**** BEGIN LOGGING AT Tue May 15 03:00:02 2018
May 15 04:30:13 <Herrie> Morning!
May 15 04:30:43 <Herrie> Tofe: On the Mido pyro build seems I have the same issue. I see the following "error" in the logs: May 14 16:07:09 mido systemd-udevd[767]: /etc/udev/rules.d/99-anbox.rules:1: NAME="%k" is ignored, because it breaks kernel supplied names; please remove
May 15 04:30:50 <Herrie> Which is caused by https://github.com/anbox/anbox/blob/master/kernel/99-anbox.rules
May 15 04:30:58 <Herrie> That might be a reason, not sure?
May 15 04:31:35 <Herrie> Also see May 14 16:11:34 mido anbox[1030]: Anbox bridge interface 'anbox0' doesn't exist. Network functionality will not be available
May 15 04:31:41 <Herrie> But that shouldn't be critical
May 15 04:36:29 <Herrie> Tofe: We'll probably need this, but shouldn't be critical to get it launched? https://github.com/anbox/anbox/pull/580/files
May 15 06:06:56 <Herrie|Laptop> Morning!
May 15 06:07:03 <Herrie|Laptop> morphis_: ping
May 15 06:50:52 <Tofe> Morning
May 15 06:52:08 <Tofe> Herrie: on Archlinux, for Anbox-git recipe, they changed slightly the rules: https://aur.archlinux.org/cgit/aur.git/tree/99-anbox.rules?h=anbox-git
May 15 06:52:33 <Tofe> maybe that would be better, I quite don't know
May 15 06:54:36 <Tofe> And you're right, we'll need the touchscreen support at some point
May 15 06:58:33 <Tofe> Also I can see that ubports guys have quite a bunch of commits https://github.com/anbox/anbox/compare/master...ubports:xenial tough I don't know which branch should be looked at
May 15 07:09:13 <Herrie|Laptop> Tofe: A lot of those are UBPorts specific with Mir etc
May 15 07:10:15 <Herrie|Laptop> Only thing that might be interesting is the touchscreen and the binder stuff, just not sure how relevant that is
May 15 07:20:41 <Tofe> I wouldn't worry too much about binder, as we don't have any hint so far that our binder is an issue
May 15 07:21:13 <Herrie|Laptop> Just testing with the new udev rules now
May 15 07:26:24 <Herrie|Laptop> Tofe: No luck either
May 15 07:30:11 <Herrie|Laptop> https://bpaste.net/show/08697035f73e
May 15 08:09:30 <Tofe> Herrie|Laptop: what I remarked on hammerhead is that surfaceflinger (i.e. android's compositor) is crashing often
May 15 08:09:59 <Tofe> and the container is probably never marked as fully booted while this happens
May 15 08:49:08 <Herrie|Laptop> Tofe: How can I check this on Mido?
May 15 08:49:16 <Herrie|Laptop> Because I don't think that happens on mido
May 15 08:57:01 <Tofe> simply with "ps axf" when you have launched the session manager
May 15 08:58:08 <Tofe> look at the processes under the container
May 15 09:00:35 <Herrie|Laptop> Looks OK? https://bpaste.net/show/b80ecb0af9a7
May 15 09:02:34 <Tofe> Herrie|Laptop: no, it looks as if you don't have the container manager started
May 15 09:03:40 <Herrie|Laptop> Well: systemctl seems to think differently? https://bpaste.net/show/fb04b3b78476
May 15 09:06:44 <Herrie|Laptop> Tofe: Seems the bridge0 is created by https://github.com/anbox/anbox/blob/e05ca245d96d2f2f5ba9e80e38a824ea9207ba96/scripts/anbox-bridge.sh?
May 15 09:06:48 <Herrie|Laptop> Not sure we have that?
May 15 09:37:43 <Tofe> Herrie|Laptop: no I didn't put it in the recipe, I'm not sure what points to it
May 15 09:38:08 <Herrie|Laptop> Tofe: Also not sure, but it seems to create that bridge0 that's missing
May 15 09:38:12 <Herrie|Laptop> That might be a missing link somehow
May 15 09:38:17 <Tofe> but we have netdev on our side
May 15 09:38:17 <Herrie|Laptop> It shouldn't but who knows
May 15 09:39:40 <Tofe> https://github.com/Tofee/meta-webos-ports/blob/anbox/meta-luneos/recipes-support/anbox/anbox/anbox-bridge.network and https://github.com/Tofee/meta-webos-ports/blob/anbox/meta-luneos/recipes-support/anbox/anbox/anbox-bridge.netdev theoretically should work
May 15 09:40:57 <Tofe> Herrie|Laptop: the container manager doesn't launch the lxc session until you start a session manager
May 15 09:41:47 <Tofe> I'm a bit puzzled, you seem to have both...
