**** BEGIN LOGGING AT Thu Sep 29 02:59:58 2016 Sep 29 04:23:57 morning Sep 29 06:52:42 Morning Sep 29 06:53:02 Herrie: are you confident with the current set of PRs for renaming services? Sep 29 07:03:17 Tofe: I think this should be fine, no? As long as we make sure that references in the other .service files are updated to the new name too? Sep 29 07:03:23 And I cross checked that Sep 29 07:03:45 Just when the naming of the .service is not consistent with the component we need to add 3-4 extra lines in each recipe :P Sep 29 07:03:59 If it's consistent we can just inherit webos_systemd instead of systemd ;) Sep 29 08:04:17 ok, let's merge then Sep 29 08:10:15 Tofe: Yeah, I'm testing it anyway before pushing the final bits Sep 29 09:52:54 Tofe: Hmmmz bootloop, I guess I must have missed or fat fingered something :P Some checking to do :P Sep 29 10:26:02 I'll compare them side by side to see what I've done wrong :P Sep 29 12:45:04 Herrie, ping Sep 29 12:45:39 Herrie|Pre3, ping Sep 29 12:47:19 novaldex: pong Sep 29 12:49:29 hey Herrie, think i've sorted that PR_Service issue from yesterday Sep 29 12:55:59 OK will try when home by disabling my config option to use the local PR_Service :P Sep 29 12:57:04 thanks! there must be a better solution going forward, but working is a good step Sep 29 12:59:01 novaldex: Not sure what was the issue and what it took to solve it? Sep 29 12:59:12 Any luck on that kernel version mystery? Sep 29 13:15:46 Herrie|Pre3: it's a manual process start to make it available again Sep 29 13:16:20 no idea on the kernel thing, it's been shared with scoutcamper & cryptk, but will probably take us a bit to figure out Sep 29 13:16:29 we don't expect it to cause problems though Sep 29 13:17:34 novaldex: Yeah I guess it shouldn't. Where are we with Bonaire? ka6sox inquired y'day about if it was OK to do an upgrade. Not sure that started already or when it's expected? Sep 29 13:18:58 we've been talking about Tahiti & Bonaire, but I think that'll be next week at the earliest Sep 29 13:19:23 i'd like to take care of existing issues before setting up new ones.. Sep 29 13:20:46 OK Sep 29 13:20:52 I'll kick off a build then :) Sep 29 13:20:58 See how things behave Sep 29 13:24:49 thanks.. Sep 29 15:14:55 Herrie, is the prserver sorted? Sep 29 16:10:46 ka6sox: I think so. Builds on Jenkins work, so I guess it is. I'm home in 30 or so to double check. Sep 29 16:12:31 Tofe: ping Sep 29 16:13:28 Seems that we might be able to use the --unlimited-storage as param for LunaWebAppManager at least if I read Chromium documentation correctly. Might be worth a try. Sep 29 16:13:54 Pending a proper quota management implementation in qtwebengine Sep 29 17:31:15 nizovn: ping Sep 29 17:31:21 Herrie: pong Sep 29 17:31:47 nizovn: You have any idea how our boot sequence works? Sep 29 17:32:05 I've been playing around renaming some systemd .service files and now seems my Mako doesn't boot :S Sep 29 17:32:28 I still have some .conf files in the images which I believe to be leftovers from when we were using Upstart, but not sure.... Sep 29 17:33:47 so you don't have even adb connection? Sep 29 17:36:10 nizovn: It seems to bootloop on Google logo :S Sep 29 17:36:41 This is what I have done: https://github.com/webOS-ports/meta-webos-ports/compare/krogoth...herrie/systemd?expand=1 Sep 29 17:36:50 you can try to get boot logs from android /proc/last_kmsg Sep 29 17:36:50 Let me reboot my wifi router, brb Sep 29 17:40:13 Herrie: pong Sep 29 17:41:17 Tofe: See my suggestion above for quota Sep 29 17:41:20 Herrie: ok, I'll investigate and white-list this option Sep 29 17:46:09 I merged DougReeders test for this in Testr Sep 29 17:47:38 nizovn: I guess I'll first compare latest nightly and my image side by side to see if there's anything obvious I'm