**** BEGIN LOGGING AT Fri Nov 30 03:00:00 2018 Nov 30 06:54:08 Some of you may have noticed that I have been looking at how time is handled. Nov 30 06:54:08 Some of our new devices have a hardware clock that increases monotonically, but is not writeable from software. Nov 30 06:54:08 That is, you cannot use it directly because that would leave you stuck in the early 1970s. Nov 30 06:54:08 The Android sub-system for these devices provide a time_daemon that reads a file ats_1 from persistent storage and sets system time according to that. Nov 30 06:54:09 ats_1 stores milliseconds since whatever /dev/rtc says so that if you add them together you get UTC now. Nov 30 06:54:12 If my tissot has a SIM in it *and* ats_1 exists and has a reasonably sensible value in it, then LuneOS picks up network time correctly. Otherwise it does not and you are stuck with system time extremely wrong and secure web browsing, for example, is not possible. Nov 30 06:54:18 I see no evidence that time_daemon ever writes ats_1, but on an Android system, evidently, something does. Nov 30 06:54:21 I haven't been able to find out what does that, nor when it happens. Nov 30 06:54:23 It occurs to me that this might not be a very webOS way of doing things anyway. Nov 30 06:54:24 Maybe we should persist and recover this time offset ourselves (on these devices only)? Nov 30 06:54:26 It would have to be picked up pretty early in the boot process so that the system clock is almost correct before anything more correct tries to set it. Nov 30 06:54:29 I would appreciate any design ideas that occur to you. Nov 30 07:30:53 Morning! Nov 30 07:31:30 JaMa: oooh I'm sorry I completely forgot to bump SRCREV Nov 30 07:40:21 ok my build is on good track now Nov 30 08:00:53 there is also: check_data_file_clashes: Package nyx-modules-hybris wants to install file /home/jenkins/workspace/luneos-testing/webos-ports/tmp-glibc/work/mido-webos-linux/luneos-dev-image/1.0-r0/rootfs/usr/lib/nyx/modules/nyxDeviceInfoMain.module Nov 30 08:00:57 But that file is already provided by package * nyx-modules Nov 30 08:01:38 JaMa: ah, that's more annoying, I didn't see it Nov 30 08:01:53 ah that's because I didn't include that recipe in the image yet, at the time Nov 30 08:01:54 and the fix for stat and bash provider seems to invalidate most of the sstate again Nov 30 08:02:20 http://jenkins.nas-admin.org/view/luneos-testing/job/luneos-testing_mido/17/console Nov 30 08:03:55 for nyx-modules-hybris, I'll have to dig a bit to know what will be the best solution here. We may want to exclude it again from the packagroupegroup until then Nov 30 08:04:45 cannot we disable the deviceinfo module in nyx-modules for MACHINEs which use it from nyx-modules-hybris? Nov 30 08:08:29 JaMa: yes, that's what I'd like to do, but it's partially in the CMakeLists.txt of nyx-modules Nov 30 08:12:45 or maybe there's a nice way through https://github.com/webOS-ports/meta-webos-ports/blob/pyro/meta-luneos/recipes-webos/nyx-modules/nyx-modules/tissot.cmake Nov 30 08:12:58 Morning Nov 30 08:13:26 JaMa: Thanks for fixes, didn't think about the SRCREV either, was low on coffee yesterday and tired as well :P Nov 30 08:17:53 no problem, just the jenkins build will take a bit longer :) Nov 30 08:18:25 first 5 MACHINEs are somewhat done Nov 30 08:18:45 disk space looks good Nov 30 08:18:46 /dev/xvdb1 493G 183G 288G 39% /home Nov 30 08:18:46 none 80G 22G 59G 28% /home/jenkins/workspace/luneos-testing/webos-ports/tmp-glibc Nov 30 08:24:34 JaMa: Well they're failing.... Nov 30 08:24:40 With various reasons Nov 30 09:00:52 and rebuilding qtwebengine every single time :/ Nov 30 09:01:19 but failing means more disk space used in tmpfs (because rm_work cannot run) and still it's not so bad Nov 30 09:05:54 Do we know what causes qtwebengine to