May 15 09:42:37 <Tofe> Maybe it couldn't start the lxc container and we didn't see it ?
May 15 11:23:25 <Herrie|Pre3> Tofe: Me too, but it seems others are having similar issues going by the git bug tracker
May 15 11:23:45 <Herrie|Pre3> Where's morphis when you need him :P ?
May 15 11:23:52 <Herrie|Pre3> morphis_: ping :P
May 15 11:28:26 <Tofe> Herrie|Pre3: on hammerhead it start the lxc container without any problem :p
May 15 12:27:08 <Herrie|Pre3> Tofe: Hmmz well I'm using the same bits as you
May 15 12:27:15 <Herrie|Pre3> So we should have same behavior really
May 15 12:27:29 <Herrie|Pre3> Unless you patched stuff :P
May 15 12:27:39 <Herrie|Pre3> Or my mido image is faulty somehow?
May 15 12:28:31 <Herrie|Pre3> I build from scratch by cloning testing and replacing meta-webos-ports with your anbox version and fixing the android-tools-conf
May 15 12:28:43 <Herrie|Pre3> (which still doesn't seem to get adb working...)
May 15 12:28:59 <Herrie|Pre3> The functionfs check doesn't do what it's supposed to it seems somehow
May 15 12:34:09 <Tofe> I don't have any patching left on my side
May 15 12:34:26 <Tofe> it could be the arm64 image isn't working very well
May 15 12:35:02 <Tofe> I don't know the status of the ubports and/or sailfishos tentatives
May 15 12:46:55 <Herrie|Pre3> Tofe: Well going by Twitter feed, SFOS and/or Mer only recently ventured into aarch64
May 15 12:47:10 <Herrie|Pre3> ubports still does 32 bits going by their branch
May 15 12:47:23 <Herrie|Pre3> I guess I could see if strace gives any hints?
May 15 12:48:53 <Tofe> Herrie|Pre3: I'm discussion that with mal on #anbox, they get it running on xperia x
May 15 12:49:14 <Herrie|Pre3> Tofe: Ah ook
May 15 12:49:49 <Herrie|Pre3> xperia x = which kernel?
May 15 12:50:06 <Herrie|Pre3> Seeing it's Sony it's usually pretty recent I guess
May 15 12:51:01 <Herrie|Pre3> OK there now too so can follow
May 15 12:51:37 <Tofe> not sure for the kernel, 3.10 it seems from hybris adaptation wiki
May 15 12:53:03 <Herrie|Pre3> Well 3.10 needed quite some hacks according to morphis, 3.18 less
May 15 12:53:21 <Herrie|Pre3> Just still not sure what exactly requires this "new" kernel
May 15 12:53:35 <Tofe> looks like noone knows for sure :P
May 15 12:54:08 <Herrie|Pre3> Just trial and error and didn't work on 3.4
May 15 12:54:19 <Herrie|Pre3> There shouldn't be really a lot I'd say
May 15 12:54:33 <Herrie|Pre3> All the snappy stuff that would require things were killed
May 15 12:55:17 <Herrie|Pre3> I don't really see many exotic things that would require fancy kernel
May 15 12:55:31 <Herrie|Pre3> Unless there's some more cgroup stuff that's needed
May 15 12:55:42 <Herrie|Pre3> But we might have patched that for systemd already anyway
May 15 13:09:50 <morphis_> Herrie: pong :-)
May 15 13:11:32 <Herrie> morphis_: Seems I'm getting close to get anbox running but not quite yet...
May 15 13:12:08 <Herrie> morphis_: https://bpaste.net/show/08697035f73e
May 15 13:12:12 <Herrie> This is with 3.18 kernel
May 15 13:12:54 <Herrie> And some more logs: https://bpaste.net/show/9d4b22f797ec and https://bpaste.net/show/fcfffebed5b8
May 15 13:13:06 <Herrie> Any clues as what & how to debug what's the issue?
May 15 13:14:23 <Herrie> This is on Xiaomi Redmi Note 4 (mido) with aarch64
May 15 13:14:37 <Herrie> We stripped the whole snap bits and just put it in simple bitbake recipes ;)
May 15 13:17:51 <Herrie> morphis_: Also you know what causes the 3.10+ kernel requirement?
