**** BEGIN LOGGING AT Wed May 27 03:04:13 2020 May 27 05:26:09 Morning May 27 07:15:04 Morning! May 27 08:23:26 Herrie: do you know where does all the "-*" suffixes come from ? https://paste.ubuntu.com/p/B8Bm89563P/ May 27 08:23:46 I guess it's automatically generated from a .bbclass May 27 08:24:03 Tofe: Yes May 27 08:24:14 eta-luneos/classes/webos_app_generate_security_files.bbclass: key = app_id + "-*" May 27 08:24:19 +m May 27 08:24:43 That's why we have "LSHUB_ROLE_FILE {} Can not find service "org.webosports.app.firstuse-*" permissions for executable "/usr/sbin/luna-qml-launcher" " May 27 08:25:02 Basically: https://github.com/webOS-ports/meta-webos-ports/blob/herrie/zeus-acg-qt5.15-large/meta-luneos/classes/webos_app_generate_security_files.bbclass#L18 May 27 08:25:22 This is what comes from OSE May 27 08:25:59 I tweaked it for com.palm to not include the -* just * as per https://github.com/webOS-ports/meta-webos-ports/blob/herrie/zeus-acg-qt5.15-large/meta-luneos/classes/webos_app_generate_security_files.bbclass#L96 May 27 08:26:03 should we add "qml" type exception too ? May 27 08:26:10 I guess we need to extend that for org.webosports as well May 27 08:26:25 I had similar issues with com.palm apps May 27 08:26:38 I'd like to understand why they add the suffix at all May 27 08:26:39 Workaround was to add a variant that doesn't have "-*" just "*" May 27 08:26:55 I guess something in their WAM somehow, not sure May 27 08:27:30 aaah, I think I can guess, they generate the appId in WAM like "id" + "-" + "pid" May 27 08:27:43 legacy used to use a space there May 27 08:27:43 Probably something like that May 27 08:28:00 Well space is not very nice in general, so I can understand - instad May 27 08:28:06 I agree May 27 08:28:14 Also, it means we could go the same May 27 08:28:21 and keep the bbclass as-is May 27 08:28:26 Yes ideally May 27 08:28:28 s/go/do/ May 27 08:28:32 I just did this to get it to work May 27 08:28:38 I wasn't sure where this originated from May 27 08:28:40 also, for the qml, we might be able to do the same May 27 08:28:48 And I saw this works ;) May 27 08:28:54 I'm all for a proper fix of course May 27 08:29:06 I'll try a proper fix then May 27 08:29:07 Sounds like it's a 1 line fix May 27 08:29:13 Or even less ;) May 27 08:29:56 It shouldn't be too hard, finding the right place to do the change will take the longest May 27 08:43:19 I'm not sure we also want that for qml all the time; it should just be for apps, a priori May 27 08:44:35 I think I did it for a reason, but happy to be proven wrong ;) May 27 08:45:00 well, for webapps we did use simply appId, not appId-pid May 27 08:45:17 so com.palm.app.email wouldn't appear right on ls2 May 27 08:45:30 I've already a fix to try to migrate these webapps May 27 08:45:36 now the question is for qml May 27 08:45:41 It probably gave some errors in /var/log/messages or journalctl for QML apps, so I inherited webos_apps so it generates the required permissions May 27 08:46:22 I think I'll check against the executable, if it is luna-qml-launcher then our qml ls2 adapter will add the -pid suffix May 27 08:46:52 With my modified bbclass it works OK for com.palm webapps May 27 08:47:26 Yes, that makes sense May 27 08:47:31 With my change in WAM we won't need to modify it May 27 08:47:45 OK great May 27 08:48:03 Because the bbclass would generate the -* variant which didn't work (I guess due to our WAM) May 27 08:48:17 exactly May 27 08:48:51 So I did what I could at my end, I prefer to not touch WAM LOL ;) May 27 08:49:07 Anyway I found some small issues as well in filecache May 27 09:20:08 