**** BEGIN LOGGING AT Mon Jun 11 03:00:03 2018 Jun 11 05:57:02 So I have built the webOS OSE image. Where should webos-ports host them? (Or I could host them on my own web site of course.) Jun 11 06:12:09 elvispre: I got access to put it, I'll check with ka6sox how he wants to proceed Jun 11 06:51:38 Morning! Jun 11 08:02:08 morning Jun 11 08:02:27 I think the plan is to build them on jenkins, not locally Jun 11 08:08:06 JaMa: I think so too Jun 11 08:08:10 But not sure Jun 11 08:14:30 ok, I'm quite sure :) Jun 11 08:15:53 doing the builds on public jenkins is much more transparent than uploading the images from somewhere (not that I don't trust elvis, but the general public doesn't know him) :) Jun 11 08:16:37 well people are downloading random ROMs from random forums and installing them on their phones.. so maybe they aren't concerned as well :) Jun 11 08:17:06 lets make some evil version of youtube app :) Jun 11 08:22:38 JaMa: I totally agree, download a binary from a random build done by a random guy is scary Jun 11 08:23:24 I've checked pyro+5.11 now in the emulator Jun 11 08:23:31 the main ui and firstuse works Jun 11 08:23:38 e.g. fingerterm works as well Jun 11 08:23:48 but many other applications don't start Jun 11 08:24:50 should I still merge it to pyro and we'll resolve the remaining issues from there? Jun 11 08:25:12 JaMa: Output from journalctl would help I guess Jun 11 08:25:13 basically just to be able to do the build with this on Jenkins for everybody to easily test Jun 11 08:25:30 Just to see if we see anything obvious there before merging Jun 11 08:26:11 ● luna-webappmanager.service loaded failed failed The Luna Next WebApp manager Jun 11 08:26:14 ● sntp.service loaded failed failed Simple Network Time Service Client Jun 11 08:27:05 https://pastebin.com/9mQ6qe0Q Jun 11 08:27:11 ^ luna-webappmanager.service log Jun 11 08:27:49 sntp just isn't configured Jun 11 08:27:50 Jun 11 10:18:46 qemux86 sntp[309]: kod_init_kod_db(): Cannot open KoD db file /var/db/ntp-kod: No such file or directory Jun 11 08:27:54 Jun 11 10:18:46 qemux86 sntp[309]: [[0;1;39mkod_init_kod_db(): Cannot open KoD db file /var/db/ntp-kod: No such file or directory[[0m Jun 11 08:27:57 Jun 11 10:18:51 qemux86 sntp[309]: ntpserver.example.org lookup error Name or service not known Jun 11 08:28:56 ERROR:zygote_host_impl_linux.cc(90)] Running as root without --no-sandbox is not supported. See https://crbug.com/638180. Jun 11 08:29:53 interesting, cannot see the crbug Jun 11 08:29:54 You do not have permission to view the requested page. Jun 11 08:29:55 Reason: User is not allowed to view this issue Jun 11 08:34:38 JaMa: sntp I've seen before already Jun 11 08:34:45 We should fix that sometimes Jun 11 08:35:51 elvispre / Herrie: thanks for the donations page tip, I can have a look at that Jun 11 08:37:08 running just luna-webappmanager.service under different user isn't as simple as adding User=foo to the Service, it needs matching owner of /tmp/luna-session Jun 11 08:37:45 and if I change the permissions then it fails as well Jun 11 08:37:46 Jun 11 08:35:40 qemux86 LunaWebAppManager[988]: DEBUG: 08:35:40.517: Adding extension "PalmSystem" Jun 11 08:37:49 Jun 11 08:35:40 qemux86 LunaWebAppManager[988]: DEBUG: 08:35:40.517: Adding extension "InAppBrowser" Jun 11 08:37:52 Jun 11 08:35:40 qemux86 LunaWebAppManager[988]: Received signal 11 SEGV_MAPERR 000000000000 Jun 11 08:38:11 will try to run at least luna-next under the same user Jun 11 08:39:47 JaMa: Isn't it a case of simply adding --no-sandbox? Jun 11 08:40:04 JaMa: if WAM