**** BEGIN LOGGING AT Mon Jun 18 03:00:07 2018
Jun 18 05:53:54 <Herrie> Morning!
Jun 18 08:16:27 <novaldex> JaMa: haven't touched DNS config on the builders, funny how they're getting that all of a sudden. any particular entries that aren't resolving or is it all of them?
Jun 18 08:28:41 <novaldex> morning all..
Jun 18 08:39:27 <Herrie|Laptop> novaldex: Morning!
Jun 18 08:40:34 <novaldex> morning Herrie|Laptop! How's things?
Jun 18 08:41:03 <Herrie|Laptop> All good here :)
Jun 18 08:41:36 <novaldex> :) glad to hear it
Jun 18 08:42:13 <novaldex> did you see JaMa's message to me over the weekend about DNS resolution?
Jun 18 08:51:42 <Herrie|Laptop> Yeah
Jun 18 08:51:48 <Herrie|Laptop> If you want me to test anything let me know
Jun 18 08:55:52 <Tofe> Morning!
Jun 18 08:56:09 <novaldex> thanks, could you give me a bit more info first please?
Jun 18 08:56:39 <Tofe> My new build, with externalized PalmServiceBridge, a central qtwebengine.conf for chromium flags, on top of Qt 5.11, is ready for testing :)
Jun 18 08:57:02 <Herrie|Laptop> Tofe: ooh nice
Jun 18 08:57:10 <Herrie|Laptop> You saw my message about Testr
Jun 18 08:57:11 <Herrie|Laptop> ?
Jun 18 08:57:17 <Tofe> yes, that'll help too
Jun 18 08:58:30 <Tofe> Ideally I would like to test it interactively, but I'm not sure if I can forward the ws:// inspector URL via adb
Jun 18 09:00:25 <Herrie|Laptop> novaldex: I can only find one reference in logs: https://bpaste.net/show/d2109220c1d9
Jun 18 09:30:19 <novaldex> thanks, do you know which builder that was on?
Jun 18 09:30:36 <novaldex> I gave bonaire, abaco & kwaj some quick tests & they seem to be okay, resolv.conf looks correct
Jun 18 10:45:45 <Herrie|Laptop> Tofe: Working on the kernel & meta-smartphone for tissot a bit. Seems it uses the new _a and _b partition layout
Jun 18 10:45:50 <Herrie|Laptop> I'll assume we need _a for now
Jun 18 10:45:52 <Herrie|Laptop> https://forum.xda-developers.com/pixel-xl/help/help-understanding-b-partitions-t3748003
Jun 18 10:45:59 <Herrie|Laptop> https://source.android.com/devices/tech/ota/ab/
Jun 18 10:46:45 <Herrie|Laptop> Partition layout: https://4pda.ru/forum/index.php?s=&showtopic=843386&view=findpost&p=65508794
Jun 18 10:51:18 <Tofe> Herrie|Laptop: first time I see that layout, let me read a bit
Jun 18 11:08:18 <Tofe> Herrie|Laptop: ok I understand; I'm surprised an Android 7 device already uses that partition scheme
Jun 18 11:24:24 <Tofe> Herrie|Laptop: we'll have to adapt our flashing script yes
Jun 18 11:24:49 <Tofe> In the end, the only solution might be to flash both to _a and _b
Jun 18 11:25:48 <Tofe> The Halium script might need to adjustments too...
Jun 18 11:25:56 <Tofe> s/to/some/
Jun 18 11:46:18 <Herrie|Laptop> Tofe: Well it could be the partition layout came from Android 8 rom actually
Jun 18 11:46:19 <Herrie|Laptop> Not sure
Jun 18 11:52:13 <Tofe> Not sure either
Jun 18 11:52:57 <Tofe> https://www.xda-developers.com/list-android-devices-seamless-updates/ "That’s why Google introduced “seamless updates”—a new A/B dual partition scheme with Android Nougat."
Jun 18 12:04:43 <Herrie|Laptop> Tofe: Well it seems it's one of the few ;)
Jun 18 12:08:04 <Herrie|Laptop> Tofe: In general for us it shouldn't really matter if a kernel is Nougat or Oreo right?
