 02 21:38:04 on the other hand, at least do have docs :-) Jan 02 21:38:09 you Jan 02 21:39:03 derRichard: we do try! :) Jan 02 21:39:25 Does OE do automatic detection of shebang-interpreter runtime dependencies for scripts in packages ? Jan 02 21:39:30 derRichard: and if you file the bug it should get improved, we do have a process for that! :) Jan 02 21:39:59 kroon: the rpmdeps code does, I'm not sure it filters to all package backends Jan 02 21:40:05 derRichard: an open source project that has a technical writer at all is novel to me :) Jan 02 21:40:06 RP: pushed that already to my todo list Jan 02 21:40:26 kergoth: lol, yes. Jan 02 21:40:32 project overemphasis on code contributions is one of my pet peeves Jan 02 21:40:36 kergoth: I remember the conversation where that came about... Jan 02 21:40:38 not an issue with this one thankfully Jan 02 21:40:47 * RP has fought for that a lot over the years Jan 02 21:41:03 it wasn't my idea but I did fight for it Jan 02 21:41:13 the single most useful thing the Yocto Project has done is pay a tech writer Jan 02 21:41:28 well user visible Jan 02 21:41:54 Crofton: layer model? there are a few others ;-) Jan 02 21:43:59 hmm, is the hddimg image type gone? i get "No IMAGE_CMD defined for IMAGE_FSTYPES hddimg...". and indeed, i don't find a IMAGE_CMD_hddimg definition Jan 02 21:44:16 RP, oh ok.. yeah, maybe the rpmdeps results isn't propagating to ipk packages then Jan 02 21:44:25 derRichard: https://www.yoctoproject.org/docs/latest/ref-manual/ref-manual.html#var-FILESOVERRIDES hints at it btw Jan 02 21:44:35 not that its at all obvious Jan 02 21:48:21 huh, pysh chokes on arithmetic expansion Jan 02 21:48:25 that's posix.. Jan 02 21:48:30 * kergoth rolls eyes Jan 02 21:50:35 derRichard: https://bugzilla.yoctoproject.org/show_bug.cgi?id=13112 - I filed it and added some comments, see if that helps Jan 02 21:50:36 Bug 13112: normal, Undecided, ---, srifenbark, NEW , FILESPATH and FILESOVERRIDES documentation lacks key detail Jan 02 21:50:45 kergoth: its an open bug Jan 02 21:50:52 https://drewdevault.com/2018/12/28/Anatomy-of-a-shell.html .. hmm. i wonder if mrsh is far enough along to write a quick python binding to libmrsh with cython and use that instead of pysh :) Jan 02 21:51:01 ah Jan 02 21:51:28 kergoth: https://bugzilla.yoctoproject.org/show_bug.cgi?id=11314 Jan 02 21:51:29 Bug 11314: enhancement, Medium, Future, unassigned, ACCEPTED , shell parser: support $(( )) Jan 02 21:52:01 ah, they're missing other things, though they do have arithmetic expansion — https://github.com/emersion/mrsh/issues/8 Jan 02 21:52:32 oh, no, the *shell* is missing bits, but the parser isn't Jan 02 21:52:37 all we'd need is the parser anyway Jan 02 21:52:45 course it'd be yet another dep.. Jan 02 21:52:59 kergoth: in some ways I like the devil we know :/ Jan 02 21:53:14 * kergoth nods Jan 02 21:53:47 has quirks but we know most of them :) Jan 02 21:57:48 RP: thx Jan 02 22:05:20 derRichard: Scott was going to ask me for a description anyway :) Jan 02 22:09:05 RP: what kind of bugs can i file to that bugzilla? Jan 02 22:15:51 derRichard: issues with bitbake/oe-core or the Yocto Project infrastructure around the project Jan 02 22:16:08 ok Jan 02 22:18:20 hmm, can't i set hostname_pn-base-files in my image recipe? the hostname is still the machine name. Jan 02 22:21:05 derRichard: image recipes can't influence other recipes. You'd have to do that at the config file level (local.conf, machine or distro config) Jan 02 22:21:40 derRichard: we don't build a new set of packages for each image, a package has to work for all images Jan 02 22:21:54 true, now it makes sense Jan 02 22:29:05 btw: meta-openembedded/meta-oe/recipes-extended/haveged/haveged_1.9.2.bb is broken, it inherits update-rc.d, but does not offer a init.d file. hence update-rc.d fails at the post install phase and bitbake unhappy. Jan 02 22:29:09 Postinstall scriptlets of ['haveged'] have failed. If the intention is to defer them to first boot, .... Jan 02 22:29:59 is this also a issue for bugzilla.yoctoproject.org? Jan 02 22:32:08 bluelightning, ping Jan 02 22:49:48 derRichard: not for meta-oe, that would be best reported on the mailing list **** ENDING LOGGING AT Thu Jan 03 02:59:57 2019