doesn't start, then all the webapps don't work, so your current status is consistent, though bad Jun 11 08:41:04 It's pretty weird, I would have restricted exactly the opposite, i.e. running as root without any sandbox Jun 11 08:41:27 Seems we need to add it here https://github.com/webOS-ports/luna-webappmanager/blob/master/files/systemd/luna-webappmanager.service#L7 and here https://github.com/webOS-ports/luna-webappmanager/blob/master/src/main.cpp#L44 Jun 11 08:42:15 --no-sandbox isn't valid option for /usr/sbin/LunaWebAppManager Jun 11 08:42:59 https://pastebin.com/JB56uhwm Jun 11 08:43:22 ah that's the 2nd link from you :) Jun 11 08:43:35 JaMa: Because it would need to be added https://github.com/webOS-ports/luna-webappmanager/blob/master/src/main.cpp#L44 and also at https://github.com/webOS-ports/luna-webappmanager/blob/master/src/main.cpp#L37 Jun 11 08:43:39 Not sure that'll fix it but it might Jun 11 08:45:38 ok, will try that Jun 11 08:45:58 you can try to export QTWEBENGINE_CHROMIUM_FLAGS ="--no-sandbox" Jun 11 08:46:58 elvispre: FYI https://github.com/Halium/halium-devices/commit/a9f0188d05d2660d6b0ad4c983afefc9956bc37b Jun 11 08:55:03 nizovn: https://paste.ubuntu.com/p/GYcW8DXtdz/ Jun 11 08:57:32 JaMa: I guess it's some advanced sandboxing features our kernel doesn't support Jun 11 08:57:41 I remember Tofe was fighting with this before Jun 11 08:58:00 I.e. https://github.com/webOS-ports/luna-webappmanager/commit/30cd41c7f79b223833642710dca5730239523388 Jun 11 09:00:23 Seems we need to add the backports seccomp patch to our kernels and enable it on runtime: https://wiki.mozilla.org/Security/Sandbox/Seccomp Jun 11 09:00:32 But that shouldn't really apply to qemux86 Jun 11 09:00:35 Since it's on new kernel Jun 11 09:02:12 Though it could be it doesn't have the options enabled in defconfig Jun 11 09:02:23 do you have access to that crbug link? does it provide any useful info? Jun 11 09:02:58 CONFIG_HAVE_ARCH_SECCOMP_FILTER=y Jun 11 09:02:59 CONFIG_SECCOMP_FILTER=y Jun 11 09:02:59 CONFIG_SECCOMP=y Jun 11 09:03:29 JaMa: No access here either Jun 11 09:05:59 JaMa: maybe adding --enable-logging --log-level=0 as well will get more info Jun 11 09:40:03 JaMa: btw do our browser or mini-browser in firstuse render any web page? Jun 11 10:07:01 sorry I got side tracked and need to finish something @work first, but I don't think we render any web page Jun 11 10:07:17 I can upload the image to the builder if someone else wants to have a look at 5.11 in emulator Jun 11 10:08:02 np Jun 11 10:08:09 i can have a look at it Jun 11 10:10:07 if you want to reproduce the build locally, just use latest meta-qt5/master-next + cherry-pick 2 top commits from meta-webos-ports/master to pyro branch Jun 11 10:14:33 JaMa: Upload would be great Jun 11 10:22:35 ok, uploading to bonaire, then will move it to milla Jun 11 11:00:06 Herrie: nizovn: http://build.webos-ports.org/luneos-testing/qt-5.11/ Jun 11 11:03:33 Note that adding --no-sandbox will probably work, but should really only be a short-term solution Jun 11 11:04:19 Though, true enough, running as a user should work too - I guess Jun 11 11:28:18 JaMa: I guess that in order to see what makes qtwebengine crash we'll need a full debug view Jun 11 11:45:18 http://jenkins.nas-admin.org/view/webosose/ I'm working on initial support for jenkins-job.sh Jun 11 11:45:33 Tofe: debug build? Jun 11 11:48:20 JaMa: I think I'll need to do that to get a usable stack trace from qtwebengine for the crash Jun 11 12:05:25 Herrie: are your latest changes for OSE (for new LS2 model etc) pushed on