May 15 13:28:31 <morphis_> LXC does that
May 15 13:28:42 <morphis_> or containerization in general
May 15 13:28:54 <morphis_> Herrie: will come back to you in a bit, about to leave right now
May 15 13:29:10 <Herrie> OK thnx
May 15 13:56:34 <Herrie> Tofe: My Anbox LXC container log: https://bpaste.net/show/20f5a96c207a
May 15 14:00:44 <Tofe> Herrie: looks like "/anbox-init.sh" ends with code 1
May 15 14:01:24 <Herrie> Tofe: Yeah
May 15 14:02:11 <Herrie> Tofe: Well that's a very short script: https://github.com/anbox/anbox/blob/master/scripts/anbox-init.sh
May 15 14:02:48 <Tofe> yes... I don't know how we could get its output
May 15 14:03:43 <Tofe> could it be it's /init that ends with code 1
May 15 14:04:08 <Herrie> Tofe: I'd say likely probably?
May 15 14:04:19 <Herrie> Interesting to see what your log looks like
May 15 14:04:47 <Herrie> It's in /var/lib/anbox/logs/container.log
May 15 14:04:54 <Tofe> I'll have a look tonight
May 15 14:05:08 <Herrie> Tofe: I guess we could modify the script to add some more debugging in it to see where it fails
May 15 14:06:26 <Herrie> At least to see if it's init or something else
May 15 14:06:40 <Tofe> if we can catch its output, yes :)
May 15 14:06:43 <Herrie> Also my lxc-checkconfig showed that I probably need to tweak soem defconfig values
May 15 14:07:02 <Herrie> https://bpaste.net/show/8f652d0ec38b
May 15 14:07:26 <Herrie> Though these are mainly IP and VLAN related not sure how critical they are or should be
May 15 14:12:10 <Tofe> not sure, doesn't look critical
May 15 14:12:38 <Tofe> do you also have a logcat ? ideally with a mark corresponding to when you start session-manager
May 15 14:20:37 <Herrie> Tofe: https://github.com/anbox/anbox/issues/3
May 15 14:20:48 <Herrie> That seems pretty similar to what we have as well
May 15 14:24:31 <Herrie> Tofe: When I run anbox container-manager manually I get: [ 2018-05-15 06:07:00] [container_manager.cpp:98@operator()] boost::filesystem::create_directories: Invalid argument
May 15 14:24:35 <Herrie> Could be we have too new boost?
May 15 14:24:51 <Herrie> I remember seeing bits where boost API changed somewere in some other repos
May 15 14:32:58 <Tofe> no, we'd have it also on hammerhead in that case
May 15 14:33:05 <Tofe> must be specific to mido somehow
May 15 14:33:32 <Tofe> mmmh though I didn't start it manually, so maybe I missed these messages
May 15 14:35:04 <Herrie> Tofe: Seems that's because I didn't specify a data-path
May 15 14:35:08 <Herrie> At least that's how I read the code
May 15 14:35:36 <Herrie> https://github.com/anbox/anbox/blob/e05ca245d96d2f2f5ba9e80e38a824ea9207ba96/src/anbox/cmds/container_manager.cpp#L46
May 15 14:35:40 <Herrie> So false alarm I guess
May 15 14:38:08 <Tofe> ok
May 15 14:41:27 <morphis_> Herrie, Tofe : ok, let me have a look
May 15 14:42:02 <morphis_> the lxc-checkconfig output looks good
May 15 14:43:20 <morphis_> Herrie: there is a default data path being set if not specified
May 15 14:44:12 <morphis_> or better said, there should
May 15 14:45:25 <morphis_> https://github.com/anbox/anbox/blob/master/scripts/container-manager.sh#L38 is how the container manager should be started
May 15 14:45:39 <morphis_> I would start it manually with ANBOX_LOG_LEVEL=debug first and then start the session manager
May 15 14:45:51 <morphis_> both with ANBOX_LOG_LEVEL=debug set
May 15 14:46:04 <morphis_> ah and you want --gles-driver=host
May 15 14:46:21 <Tofe> morphis_: that was one question I had: isn't "host" chosen by default ?