rebuild for every target? Nov 30 09:13:28 Herrie|Laptop: qtwebengine has a lot of dependencies, unfortunately: qt*, pulseaudio, gstreamer, etc Nov 30 09:15:36 http://jenkins.nas-admin.org/view/luneos-testing/job/luneos-testing_workspace-compare-signatures/ shows a lot of signature issues Nov 30 09:15:54 one of them is for qtwebengine Nov 30 09:52:23 bitbake-diffsigs tmp-glibc/sstate-diff/1543570691/mido/aarch64-webos-linux/qtwebengine/5.11.2+gitAUTOINC+89afebb830_7085115f51-r0.do_prepare_recipe_sysroot.sigdata.fdb1b4e092db199d2f18c5b5c027d805 tmp-glibc/sstate-diff/1543570691/rosy/aarch64-webos-linux/qtwebengine/5.11.2+gitAUTOINC+89afebb830_7085115f51-r0.do_prepare_recipe_sysroot.sigdata.283a3461bf574dcd6b9044c9123f1555 Nov 30 09:52:29 runtaskdeps changed: Nov 30 09:52:32 luna-service2/luna-service2.bb.do_populate_sysroot with hash 33d65fea7c246c0b64aa3827f3b068ce Nov 30 09:52:35 changed to Nov 30 09:52:37 luna-service2/luna-service2.bb.do_populate_sysroot with hash 54c814f49bc3368b4e3039356d56428f Nov 30 09:52:40 Hash for dependent task luna-service2/luna-service2.bb.do_populate_sysroot changed from 33d65fea7c246c0b64aa3827f3b068ce to 54c814f49bc3368b4e3039356d56428f Nov 30 09:53:35 bitbake-diffsigs tmp-glibc/sstate-diff/1543570691/mido/mido-webos-linux/luna-service2/3.21.2-1+gitAUTOINC+17521d3923-r0.do_populate_sysroot.sigdata.33d65fea7c246c0b64aa3827f3b068ce tmp-glibc/sstate-diff/1543570691/rosy/rosy-webos-linux/luna-service2/3.21.2-1+gitAUTOINC+17521d3923-r0.do_populate_sysroot.sigdata.54c814f49bc3368b4e3039356d56428f Nov 30 09:53:39 basehash changed from 909052ce9152cbddec58baef5cec3fa1 to 7df37b6daf1c170463f132473d8690a0 Nov 30 09:53:42 Variable MACHINE value changed from 'mido' to 'rosy' Nov 30 09:53:45 Variable SSTATE_MANMACH value changed from 'mido' to 'rosy' Nov 30 09:55:03 inherit webos_machine_impl_dep was added to luna-service2 in "Initial work for webOS OSE migration" Nov 30 09:55:11 Herrie|Laptop: ^ Nov 30 09:56:11 be careful with OSE stuff Nov 30 09:56:46 I'm checking if this is the only reason for qtwebengine being MACHINE_ARCH Nov 30 10:00:19 pushing a removal in few mins Nov 30 10:00:31 expect another big rebuild Nov 30 10:02:26 FYI, I published a french version of our release press for Doppio on LinuxFr.org: https://linuxfr.org/news/luneos-doppio-est-sortie Nov 30 10:03:25 JaMa: Thnx Nov 30 10:03:34 I tested most of OSE stuff only for 1 or 2 targets locally Nov 30 10:03:40 So I didn't spot it there Nov 30 10:03:54 Herrie|Laptop: meta-luneui/recipes-webos/libpbnjson/libpbnjson.bb also lost PV and my fix with OECMAKE_GENERATOR Nov 30 10:05:23 These things show up when we do all machines build ;) Nov 30 10:05:36 JaMa: well my OSE work was from March from before you fixed the OECMAKE one ;) Nov 30 10:05:41 But yeah should be put back Nov 30 10:05:44 I can do it, or you? Nov 30 10:05:47 JaMa: not sure how we missed that one, there should have been a merge somewhere Nov 30 10:06:56 doing it now Nov 30 10:08:31 there are one more conflict in nyx-modules Nov 30 10:08:32 check_data_file_clashes: Package nyx-modules wants to install file /home/jenkins/workspace/luneos-testing/webos-ports/tmp-glibc/work/rosy-webos-linux/luneos-dev-image/1.0-r0/rootfs/usr/lib/nyx/modules/nyxSystemMain.module Nov 30 10:08:36 But that file is already provided by package * nyx-modules-hybris Nov 30 10:08:44 shown in http://jenkins.nas-admin.org/view/luneos-testing/job/luneos-testing_rosy/15/console Nov 30 10:08:54 awww sorry for that Nov 30 10:10:15 I'm still trying to figure out a good way to exclude them from nyx-modules Nov 30 10:10:51 will drop it from the image until it's fixed Nov 30 10:10:57 yes, thanks Nov 30 10:11:05 you can just revert the corresponding commit Nov 30 10:11:21 https://github.com/webOS-ports/meta-webos-ports/commit/aba542e9290eb612378bb706d82b7b981a927c18 