GH ? Jun 11 12:15:51 Tofe: Yeah everything I did is in https://github.com/webOS-ports/meta-webos-ports/tree/herrie/webOS-OSE/ Jun 11 12:15:55 Nothing locally pending Jun 11 12:19:15 Because nyx-modules generates a dependency loop Jun 11 12:19:55 Note that I've rebased it on top of current pyro, maybe I made a mistake there Jun 11 12:41:12 Tofe: Hmmz I haven't seen that one locally I think Jun 11 12:44:33 ok, must be on my side then Jun 11 12:48:50 Tofe: Well I've got quite a frankenbuild Jun 11 12:48:55 So could be that I need to start afresh Jun 11 12:49:25 chrome issue seems to be related to opengl. i was able to get browser working until page load completes, then it crashes. https://bpaste.net/show/aece30494923 Jun 11 12:49:45 For the 5.11 build on arm, it seems there's a little regexp to update :) https://git.merproject.org/mer-core/qtscenegraph-adaptation/blob/master/customcontext/customcontext.pro#L6 Jun 11 13:01:07 Here we go... https://git.merproject.org/Tofe/qtscenegraph-adaptation/commit/1410d45704cf65224c9d8ec2d9bc7128e65b27d5 Jun 11 14:37:07 Herrie|Laptop: I wonder how the PalmService bridge looks like on LG's side Jun 11 14:37:48 Tofe: https://github.com/webosose/chromium53/tree/e71abe77551664c36f50433595fbc06e895219ee/src/injection/palmsystem Jun 11 14:42:26 oh, I missed the easiest source :p thanks Jun 11 14:44:13 But I guess to see the full extend you'd need to do a diff between Chromium 53 and LG's version Jun 11 14:44:24 Which will then also show all LG's customizations :P Jun 11 14:48:14 Herrie|Laptop: oh for the bridge it's easy: completely different approaches, similar functionalities Jun 11 14:54:25 But I think even LG's approach runs into changing API, won't it? Jun 11 15:01:30 You mean regarding Chromium's everchanging APIs ? Jun 11 15:02:25 Yes, they are even more impacted than us I think. However, they have people in their team that probably work fulltime with V8 APIs Jun 11 18:02:04 I just received my Redmi 5 ! Jun 11 18:02:21 \o/ Jun 11 18:02:26 but... Jun 11 18:02:46 Xiaomi wants me to wait 360 hours to unlock the bootloader /o\ Jun 11 18:03:05 Their process is just completely dumb Jun 11 18:03:47 Anyway. I've got work to do in parallel, beginning with this 5.11 thing Jun 11 18:04:53 heh 360 hours? that's really long Jun 11 18:08:53 yeah... Jun 11 18:09:02 JaMa: I get this on hammerhead: /usr/sbin/LunaWebAppManager: error while loading shared libraries: libQt5Core.so.5: cannot open shared object file: No such file or directory Jun 11 18:09:17 the file is there and seems correct Jun 11 18:09:45 that's strange Jun 11 18:09:49 Agreed Jun 11 18:09:51 was it a "clean" build? Jun 11 18:10:04 ...mostly? Jun 11 18:10:10 other binaries linked to libQt5Core.so.5 work ok? Jun 11 18:10:28 always based on pyro, but I switched back and forth for meta-qt5 Jun 11 18:10:30 I would try with LD_DEBUG=all Jun 11 18:11:25 it seems it's only for qt5core Jun 11 18:11:41 let me paste the output Jun 11 18:12:03 if anyone wants to try youtube app on rpi3, we have a build on our jenkins: http://build.webos-ports.org/webosose/ qemux86 will be next Jun 11 18:12:06 https://bpaste.net/show/d0a208a17cf2 Jun 11 18:12:42 does the libQt5Core.so.5.11.0 exist? Jun 11 18:12:58 yep, see beginning of the paste Jun 11 18:13:08 ah didn't notice -L Jun 11 18:13:32 I'll reflash it, just in case Jun 11 18:14:09 can you copy the file? Jun 11 18:14:19 just in case the fs is broken? Jun 11 18:15:32 yes, adb pulled it without any issue Jun 11 18:15:54 let me readelf it Jun 11 18:16:29 no error, all is output Jun 11 18:16:40 pretty