Jun 18 12:08:25 <Tofe> I guess so
Jun 18 12:09:40 <Herrie|Laptop> Just there is Xiaomi's kernel which is 3.18.31 for Nougat and there's also some "stable" one which seem to be 3.18.133 which is for Oreo
Jun 18 12:10:46 <Herrie|Laptop> Ehm 3.18.113 that is: https://github.com/android-linux-stable/tissot
Jun 18 12:11:12 <Tofe> Both should work I guess
Jun 18 12:12:43 <Herrie|Laptop> I guess so too
Jun 18 12:12:57 <Herrie|Laptop> Just having 82 minor kernel updates wouldn't hurt
Jun 18 12:13:08 <Herrie|Laptop> I guess I can try both to see which one works or at least compiles
Jun 18 12:13:19 <Herrie|Laptop> I remember I tried "newest" one on Mido and wouldn't compile even
Jun 18 12:13:56 <Tofe> Well, some repositories out there are really in alpha shape :)
Jun 18 12:16:43 <Herrie|Laptop> Yup
Jun 18 12:16:59 <Herrie|Laptop> I think I have a defconfig more or less complete now
Jun 18 12:17:06 <Herrie|Laptop> Will put together a recipe and see if stuff builds
Jun 18 12:19:35 <Tofe> https://github.com/Halium/initramfs-tools-halium/pull/14 <-- good, I'll have to check how far it is from our current halium script
Jun 18 12:27:59 <Herrie|Laptop> Tofe: https://github.com/Halium/initramfs-tools-halium/pull/14
Jun 18 12:28:03 <Herrie|Laptop> Ah
Jun 18 12:28:06 <Herrie|Laptop> You found it too :P
Jun 18 12:29:23 <Tofe> Seems they made it backward compatible, I hope it works :p
Jun 18 12:32:00 <Tofe> https://github.com/webOS-ports/meta-webos-ports/blob/master/meta-luneos/recipes-core/initrdscripts/initramfs-boot-android_1.0.bbappend#L4 I should update that sometime...
Jun 18 12:33:14 <Herrie|Laptop> Tofe: Yes seems so :P
Jun 18 13:08:09 <novaldex> JaMa: ping?
Jun 18 13:09:26 <JaMa> pong
Jun 18 13:09:47 <novaldex> hey, I got your message about DNS, but i'm still a bit puzzled
Jun 18 13:10:34 <novaldex> it appears to be fine & resolving, but that's from the command line on bonaire
Jun 18 13:10:56 <JaMa> it works most of the time
Jun 18 13:11:06 <JaMa> it fails in 0.1% maybe
Jun 18 13:11:25 <novaldex> are the failures completely random? always at a different point?
Jun 18 13:11:35 <JaMa> it uses this PRservice maybe a 1000 times per build, but only 1-2 of them fail per build
Jun 18 13:11:39 <JaMa> yes
Jun 18 13:11:56 <novaldex> I notice the difference between bonaire & abaco/kwaj, the latter two have additional external DNS servers listed as well
Jun 18 13:12:04 <novaldex> something related to PRserver instead perhaps?
Jun 18 13:12:16 <novaldex> *PRService
Jun 18 13:12:30 <JaMa> abaco/kwaj aren't using this PRservice, so cannot say if they might be failing as well or not
Jun 18 13:12:46 <novaldex> okay, looking for similarities
Jun 18 13:13:00 <JaMa> maybe we can add this host to /etc/hosts to make it a bit more reliable
Jun 18 13:13:29 <novaldex> which host? happy to try it
Jun 18 13:17:46 <JaMa> jenkins.nas-admin.org
Jun 18 13:18:42 <novaldex> to /etc/hosts on bonaire? sorry, just wanting to make sure i'm getting it right
Jun 18 13:19:47 <novaldex> hmmm, just spotted something.. checking
Jun 18 13:27:24 <novaldex> okay, i've not modified the hosts file yet, but I have tweaked resolv.conf to add a new entry
Jun 18 13:27:56 <novaldex> can we try this for now & we can go with the hosts changes if it's no help?
Jun 18 13:30:36 <JaMa> the builds are already running, lets see if some of them fail this time, thanks
Jun 18 13:30:41 <novaldex> thanks
Jun 18 13:47:20 <bshah> Tofe: ping
Jun 18 13:47:41 <bshah> (or Herrie|Laptop )
Jun 18 13:47:49 <Tofe> bshah: pong
Jun 18 13:48:01 <bshah> Tofe: you mentioned issues with qt 5.11 and libhybris
Jun 18 13:48:06 <bshah> I think I am hitting that
Jun 18 13:48:15 <bshah> kwin_wayland doesn't start at all
Jun 18 13:48:28 <bshah> can you give more details/
Jun 18 13:48:40 <Tofe> ah, mmh, wait just a bit that I remember
Jun 18 13:48:42 <Herrie|Laptop> bshah: We're in a better shape then ;) stuff boots for us ;)
Jun 18 13:48:51 <Herrie|Laptop> But Tofe did some magic :P
Jun 18 13:48:52 <JaMa> bshah: hi, can you please merge https://github.com/Halium/android_kernel_lge_hammerhead/pull/2 ?