May 15 14:47:23 <Tofe> because I excluded the Glx translator (we have no X11 here) so the Translator one sure won't work; but as I didn't get any error message, I thought it chose host
May 15 14:47:41 <morphis_> if I remember right there is a check for /dev/nvidia*
May 15 14:47:50 <morphis_> and if that is found it selects translator
May 15 14:48:03 <morphis_> there is a change coming soon which will remove the translator entirely
May 15 14:48:22 <Tofe> oh, ok, so no need to bother about a wayland translator then :p
May 15 14:48:45 <morphis_> you don't need one, this is just the GL level
May 15 14:48:55 <morphis_> SDL should do all the wayland work for you
May 15 14:49:11 <morphis_> the ubports guys did a few hacks on top of Anbox https://github.com/ubports/anbox/commits/xenial
May 15 14:50:05 <Tofe> well they've also introduced some Mir dependencies which we don't want; but we do want the touchscreen
May 15 14:50:18 <morphis_> there is a PR from Marius pending
May 15 14:50:33 <morphis_> https://github.com/anbox/anbox/pull/580
May 15 14:51:06 <Tofe> yes, we've seen that one, we'll probably pick it
May 15 14:51:31 <Tofe> Also I did this change, but which might be bugged, I didn't debug it yet https://github.com/Tofee/meta-webos-ports/blob/anbox/meta-luneos/recipes-support/anbox/anbox/0004-Fix-build-with-Wayland-SDL2.patch
May 15 14:52:09 <morphis_> looks fine to me, if we clean that up a bit we can have it upstream
May 15 14:52:28 <Tofe> ok, I'll do a PR once I can confirm it works
May 15 14:52:36 <morphis_> thanks
May 15 14:53:55 <morphis_> Tofe, Herrie: are you guys using the armhf image from build.anbox.io?
May 15 14:54:17 <Tofe> morphis_: yes: https://github.com/Tofee/meta-webos-ports/blob/anbox/meta-luneos/recipes-support/anbox/anbox-data.bb
May 15 14:54:40 <morphis_> great, I would like to keep the images central as there should be no need to customize them
May 15 14:55:00 <Tofe> I have no will whatsoever to touch that part :)
May 15 14:55:12 <morphis_> :-D
May 15 14:55:35 <morphis_> for https://github.com/Tofee/meta-webos-ports/blob/anbox/meta-luneos/recipes-support/anbox/anbox/0001-Fix-dependencies-for-LuneOS.patch we could add a NO_TESTS flag for cmake
May 15 14:56:49 <Tofe> morphis_: yes I saw there is a PR like this somewhere
May 15 14:57:07 <Tofe> or a commit, but I don't remember where
May 15 14:57:31 <Tofe> maybe in the archlinux PR
May 15 14:58:20 <Tofe> morphis_: for now these patches work but having configuration flags is sure a much better way
May 15 14:58:46 <Tofe> It started as a quick hacky project, like it often does :)
May 15 15:03:18 <morphis_> which is absolutely fine :-)
May 15 15:41:53 <Herrie|Pre3> Back from mobile now ;)
May 15 15:42:02 <Herrie|Pre3> Outside with the 2 little ones for a bit
May 15 15:42:29 <Herrie|Pre3> morphis: the lxc-config for the IPv6 bits shouldn't be critical right?
May 15 15:42:46 <Herrie|Pre3> The missing anbox0 bridge also shouldn't be right?
May 15 15:42:51 <Herrie|Pre3> At least to get it started
May 15 15:43:16 <Herrie|Pre3> I understand that without the bridge we probably don't have any networking to outside world.
May 15 15:43:31 <Herrie|Pre3> It seems that anbox-init.sh somehow fails
May 15 15:43:39 <Herrie|Pre3> Just not sure where & how yet
May 15 17:17:14 <morphis_> Herrie: the container.log would be interesting
May 15 17:17:56 <morphis_> and you may want to run container manager with --privileged and unshift the rootfs back to 0 instead of 100000 as the base UID
May 15 17:18:15 <morphis_> I guess UID namespaces are not complete with these old kernels
May 15 17:32:51 <Herrie> morphis_: https://bpaste.net/show/20f5a96c207a
May 15 17:32:57 <Herrie> That's the container.log
May 15 17:35:44 <Tofe> back home
May 15 19:12:01 <Herrie> morphis: I guess this looks a little better? https://bpaste.net/show/feca61378111
May 15 19:14:25 <Herrie> morphis_: That is ^ Container log: https://bpaste.net/show/d0fedba2fbe3
May 15 19:14:32 <Herrie> Still seems init fails inside lxc somehow
May 15 19:15:14 <Tofe> Herrie: your data path looks wrong
May 15 19:15:24 <Tofe> it should be anbox container-manager --daemon --privileged --data-path=/var/lib/anbox
May 15 19:15:57 <Herrie> Tofe: well it doesn't find it there :P
May 15 19:16:26 <Herrie> It seems some mounts didn't happen
May 15 19:16:45 <Herrie> Maybe those happen when you do systemctl start anbox-container-manager instead?