Which caused the email body not to show, I'll relax the ACG a bit so we use wildcards instead of hardcoded app/services names May 27 09:25:17 https://github.com/Tofee/luna-webappmanager/commits/herrie/acg and https://github.com/webOS-ports/luneos-components/commits/tofe/acg are now my inwork branches, and I disabled the "com.palm" exception in the bbclass May 27 09:28:50 Tofe: You know why we have this special case for systemui there? May 27 09:29:05 yes, I just moved it a bit May 27 09:29:56 and made it adopt the same convention too; hope that works... May 27 09:30:10 But why is it there in general I wonder May 27 09:30:45 I guess the incoming systemui is started in a very unconventional way May 27 09:30:56 -the incoming May 27 09:31:27 flashing, let's hope I didn't break everything May 27 09:31:49 Hehe well it's easy enough to fix usually May 27 09:33:46 No bootloop, that's a good beginning May 27 09:34:15 Damn LSHUB_ROLE_FILE {} Can not find service "org.webosports.app.firstuse-*" permissions for executable "/usr/sbin/luna-qml-launcher" May 27 09:34:48 Can you paste full log? May 27 09:36:17 Herrie: https://paste.ubuntu.com/p/DMtXcZpmS4/ May 27 09:37:06 the rest isn't that bad May 27 09:39:19 well, firstuse is described in the role file as a "regular" app, with an "appId" and therefore no "exeName" May 27 09:41:02 I do have the container file for qml-launcher, so it's not that May 27 09:41:22 and the message only appears when firstuse is started, so the syntax of the role file is correct May 27 09:47:49 I get "org.webosports.app.firstuse-legacy" in ls-monitor, that's not good May 27 09:56:06 ok, that comes from libwebos-application May 27 09:58:51 ah, could it be that it fails to match org.webosports.app.firstuse-* against org.webosports.* ? May 27 10:54:25 That's a possibility May 27 10:54:33 Though it should be able to match it... May 27 10:58:01 going with the debugger in ls-hubd, I find that it takes "exe_path = org.webosports.app.firstuse" to begin the search May 27 10:58:13 that's bound to fail May 27 10:59:08 It could be we simply don't use/implement luna-qml-launcher as a real container, now that I think more about it. May 27 10:59:33 It we look at WAM, it has its own appId, and forwards the calls with LSCallFromApplication May 27 10:59:35 How do OSE guys do it? They also have a QML launcher I guess May 27 10:59:46 but luna-qml-launcher adopts the same id as the app it starts May 27 11:00:09 good idea, let's have a look :) May 27 11:00:39 Seems they don't really do QML apps? May 27 11:00:44 Most seem Enact? May 27 11:02:37 Ah they probably put in in SAM May 27 11:02:59 Yup: https://github.com/webosose/sam/search?q=QML&unscoped_q=QML May 27 11:03:07 https://github.com/webosose/qml-webos-framework/blob/master/tools/runner/main.cpp it's around here May 27 11:03:45 https://github.com/webosose/website/blob/0596c5629ffd24d1da76fd10703a11ba194c16d7/content/en/docs/tutorials/qml-apps/developing-built-in-qml-apps.md May 27 11:05:21 Tofe: Yeah that seems the right place May 27 11:06:03 https://github.com/webosose/qml-webos-framework/blob/master/tools/runner/lunaservicewrapper.cpp#L82 May 27 11:07:40 It's quite annoying to have links to all kinds of not publically accessible LG URLs.... May 27 11:11:26 it doesn't look like their examples are using appId-pid for qml apps, just for the runner May 27 11:12:31 https://github.com/webosose/meta-webosose/blob/master/meta-webos-smack/recipes-webos/com.webos.app.test.smack.native/com.webos.app.test.smack.native.bb that one could be a starting point for an example May 27 11:12:59 Herrie: will "inherit webos_app" trigger