missing Sep 29 17:48:09 ok Sep 29 17:48:34 * DougReeder checks in Sep 29 17:52:11 nizovn: I get the qemux86 doing the same :S Sep 29 17:52:19 Must have missed something crucial somewhere Sep 29 17:52:27 I.e. forgot to change a name somewhere or something Sep 29 17:52:43 It just get stucks on "booting the kernel" Sep 29 17:54:14 you can see vbox logs somewhere Sep 29 17:54:33 maybe $HOME/VirtualBox VMs/{machinename}/Logs Sep 29 18:03:09 Herrie: I confirm this kind of policy isn't managed yet: https://github.com/qt/qtwebengine/blob/dev/src/core/browser_context_qt.cpp#L160 Sep 29 18:06:02 But if we want to put some limits depending on the origin URL (local or remote), we could write a simple implementation in https://github.com/qt/qtwebengine/blob/dev/src/core/content_browser_client_qt.cpp#L339 Sep 29 18:06:56 The values would be quite hardcoded though Sep 29 18:08:17 Tofe: That's OK Sep 29 18:08:22 I think 5MB would do? Sep 29 18:09:13 Ah we did 64 max in our previous implementation: https://github.com/webOS-ports/luneos-components/blob/master/modules/LuneOS/Components/LunaWebView.qml#L112 Sep 29 18:10:04 5MB, up to 64MB for local ? And only 5MB for remote ? Sep 29 18:11:48 When we allow up to 64 MB will it reserve it right away or adjust on the fly? Sep 29 18:12:00 If it adjusts on the fly can simply make it 64 Sep 29 18:12:42 should adjust, I guess. It just asks us if x bytes would be ok, and we say yes/no Sep 29 18:13:25 But remote URL would include the browser, that's why I wanted to be more restrictive for that Sep 29 18:14:50 OK Sep 29 18:15:03 Sounds like a good idea Sep 29 18:15:25 I've already written the logic, now I've got to use the actual correct variables :p Sep 29 18:15:26 Adjusting values when needed anyway should be straight forward Sep 29 18:15:30 :P Sep 29 18:22:31 Herrie, Testr, and the SDL2 OpenGLES* apps really belong in their own tab. Sep 29 18:22:54 But if you want ot put them in Prefs, that’s probably less bad than “Apps” Sep 29 18:24:49 If you’re busy, I can work on that. Sep 29 18:25:45 Herrie: how can I test the quota thing works ? I update my Testr app ? Sep 29 18:26:28 DougReeder: actually, maybe you do know that :) Sep 29 18:26:46 Yes, Testr runs a cmplete scenario now using quota. Sep 29 18:27:11 ok, great! Sep 29 18:28:19 On my Pre 3, I had a launcher tab “Seldom” for apps I rarely needed. I wonder if there’s some broad concept along those lines we could exile Testr and the SDL2OpenGLES* apps to. Sep 29 18:28:52 DougReeder: I think we don't have an option to dynamically add tabs yet Sep 29 18:28:59 I.e. User created ones like in 2.x Sep 29 18:29:03 Testr and SDL2* apps are some quite advanced apps, more related to dev Sep 29 18:29:13 * DougReeder nods Sep 29 18:29:14 can't we have something along the "Dev" line? Sep 29 18:29:15 3.x didn't support that unless you used some hacks Sep 29 18:30:06 nizovn: Log from vbox: Anything that looks suspicious to you? https://bpaste.net/show/ea99c192873d Sep 29 18:31:07 Tofe: Latest Testr IPK: http://build.webos-ports.org/luneos-testing/ipk/all/org.webosports.app.testr_0.1.0%2Bgitr62%2B915cd86d92-r0.0_all.ipk Sep 29 18:31:19 Would “Advanced” make sense as a tab? Sep 29 18:32:23 Herrie: thanks Sep 29 18:33:08 DougReeder: I guess that's a bit more generic Sep 29 18:33:12 The distinction beween “Apps” and “Downloads” doesn’t matter to the user (or shouldn’t) Sep 29 18:33:49 https://github.com/webOS-ports/luna-next-cardshell/blob/master/qml/LaunchBar/FullLauncher.qml#L232 Sep 29 18:33:50 I’ve wondered if “System” is a reasonable tab, but suspect that’s not quite right. Sep 29 18:34:53 cid3baf5 Sep 29 18:36:08 DougReeder, Herrie: does that look ok? https://dl.dropboxusercontent.com/u/4679068/luna-next/Testr_QuotaRequest_WebEngine.png Sep 29 18:37:47 If I request 