strange. Jun 11 18:17:01 It's same for luna-next, so it comes from my libQt5Core binary somehow Jun 11 18:17:09 I'll reflash, who knows. Jun 11 18:18:14 Cool! Saves me a job - http://build.webos-ports.org/webosose/ Jun 11 18:19:37 How do you decide which source tar balls to include? (Guess: "I don't; that's all of them.") Jun 11 18:20:30 everything from downloads directory except SCM dirs Jun 11 18:20:39 see jenkins-job.sh changes from today :) Jun 11 18:20:56 now doing the qemux86 build Jun 11 18:21:51 I was not aware that there was a qemux86 target. ("There is now"?) Jun 11 18:21:54 nope, same, must come from my build Jun 11 18:22:20 elvispre: officially there isn't, but that doesn't stop me from fixing it with one small sed call :) Jun 11 18:22:29 :) Jun 11 18:22:59 elvispre: but you still need to build your own qemu with virglrenderer and sdl2 (with opengl enabled) support Jun 11 18:23:20 elvispre: only distributions like Gentoo allow you to do that easily Jun 11 18:23:53 JaMa: I'll stick to my RPi3 then. Jun 11 18:24:10 I've modified the build for Ubuntu as well, so if you want to build your own .deb packages for Ubuntu I can share the necessary changes somewhere Jun 11 18:25:46 JaMa: I'll concentrate on trying to get LuneOS working on my onyx, I think Jun 11 18:26:26 JaMa: but I will download the latest RPi3 image and confirm that it's good a bit later on. Jun 11 18:27:18 ok Jun 11 18:53:47 JaMa: how can I make so that I rebuild qtbase, without triggering the rebuild of all the rest ? Jun 11 18:54:22 I fear that if I do a clean(sstate), all my qt build will be lost Jun 11 18:55:07 if you do cleansstate without changing any metadata then you can rebuild qtbase without invalidating the rest of the build Jun 11 18:55:14 ok great Jun 11 18:55:16 will do Jun 11 18:58:01 hmm vboxvideo from upstream linux-yocto oopses even before getting to the console Jun 11 19:44:02 mmmh it didn't solve the issue; pretty strange Jun 11 20:22:35 JaMa: I've read in a forum: "You problem is indeed the old kernel. Qt 5.10 uses the renameat2 system call which is only available since kernel 3.15" https://bbs.archlinux.org/viewtopic.php?pid=1755257#p1755257 Jun 11 20:23:18 But I don't see how that would prevent library loading Jun 11 20:36:47 there is PACKAGECONFIG for renameat2 in qtbase (or was), let me check Jun 11 20:37:36 hmm there isn't https://patchwork.openembedded.org/patch/148219/ Jun 11 20:38:28 but at least you can pass -no-feature-renameat2 to see if it fixes the issue for you Jun 11 20:38:47 and -no-feature-getentropy while you're at it Jun 11 20:39:12 but this will invalidate all other qt* builds for you Jun 11 20:39:29 oh. Jun 11 20:39:44 well well, I'll prepare this and start the build tomorrow Jun 11 20:39:55 you can rebuild just qtbase and then update just the qtbase on the target Jun 11 20:40:00 but that's always a bit tricky Jun 11 20:40:10 that'll be fine :) Jun 11 20:40:27 if it works for you, I'll merge it to meta-qt5, but as disabled by default Jun 11 20:40:52 That would be nice; but let's see if it works first :) Jun 11 20:41:40 I just add it to QT_CONFIG_FLAGS right ? Jun 11 20:43:14 JaMa: The Rpi3 image from jenkins works just fine. Jun 11 20:46:34 Tofe: yes Jun 11 20:46:38 elvispre: thanks for test Jun 11 20:48:31 JaMa: ok good, let's see how this will go then Jun 11 21:16:43 Tofe: might be better for you to just upgrade meta-qt5 to latest jansa/master branch Jun 11 21:16:56 if it works I'll merge it as is and it won't trigger you another rebuild later **** ENDING LOGGING AT Tue Jun 12 03:00:04 2018