the "-" suffix ? May 27 11:13:46 ah, wait, it's native, not qml May 27 11:15:23 Tofe: Normally yes, inherit webos_app will generate the ACG files based on the requiredPermissions in appinfo.json May 27 11:15:57 looking at the help page you pointed to, a QML app recipe should also inherit that May 27 11:18:05 Anyway, one conclusion is that our qml launcher should have its own id May 27 12:03:53 Tofe: In general any app that makes LS2 calls should inherit webos_app and have the requiredPermissions in their appinfo.json May 27 12:05:14 At lesat that's my understanding May 27 12:07:43 yes, me too May 27 12:08:20 I just have the impression that OSE's qml runner is doing the calls with the real app id, not the one with a -pid suffix May 27 12:08:31 and that puzzles me a bit May 27 12:13:39 Herrie: you have a working OSE image, right ? May 27 12:13:51 Tofe: I guess so May 27 12:14:12 Just need to get it working in VBox May 27 12:14:17 I'd be interested in a zip of luna-service2 folder May 27 12:14:43 to double-check that they have this "-*" for their qml apps too May 27 12:15:11 Let me check May 27 12:15:45 I don't have a running image but got rootfs May 27 12:18:31 ah, great May 27 12:20:28 luna-service2-OSE.zip on ftp May 27 12:21:57 ok, the "-*" suffix is there May 27 12:32:20 It's a wonder their LS2 calls work from their QML apps. Or do they?... May 27 12:35:36 Well that's the question, do they May 27 12:35:39 Let me see what they are May 27 12:36:27 eos.widgetgallery.app would be a good candidate May 27 12:37:26 All appinfo.json in the image I have (might be a bit old one though) are web except for browser May 27 12:37:51 Which is "type": "native_appshell", May 27 13:09:49 I'll have to rework a bit libwebos-application May 27 13:10:11 and (again) our QML LS2 adapter May 27 13:15:10 Tofe: I don't think I have that app in my image May 27 13:15:15 At least I don't see it May 27 13:20:37 I should unpack the rootfs of latest OSE image May 27 13:20:54 JaMa: ^Seems there's no new OSE image on the build server yet, due to I/O problems? May 27 13:26:24 Herrie: could you give me write access to webOS-ports/libwebos-application.git ? May 27 13:29:54 Tofe: Done May 27 13:30:08 thanks May 27 14:32:44 Herrie: yes, all images and sstate are missing on file server May 27 14:33:05 Herrie: the builds are just to update sstate on the actual build server (and then it fails in rsync jobs) May 27 14:33:52 2020-05-20.log:10:42 <+JaMa> ka6sox: looks like /media/ra_build_share is no longer mounted on bonaire, are you still working on fixing those I/O issues? I don't have sudo (as jenkins user) to try re-mounting it May 27 14:34:52 novaldex: can you please check ^^ ? I don't know in what state file server is, but it would be pity if it's blocked just because I cannot re-mount it on bonaire (or reboot bonaire) May 27 14:39:21 novaldex: ideally if you can set some password on jama@bonaire and send it to me, I never had a password there and was using ssh key while nasadmin VPN worked for me, but now I need to login there from kaylee or milla where I don't have the right key (and don't want to copy my "default" one there), so I'm using sudo to jenkins on milla or kaylee and then ssh to bonaire, but then I cannot sudo May 27 14:40:04 well I guess i can tunnel it, will check that now May 27 14:45:03 OK, that works jama@bonaire:~ May 27 14:45:58 I've mounted it now, lets see if rsync jobs fail with IO again May 27 15:00:03 Tofe: Herrie: local build with final 5.15.0 finished, if it works are you OK if I push it to zeus and dunfell or should we do big rebuild together with ACG changes (as you seem to be very close to get it merged as well)? May 27 15:00:59 JaMa: Well if all OK I'm not against merging, but with Tofe's slower builder might be better to wait May 27 15:01:06 We're 98% there with ACG I'd say May 27 15:01:09 So should be soonish May 27 15:02:06 JaMa: as you wish. We'll have to do one after the other in all cases May 27 15:02:19 I'm already testing using a 5.15rc, so let's do 5.15 first May 27 15:03:50 JaMa: Corsair is sending new memory kit... DHL tracing received 2 days ago, still not picked up in Taiwan it seems, so might be sometime next week it arrives... May 27 15:05:19 Herrie: cool, hopefully it will be better than the previous 2 May 27 15:06:28 JaMa: Yes let's see.... May 27 15:06:39 Previous one was actually better compared to this one... May 27 15:06:48 This one had a way larger number of errors.... May 27 15:11:07 final 5.15.0 Qt seems to work the same as the previous SRCREVs (they were very close to this anyway), qtwebengine is actually newer than 5.15.0 because of fixes for gcc10 build May 27 15:11:26 will do few more tests without LuneOS and merge it to meta-qt5 soon May 27 15:11:35 great :) May 27 15:11:45 Then we'll stay on that LTS for some time I guess May 27 15:11:55 if you want to build with it already just use master-next branch of meta-qt5 and 2 top commits from jansa/zeus in meta-webos-ports May 27 15:12:17 yes, before Qt 6 is usable :) May 27 15:12:59 hopefully it won't take so long as migration from qt4 to qt5, but still it might be significantly longer then 5.14 -> 5.15 May 27 15:13:10 Herrie: if you have a list somewhere, I modified libwebos-application, luna-qml-launcher and luneos-components for ACG. But maybe it doesn't work at all, let's see :p May 27 15:13:21 especially if try to get rid of qmake as well :) May 27 15:13:41 What is the targetted build system ? May 27 15:13:57 meson ? cmake ? May 27 15:13:59 https://wiki.qt.io/CMake_Port May 27 15:14:12 Ok, I know that one a bit better May 27 15:18:38 Tofe: db8, filecache, webos-connman-adapter for now it seems May 27 15:19:05 Pushed those changes just not the bb updates yet May 27 15:20:05 oops : user.err luna-qml-launcher [] LS_INVALID_HANDLE {"COND":"sh != NULL","FUNC":"_LSCallFromApplicationCommon","FILE":"callmap.c","LINE":1818} sh != NULL: failed May 27 15:20:17 seems like I messed up my service handle pointers somewhere May 27 15:28:20 Herrie: let's not forget this little patch I did for luna-next May 27 15:28:43 Yes May 27 15:58:12 Tofe: All working now? May 27 16:04:40 Herrie: no :( May 27 16:05:07 I still get things like user.warning webos-telephonyd [] LS_REQUIRES_SECURITY {"SENDER":"org.webosports.luna-qml-launcher-4676","RECEIVER":"com.palm.telephony","SERVICE":"org.webosports.luna-qml-launcher-4676","CATEGORY":"/","METHOD":"networkIdQuery"} Service security groups don't allow method call. May 27 16:05:52 and to make it work I have to modify qml-launcher's client permissions... so it's not doing what I want May 27 16:06:40 I guess LSCallFromApplication isn't doing what I think it does May 27 16:06:55 I thought it permitted a service to send a call under a specific app name May 27 16:10:43 Tofe: Well you miss networking.internal in https://github.com/webOS-ports/luna-qml-launcher/blob/tofe/work/files/sysbus/org.webosports.luna-qml-launcher.perm.json#L3 May 27 16:11:30 Or networking.query May 27 16:11:49 I killed public/private as well in my reworks, could be that's why I didn't show upo May 27 16:12:08 I think in Qemu it might not do this query May 27 16:13:30 yes, I added networking.query for the launcher and it worked a bit