20MB, it's granted too. Then if I request 80MB, it only grants 20MB. Sep 29 18:38:14 Looks good to me Sep 29 18:38:25 The last line should read “wrote 200000 bytes…” Sep 29 18:39:03 … but something similar happens in Chrom when you attempt to write 300000 bytes. Sep 29 18:39:29 maybe we get confused by the "progress", but the 2MB get eventually written? Sep 29 18:40:11 In Chrome, you can write up to abut 250000 bytes, without a PreogresEvent. Sep 29 18:40:53 So, while something didn’t go as expected here, it might be Chrome’s fault. Sep 29 18:41:11 ok; anyway it's better than before. Sep 29 18:41:22 Can you write smaller amounts without the ProgressEvent showing up? Sep 29 18:41:31 Certainly this is better! Sep 29 18:42:13 Herrie: https://github.com/webOS-ports/qtwebengine/pull/37 here you go Sep 29 18:43:04 30,000 bytes does that Sep 29 18:43:53 Even 400,000 seems fine Sep 29 18:45:25 Tofe: Nice :D Sep 29 18:46:05 It’s possible that the ProgressEvent is sent when you exhaust your quota, because of the current file and other files. Sep 29 18:46:44 Maybe... but we shouldn't exclude a Chromium bug here either Sep 29 18:47:31 Serene always asks for an increase of quota beforehand, if necessary Sep 29 18:47:33 Herrie: with the DB quota back to life, we should be able to support a whole bunch of legacy apps now, correct? Sep 29 18:49:48 Okay, the 32768 value is not the number of bytes written to the file. Sep 29 18:51:00 Currently the file is truncated before writing, but probably it should be completely deleted Sep 29 18:51:19 ok Sep 29 18:51:36 Tofe, can you nuke the sandbox filesystem, and then try with 200000? Sep 29 18:51:50 ..how do I nuke it? Sep 29 18:52:01 Good question. :-S Sep 29 18:52:05 hehe :p Sep 29 18:55:39 "rm -rf ./media/cryptofs/.sysmgr/cache/LunaWebAppMgr/QtWebEngine/Default/Cache/" did it Sep 29 18:55:45 DougReeder: now it writes it fine Sep 29 18:56:09 ok, my test is poorly written; I’ll fix it. Sep 29 18:56:24 BUt now I know what quota exhaustion actually looks like! Sep 29 18:56:46 oh wait Sep 29 18:56:55 it wasn't the correct directory Sep 29 18:58:14 It's in "/media/cryptofs/.sysmgr/data/LunaWebAppMgr/QtWebEngine/Default/File System" Sep 29 18:58:40 I get a NotFoundErr for the FileWriter Sep 29 18:58:56 On the second try, it works well Sep 29 18:59:04 Hmm Sep 29 18:59:27 but it could be I nuked it too violently... Sep 29 18:59:34 * DougReeder nods Sep 29 19:00:21 yup, restarting LunaWebAppManager after the nuke does help Sep 29 19:00:25 it works fine Sep 29 19:00:51 probably some caching data in chromium Sep 29 19:00:57 Herrie: imho the log looks ok. Can you ssh into VM? Sep 29 19:04:59 One thing that concerns me, is the name of the sandbox filesystem: file__0:Persistent Sep 29 19:05:20 If all apps use that name, I wonder if they’re using the same sandbox file system. Sep 29 19:10:28 DougReeder: well, we use of course separate WebEngine instances for each app Sep 29 19:10:39 * DougReeder nods Sep 29 19:11:03 so it's like different tabs in chromium Sep 29 19:11:36 but same process Sep 29 19:12:42 In Chrome, different tabs running Testr share the same sandbox filesystem. Sep 29 19:12:59 … because they are the same origin. Sep 29 19:13:53 I wonder how I could test that Sep 29 19:14:26 I can open a second Testr window, but then what? Sep 29 19:14:47 We’d want to test with a second app with a different ID. Sep 29 19:14:53 ah yes, right Sep 29 19:15:08 can I just duplicate testr and change appinfo? Sep 29 19:15:17 I think so. Sep 29 19:15:40 Put the copy in a different diretory, and edit appinfo to match Sep 29 19:16:52 done, launched Sep 29 19:17:29 for Testr, I created a 200K file. What should I do with Testr2 ? Sep 29 19:18:07 Any size should work Sep 29 19:18:10 Create a 300K file, and look at the storage? Sep 