May 15 19:17:38 <Tofe> Here's my ps axf output: https://bpaste.net/show/a913edbc5750
May 15 19:19:25 <Tofe> ... and the associated container.log https://bpaste.net/show/aeffdfc8ee5b
May 15 19:21:53 <Tofe> just surfaceflinger is dying every second or so
May 15 19:23:15 <Herrie> Seems something might have gone awol at my end
May 15 19:23:18 <Herrie> I'll reflash
May 15 19:45:07 <Herrie> Tofe: Seems the issue with 3.4 kernel might be:  lxc 20171223092550.533 ERROR    lxc_start - start.c:preserve_ns:153 - No such file or directory - Kernel does not support attaching to mnt namespaces.
May 15 19:45:07 <Herrie>             lxc 20171223092550.533 INFO     lxc_start - start.c:lxc_spawn:1168 - Failed to preserve namespace for lxc.hook.stop.
May 15 19:45:15 <Herrie> I don't have those
May 15 19:46:15 <Tofe> yes, I saw this but I don't know how fatal that would be
May 15 19:46:37 <Herrie> Tofe: Well it's patched upstream: https://github.com/lxc/lxc/pull/2199
May 15 19:47:14 <Herrie> But only recently
May 15 19:47:21 <Herrie> So I doubt our lxc version has it
May 15 19:47:41 <Tofe> right
May 15 19:47:59 <Tofe> and that would be lxc-3
May 15 19:48:25 <Herrie> Seems that this is a 3.8+ kernel feature
May 15 19:48:33 <Herrie> So that could explain teh 3.10 requirement
May 15 19:48:40 <Herrie> I guess you could try the patch ;)
May 15 19:48:59 <Tofe> I could
May 15 19:49:54 <Herrie> Not sure it'll fix stuff
May 15 19:50:07 <Herrie> But it's the only LXC ERROR in the container.log
May 15 19:50:12 <Herrie> That I don't have
May 15 19:50:22 <Herrie> And might explain the 3.10+ requirement, so worth a try really
May 15 19:50:31 <Herrie> I suspect the code in 2.x is similar
May 15 19:54:44 <Tofe> no the source code has change quite a bit
May 15 19:54:48 <Tofe> changed*
May 15 19:56:26 <Herrie> Tofe: Well this might also work? http://lxc.sourceforge.net/patches/linux/3.0.0/3.0.0-lxc1/patches/0016-ns-proc-Enable-setns-for-support-for-the-pid-namespa.patch
May 15 19:56:37 <Herrie> It's for 3.0 but hammerhead code for 3.4 loosk similar
May 15 19:56:48 <Herrie> https://github.com/Halium/android_kernel_lge_hammerhead/blob/halium-5.1/kernel/pid_namespace.c#L271
May 15 19:57:29 <Herrie> There's a whole set of LXC patches for 3.0 kernels actually: http://lxc.sourceforge.net/patches/linux/3.0.0/3.0.0-lxc1/patches/
May 15 19:57:57 <Herrie> Of which most we wouldn't need since they're in 3.4 already
May 15 19:58:02 <Herrie> Adn you added some already
May 15 19:58:41 <Tofe> I can try the one you pointed out
May 15 19:59:32 <Herrie> Not sure that fixes it but it could
May 15 19:59:45 <Herrie> Since it's only a few lines, should be easy to test on kernel
May 15 20:01:25 <Tofe> oh it's already building
May 15 20:05:57 <Tofe> no that didn't solve the issue: the error is still there
May 15 20:06:42 <Tofe> I'll stop here tonight, we'll see tomorrow :)
May 15 20:06:44 <Tofe> gn8 !
May 15 20:11:04 <Herrie> Tofe: OK gn8!
May 15 20:26:33 <Herrie> Tofe: For tomorrow might be worth to try & bump lxc a bit
May 15 20:26:39 <Herrie> I.e 2.0.9 or 2.1.1
May 15 20:27:34 * JaMa  just finished building with yocto 2.6 and anbox
May 15 20:36:26 <JaMa> hmm I didn't because I've built only luneui-example-image which doesn't include the packagegroup changes from Tofe
**** ENDING LOGGING AT Wed May 16 03:00:11 2018