better, but it the app client perm that should be modified, not the launcher May 27 16:16:43 I still don't have the country codes, it's maybe something unrelated to qml May 27 16:17:50 ah, ls-hubd LSHUB_NO_OUT_PERMS {"DEST_APP_ID":"org.webosports.luna-qml-launcher-3111","SRC_APP_ID":"com.palm.db","EXE":"/usr/sbin/mojodb-luna","PID":1372} "com.palm.db" does not have sufficient outbound permissions to c..... May 27 16:39:10 I'm trying another approach, where the qml app has its own service handle, separate from the handle of the qml launcher... May 27 16:41:47 damn: Invalid permissions for org.webosports.app.firstuse-3744 May 27 16:41:58 I can't even regiser the handle May 27 16:42:07 register* May 27 16:50:00 so, actually when reading more carefully OSE's code, the ls2 name of the qml runner is appId- May 27 16:51:06 which rises other questions, but anyway. May 27 17:09:07 http://jenkins.nas-admin.org/job/LuneOS/view/unstable/job/luneos-unstable_workspace-rsync/470/console took much longer but finished OK, fileserver might be OK now, will rebuild all to populate it May 27 17:09:14 sent 34,307 bytes received 71,861 bytes 212,336.00 bytes/sec May 27 17:09:31 2:18 May 27 17:10:33 90GB ram used (mostly tmpfs) and 2/2GB of swap as well, will reboot it as well to boot newer kernel May 27 18:06:46 ok, I *might* have something that begins to work May 27 18:09:52 but I think it's not really what should be May 27 18:10:31 For instance, I'm using LSRegisterApplicationService, which is nowhere to be seen in OSE May 27 18:12:32 Ah wait, maybe not. https://github.com/webosose/qml-webos-bridge/blob/66295ee8b3dd82597003fb62ac3b49891f0b61b3/plugin/lunaservicemgr.cpp#L132 May 27 18:15:55 Ah, if only I had remarked that before... May 27 18:16:02 Now I have some new idea :) May 27 18:16:04 ideas* May 27 19:02:16 Herrie: ok, it's working a bit better now. I don't get the country list, but I think it's an issue with palmdb May 27 19:02:33 I need to patch the phone app to add the requested permissions May 27 19:02:54 Herrie: or did you already do that? May 27 19:20:53 Herrie: I haven't PR'ed or merged anything, but here is the diff: https://github.com/Tofee/meta-webos-ports/commit/61bc793c6ac326bb4ed3fa603f1e71f6e528d450 May 27 19:21:53 That'll be all for today I think :) May 27 19:24:40 NotaBene: in my branch I already rebased it, so it might be incompatible with your, a little bit. But I only changed what's in the commit. May 27 19:25:50 This commit also includes the webapp scheme appId-, but I didn't test it as much as I should have May 27 19:35:29 I forgot a bb file. Now it is https://github.com/Tofee/meta-webos-ports/commit/69db01df28869c6e0e8b4bcc29c302b67754ef1d May 27 20:21:55 mido, rosy, tissot testing + qt 5.15-rc1 builds are now on fileserver May 27 20:31:35 rc1 ? May 27 20:44:27 "When qml-runner launches QML app, qml-runner calls the method to register the app to SAM. To enable qml-runner to call this method, put "applications" group" (from OSE's doc about doing a QML app). Ok, so it's what I thought, they put an external constraint (qml-runner needs a permission) onto the app developper (put "applications" in your app, pleaaase). I prefer the way I did, which separates the two needs. May 27 20:59:41 Tofe: yes, zeus currently builds rc1, I haven't pushed the final to meta-qt5 yet May 27 21:00:54 ok I see :) May 27 21:01:08 Anyway, bedtime now :p May 27 21:01:12 gn8! May 27 21:19:51 Tofe: was out for the night, looks like good progress! May 27 21:20:15 Did you take my db8 update? May 27 21:20:52 For phone app palmdb issues? **** ENDING LOGGING AT Thu May 28 03:00:05 2020