29 19:18:20 since the first should have exhaused storage. Sep 29 19:18:47 300K written Sep 29 19:19:38 Whatever the largest quota you’ve asked for in Testr, do that again, and that should exhaust the quota. In Testr2, any size write should fail. Sep 29 19:19:39 Physically, the same file has been modified on disk Sep 29 19:19:46 Oi! Sep 29 19:20:17 Okay, my proceedure was not a proper test. Sep 29 19:20:49 well, that still worries me... Sep 29 19:21:06 If they modify the same file, we’ve got a probelm Sep 29 19:21:41 they're playing in the very same sandbox, yes Sep 29 19:21:54 Quite a useless sandbox... Sep 29 19:22:01 * DougReeder nods Sep 29 19:22:34 They are probably the same origin for other purposes, which is also a security hole. Sep 29 19:22:50 yep Sep 29 19:23:48 So, I guess we add that to the to-do list. Sep 29 19:23:52 :-( Sep 29 19:23:54 ahah, I think I found a possible hint Sep 29 19:24:40 among many other switches that we put on the child WebEngine process, there is a ... "--no-sandbox" Sep 29 19:25:02 Firefox OS also confronted this. Different apps had unique (weird) origins Sep 29 19:25:38 nizovn: No SSH Sep 29 19:25:41 I'll dig a bit more Sep 29 19:33:30 Tofe: That switch doesn't help I'm sure Sep 29 19:34:46 I suspect we need code to set the origin for each app. Sep 29 19:35:11 From a web browser’s perspective, they’re all just files on the local file system. Sep 29 19:38:48 nizovn/Tofe: You know if we still use the stuff in /etc/event.d/ ? Sep 29 19:39:58 Same question would apply to /etc/init Sep 29 19:40:18 I'm not sure how our whole boot process works before it ends up with systemd? Sep 29 19:40:39 not sure Sep 29 19:41:28 wanna try https://github.com/qt/qtwebengine/commit/7a49313c84ccd4779e396b7bc8341a331d90f96f ? :p Sep 29 19:42:19 Tofe: Hell why not :P Sep 29 19:42:42 Just my builds are broken due to me playing with systemd :P Sep 29 19:42:46 Trying to sort that first Sep 29 19:44:30 I thought systemd was process 1 ? Sep 29 19:44:52 Oh, you mean the firmware boot, before linux. Sep 29 19:45:38 DougReeder: I'm not sure... I tried to innocently rename some of the .service files (while updating references to them in other .service files so simplify the bitbake recipes. Sep 29 19:46:12 When we inherit systemd we need 3-4 lines in each recipe to get the .service installed. When we have them the same name as component we can just inherit webos_systemd and all is taken care off ;) Sep 29 19:46:24 So I renamed LunaSysMgr.service to luna-sysmgr.service etc Sep 29 19:46:36 But now I'm stuck at Google logo :P Sep 29 19:47:46 Ah seems I forgot: https://github.com/webOS-ports/meta-webos-ports/blob/krogoth/meta-luneos/recipes-core/android-system/android-system/change-before-dependencies.conf#L2 Sep 29 19:47:55 Let me try with that fixed Sep 29 19:48:47 That seems like a crucial one :P Sep 29 19:49:03 And yepee for systemd not helping here... Sep 29 19:49:27 Tofe: Well at least it gets cleaner when it works ;) Sep 29 19:49:35 that won't help with vbox Sep 29 19:49:43 nizovn: Ehm no.... Sep 29 19:49:53 But worth a try with the mako at least Sep 29 19:54:54 i don't think those packages that you've changed are crucial for boot, maybe it's systemd fault itself.. Sep 29 19:59:39 You can always revert your changes, and add them back one by one util the builds fail. :-S Sep 29 20:00:27 Herrie: one way to check it is to look into the lune-dev-image.tar.gz archive, and grep the whole /lib/systemd/system directory for old strings Sep 29 20:01:20 Yeah let me do a full rebuild first Sep 29 20:01:29 Could be I simply missed something somewhere Sep 29 20:03:18 ok, my naïve way of activating sandboxing didn't help. There must other pieces I'm missing, unsurpringly :) Sep 29 20:03:24 +si **** ENDING LOGGING AT Fri Sep 30 02:59:58 2016