Nov 30 10:11:45 already did, just pushing it now Nov 30 10:12:11 will restart the builds now Nov 30 10:14:57 qtwebengine seems to be aarch64 now Nov 30 10:18:23 Tofe: it is and was, but effectivelly MACHINE specific because of MACHINE_ARCH luna-service2 Nov 30 10:18:40 so each MACHINE rebuilt it separately even while producing aarch64 .ipk file Nov 30 10:20:59 oh, that's bad. Nov 30 10:24:39 yes OSE is really bad example, OSE has everything MACHINE_ARCH which sucks Nov 30 10:28:36 JaMa: They only have 1 machine, so they don't care :P Nov 30 10:28:45 Well 2 now ;) Nov 30 10:30:03 Herrie|Laptop: they have more internally, but still don't care (don't listen) Nov 30 10:30:19 JaMa: Hmmz that's not good Nov 30 10:31:22 I hate it since I was pushed to switch the default to MACHINE_ARCH (as supposedly temporary work around for design issues and signature issues in recipes) Nov 30 10:32:12 I was maintaining the changes to switch back to TUNE_PKGARCH and fix most of the recipe issues for few years, but they still didn't listen about the design issues, so I gave up and stopped maintaining even my fixes Nov 30 10:32:41 that's why we need to be careful about what we import from this stuff Nov 30 10:33:20 LuneOS is far superior in some aspects Nov 30 10:37:02 \o/ Nov 30 10:40:42 JaMa: OK Nov 30 10:40:53 Well most of these things should be relatively easy & minor to fix Nov 30 12:00:36 ok, my tissot is now built Nov 30 13:27:53 my qemu builds are doing qtwebengine with pyro and thud, I'll need another coffee or 2 before it's finished :) Nov 30 14:15:09 Herrie: db8 also lost PV it seems Nov 30 14:17:00 JaMa: Let me fix db8 Nov 30 14:17:03 There might be more Nov 30 14:20:30 I'm cross checking all now Nov 30 14:21:53 Ok found 8 altogether in the original PR where PV was dropped Nov 30 14:22:03 I'll double check to make sure which ones are back already and which not Nov 30 14:22:15 db8, luna-sysservice, nyx-modules, storaged, cmake-modules-webos-native, libpbnjson, pmloglib-private, pmloglib Nov 30 14:23:53 I have a fix for db8, but it fails to build with newer Yocto so I'm checking a bit more Nov 30 14:24:50 5http://paste.ubuntu.com/p/zzGdySS7Fr/ Nov 30 14:24:52 http://paste.ubuntu.com/p/zzGdySS7Fr/ Nov 30 14:25:13 WARNING: luneos-dev-image-1.0-r0 do_rootfs: The license listed Apache-2.0 was not in the licenses collected for recipe luna-service2 Nov 30 14:27:29 JaMa: newer gcc ? Nov 30 14:30:57 JaMa: also, it's not the latest commit in db8 repo... Nov 30 14:31:14 ah, oops, that's what you just said Nov 30 14:35:38 tissot build finished in 18minutes, much better Nov 30 14:36:21 Tofe: yes, newer everything but it wasn't failing +- 2 weeks ago and doesn't fail when I build the OSE version with Thud as well Nov 30 14:36:28 something strange is going on Nov 30 14:40:39 definitely... where are the leveldb includes coming from ? Nov 30 14:42:48 leveldb = option in bb I think Nov 30 14:43:53 I mean "recipe-sysroot/usr/include/leveldb/any_db.hpp" Nov 30 14:44:37 Tofe: I think leveldb gets included from https://github.com/webOS-ports/meta-webos-ports/blob/pyro/meta-luneos/recipes-webos/db8/db8.bb#L27 Nov 30 14:45:21 but I don't see in which repo the file actually is Nov 30 14:50:12 Tofe: I guess from https://github.com/openembedded/meta-openembedded/blob/pyro/meta-oe/recipes-extended/leveldb/leveldb_git.bb ? Nov 30 14:51:01 actually I found https://github.com/webOS-ports/meta-webos-ports/blob/3a29811836550a687031363c673f0362865b25d5/meta-luneos/recipes-upstreamable/leveldb-tl/leveldb-tl_git.bb Nov 30 14:51:48 Ah that might be more likely Nov 30 14:51:52 Though 1.06 is quite old Nov 30 14:52:25 2014, with a patch for gcc7... Nov 30 14:53:19 but OSE compiled fine 2 weeks ago, then it's unlikely to be the cause. Nov 30 14:56:03 I'm adding