Jun 18 13:49:06 <bshah> JaMa: let's see
Jun 18 13:49:30 <Tofe> bshah: ah yes, TLS corruption
Jun 18 13:49:34 <bshah> Tofe: I get "Thread 1 "kwin_wayland" received signal SIGBUS, Bus error."
Jun 18 13:49:38 <bshah> is it one?
Jun 18 13:49:40 <Tofe> yes
Jun 18 13:50:03 <Herrie|Laptop> Tofe: Was this the qpa patch for 5.1x with the regex or another one?
Jun 18 13:50:15 <Tofe> There are two ways or fixing that: one is just a workaround, but an easy one
Jun 18 13:50:33 <Tofe> Herrie|Laptop: no, that's another one
Jun 18 13:50:39 * bshah afk for 5 min, brb
Jun 18 13:51:26 <Tofe> (let me find my patch...)
Jun 18 13:54:07 <bshah> JaMa: merged
Jun 18 13:54:54 <Tofe> aaaargh, JaMa, do you remember where I committed my patch ? I did #undef HAVE_TLS instead of #define HAVE_TLS in qtbase somewhere
Jun 18 13:57:33 <Tofe> bshah: so basically, https://github.com/qt/qtbase/blob/5.11/src/corelib/thread/qthread_unix.cpp#L114 here , you undef HAVE_TLS
Jun 18 13:58:06 <Tofe> the second option is to try the libhybris patch that shifts a bit more the TLS slots in libhybris; so far it has only been tried on x86
Jun 18 13:58:27 <Tofe> --> this one https://github.com/mer-hybris/android_bionic/commit/493fae6a60b65a9183cd04f4fcc23dfbecce5141
Jun 18 14:00:31 <JaMa> Tofe: sorry I was deleting e-mails from github and there were more then expected :)
Jun 18 14:00:34 <JaMa> bshah: thans
Jun 18 14:00:37 <JaMa> +k
Jun 18 14:01:14 <Tofe> JaMa: no problem, I found the file in the end
Jun 18 14:22:13 <bshah> Tofe: what patch you use?
Jun 18 14:22:44 <Tofe> bshah: right now I'm disabling TLS in Qt
Jun 18 14:22:58 <Tofe> But I'd prefer patching libhybris, if it works
Jun 18 14:23:12 <Tofe> s/libhybris/bionic/
Jun 18 14:23:26 <bshah> urgh.. requires rebuilding android image
Jun 18 14:23:30 <bshah> *yawn*
Jun 18 14:23:35 <Tofe> exactly why I didn't do that yet :)
Jun 18 14:24:31 <bshah> Tofe: havee you pin-pointed exactly what changed in qtbase to cause this?
Jun 18 14:25:07 <bshah> and wouldn't this have any side-effects on other platforms?
Jun 18 14:25:21 <bshah> (we have shared Qt between everything)
Jun 18 14:26:57 <bshah> wait what
Jun 18 14:27:11 <bshah> I am confused
Jun 18 14:30:06 <Tofe> I'm currently only patching for arm
Jun 18 14:32:47 <Tofe> bshah: more in detail, what you can see is that currentThreadData (https://github.com/qt/qtbase/blob/5.11/src/corelib/thread/qthread_unix.cpp#L121) gets corrupted by some call in bionic; the result is that you get a non-NULL pointer, but still invalid. As a result (for me) the Qt OpenGL texture object will get wrong data and trigger a SIGBUS.
Jun 18 14:34:16 <Tofe> This probably happens now because of some change in the threading model of Qt's SceneGraph renderer
Jun 18 14:34:46 <bshah> I am leaning towards libhybris/bionic change most likely
Jun 18 14:35:30 <Tofe> bshah: that's the cleanest yes; I'll do that too in the end, for sure. I mostly wanted to move on and see if there were other surprises ahead
Jun 18 14:36:26 <Tofe> bshah: you probably noticed you needed patching qpa
Jun 18 14:36:45 <bshah> Tofe: oh?