snappy, but that doesn't fix it as well Nov 30 14:59:13 I added the missing PV's Nov 30 15:01:09 snappy ? how is it related ? Nov 30 15:09:58 snappy is optional dependency of leveldb Nov 30 15:10:11 enabled by leveldb bbappend in OSE Nov 30 15:10:56 that's the snappy from meta-snappy, or complelely unrelated ? Nov 30 15:11:43 the recipe is in meta-webosose/meta-webos Nov 30 15:14:03 ooh I understand :) Nov 30 15:41:32 ok, I have a fix for db8 with new Yocto Nov 30 15:53:20 JaMa: Weird fix LOL Nov 30 15:56:22 That's unexpected yes :) Nov 30 16:00:51 Herrie|Laptop: Parsing recipes...ERROR: ParseError at /OE/build/wpb/webos-ports/meta-webos-ports/meta-luneui/recipes-webos/cmake-modules-webos-native/cmake-modules-webos-native.bb:17: unparsed line: 'PV = "1.6.3-1+git${SRCPV}' Nov 30 16:00:55 will fix Nov 30 16:00:56 elvispre: i had a look at time issue as well some time ago, but it seems my messages are lost as irc logger wasn't working during server's move. Nov 30 16:00:56 there is https://github.com/ubports/timekeeper in my notes, maybe we can use it as well Nov 30 16:01:26 nizovn: Ah I was looking for that one, checked mer, halium but didn't find it Nov 30 16:01:31 Seems it's at ubports :P Nov 30 16:01:51 :) Nov 30 16:01:51 JaMa: Need a new keyboard it seems :P Nov 30 16:06:35 this rebuild from sstate goes really well :) Nov 30 16:07:06 we're invalidating it faster than it could be reused by other MACHINEs :) Nov 30 16:08:00 cmake-modules-webos-native is used by nearly every webos components and the rest depend on one of those :) Nov 30 16:11:24 JaMa: Well this is temporary until we have something building Nov 30 16:16:20 JaMa: seems Mako is stuck on NOTE: recipe mediaindexer-0.1.0-14+gitAUTOINC+b1193b36d3-r0: task do_rm_work: Succeededfor quite some time Nov 30 16:33:01 JaMa: Seems it's now continuing and failing again on the ERROR: android-kernel-bootimg-1.0-r0 do_compile: Function failed: do_compile (log file is located at /home/jenkins/workspace/luneos-testing/webos-ports/tmp-glibc/work/mako-webos-linux-gnueabi/android-kernel-bootimg/1.0-r0/temp/log.do_compile.22060) Nov 30 16:33:13 "/home/jenkins/workspace/luneos-testing/webos-ports/tmp-glibc/deploy/images/mako/zImage: No such file or directory" Nov 30 16:48:24 Herrie|Laptop: it wasn't stuck it was building qtwebengine Nov 30 16:48:41 JaMa: Ah OK but wasn't showing this in the log at that time :P Nov 30 16:49:16 it shows do_compile started somewhere, than does some other stuff until it has nothing else to do and then it looks stuck Nov 30 16:50:35 missing /home/jenkins/workspace/luneos-testing/webos-ports/tmp-glibc/deploy/images/mako/zImage: is sign of missing virtual/kernel.do_deploy dependency in android-kernel-bootimg, I'll check Nov 30 16:51:31 do_install[depends] += "virtual/kernel:do_deploy" Nov 30 16:51:43 but it needs it in do_compile already Nov 30 16:53:50 strange the do_install is only in meta-smartphone thud, in pyro it's do_compile[depends] += "initramfs-android-image:do_image_complete virtual/kernel:do_deploy" which should be enough Nov 30 17:09:27 nizovn: Thanks for the timekeeper link. Nice and straightforward. Nov 30 17:22:27 elvispre: There's quite some useful things in ubports & mer that we could easily adapt in the future :) Nov 30 17:25:08 Herrie|Laptop: :) Nov 30 17:25:23 For stuff like FM radio etc Nov 30 17:25:40 We've been using quite some things in the past like voiceall, qt5-qpa-plugins etc Nov 30 17:25:46 BT bits as well Nov 30 23:18:41 * elvispre has built the latest tissot dev image and flashed it onto the device. Nov 30 23:19:45 It sticks on the boot screen, but it has actually booted and done most of the things. (Time is picked up from the phone network, for example.) Nov 30 23:20:10 There's just no display. **** ENDING LOGGING AT Sat Dec 01 03:00:00 2018