Jun 18 14:37:34 <Tofe> yes, if you use the code base from mer
Jun 18 14:37:50 <Tofe> https://git.merproject.org/Tofe/qtscenegraph-adaptation/blob/master/customcontext/customcontext.pro#L6 this test becomes invalid for Qt 5.10+
Jun 18 14:40:51 <Tofe> I've fixed it like this: https://git.merproject.org/Tofe/qtscenegraph-adaptation/blob/tofe/qt5.8/customcontext/customcontext.pro#L6
Jun 18 14:41:16 <Tofe> I should PR this when I get some time...
Jun 18 14:41:18 <bshah> We don't really use qtscenegraph-adaption
Jun 18 14:41:51 <Tofe> ah ok
Jun 18 15:06:07 <ka6sox-away> Good morning
Jun 18 15:07:37 <Herrie|Laptop> ka6sox-away: Morning
Jun 18 15:08:08 <ka6sox-away> hi hi...I see builds for webosOSE happening... This seems godo.
Jun 18 15:08:10 <ka6sox-away> good
Jun 18 15:08:22 <Herrie|Laptop> Yes seems so
Jun 18 15:19:50 <JaMa> they aren't triggered automatically yet, but with the frequency of the build-webos changes I think it's ok
Jun 18 15:20:23 <JaMa> I've prepared them to be triggered by github event from build-webos (and blocked when any luneos-* build is running - because they share tmpfs)
Jun 18 16:29:28 <bshah> Tofe: ping
Jun 18 17:06:05 <Tofe> bshah: pong
Jun 18 17:06:32 <bshah> see #libhybris
Jun 18 17:31:41 <Tofe> JaMa: I found out why I couldn't find my Qt patch ! I never committed it...
Jun 18 17:45:31 <Tofe> ah no, it was the -no-feature-renameat2 stuff that I forgot
Jun 18 17:45:47 <Tofe> JaMa: ^ did you push something in meta-qt5 regarding this ?
Jun 18 19:22:52 <Herrie> Tofe: The new fancy kernel gave me some headaches for Halium, but sorted that now
Jun 18 19:50:37 <Tofe> Herrie: good job :)
Jun 18 19:55:43 <Herrie> Just need to PR and test the meta-smartphone bits
Jun 18 19:55:52 <Herrie> See if we have surprises in our build :P
Jun 18 20:44:15 <Tofe> ok well, my PalmServiceBridge needs some tweaking, but the basics seem to work. Testr can call it, for instance. There are a couple of bugs to fix, and then we're good to go :)
Jun 18 21:06:31 <Tofe> ok after a few fixes it already works pretty well
Jun 18 22:18:58 <JaMa> Tofe: yes, I've pushed the PACKAGECONFIG which will disable it by default
Jun 18 22:48:43 <elvispre> ping saidinesh5
Jun 18 22:48:52 <saidinesh5> what's up elvispre?
Jun 18 22:49:19 <elvispre> saidinesh5: did you ever resolve "do you know what kernel config/ patch i need to enable (nested namespaces/cgroups?)"
Jun 18 22:49:37 <elvispre> "5
Jun 18 22:49:37 <elvispre> ERROR    lxc_utils - utils.c:mkdir_p:253 - Invalid argument - failed to create directory '/sys/fs/cgroup/net_cls//lxc/default"
Jun 18 22:49:45 <saidinesh5> Ahh
Jun 18 22:49:48 <saidinesh5> yeah
Jun 18 22:49:54 <saidinesh5> you using 3.4 kernel?
Jun 18 22:50:00 <elvispre> Yes
Jun 18 22:50:41 <saidinesh5> simply disable that namespace feature.. after so much googling and so much nagging people, i found out that that wasn't ready in 3.4 kernel and it wasn't useful for anbox anyway
Jun 18 22:50:45 <elvispre> I am googling and your name came up :)
Jun 18 22:51:44 <saidinesh5> it is mostly used for "accounting where your data goes" iirc
Jun 18 22:51:56 <elvispre> Disable that namespace feature in the kernel or in lxc? (Does that make sense?)
Jun 18 22:52:15 <saidinesh5> in kernel .. then lxc will automatically try to not use it
Jun 18 22:52:26 <elvispre> Cool. Thanks.
Jun 18 22:52:26 <saidinesh5> the problem was that the kernel says that feature was working, but it wasn't
Jun 18 22:53:01 <saidinesh5> i think you can safely disable any of the optional lxc-checkconfig requirements
**** ENDING LOGGING AT Tue Jun 19 03:00:03 2018