Thursday, 2019-01-10

*** ChanServ sets mode: +v T402:25
Benclark006good morning, just wondering if anyone knew some common reasons that the ui isn't working (or some places to look) thanks :)06:54
T4<adampigg> R0kk3rz,.hmm,.not thought about it....07:21
T4<adampigg> @benclark006 test_hwcomposer, check egl plugins are correct etc, all in the faq07:25
T4<adampigg> Logcat, journal07:26
*** ChanServ sets mode: +v T408:19
sledgesPSA: (was already a while in kernel checker)
sledgesr0kk3rz: @adampigg: please create topics for community meetup: guides/decisions on NFC and fingerprints need time from sailors that work on them (better than 'its closed were not ready' answer :))08:50
sledgesPSA: Sailfish OS community #mer-meeting <- /join (or connect via any other applicable means) this channel in 10mins08:50
T4<adampigg> sledges, ok09:07
T4<adampigg> thx for the merge too09:07
T4<adampigg> someone should /connect t4 to mer-meeting09:08
sledgesit's there09:13
krnlyng!seen dcaliste09:41
merbotkrnlyng: dcaliste was last seen in #sailfishos-porters 3 weeks, 1 day, 0 hours, 52 minutes, and 21 seconds ago: <dcaliste> I'm trying to add some debug in /usr/lib/libhybris/ which is created by compiling libhybris with --with-default-egl-platform=wayland. I'm wondering if there is a race condition in finishSwap() call there.09:41
r0kk3rzhes in mer-meeting right now09:45
T4<adampigg> channel link someone?10:04
T4<adampigg> actually, i cant join from depecher anyway ... next time!10:04
dcalistemal: may I ask? In your wayland MR in gitlab, you are building new wayland-egl package that was previously built by mesa. How do you plan to migrate?11:24
dcalisteThere should be a conflict with mesa < x.y.z I guess and an upgrade in mesa packaging?11:24
maldcaliste: hmm, not sure what you mean, I think we use wayland-egl from libhybris11:25
dcalisteI try to install wayland-egl-devel in SDK to compile SDL with the new version and it said that file /usr/lib/ was providing by mesa-llvmpipe-libwayland-egl-9.2.5+git3-1.1.7.jolla.armv7hl11:27
maldcaliste: uninstall mesa-llvmpipe-libwayland-egl ?11:27
malit should then install the wayland version11:28
dcalisteYes, I may do this, but in term of packaging, if package B is providing a file that was previously in package A, it should conflict with package A so package A can be removed.11:29
dcalisteWell, it's not really important, just wondering if it may raise later an issue when Mer MR will be accepted.11:29
maldcaliste: good point, I'll fix the packaging of wayland11:29
malin the MR11:30
maldcaliste: need to figure out when we need mesa for building11:32
dcalisteOtherwise, wayland upgrade and SDL upgrade compile well in SDK, will going to test them on device.11:32
dcalisteBesides, this apply in SDK, the same library is provided on device by libhybris-libwayland-egl indeed, as you mentioned.11:35
dcalisteI'm not packaging expert at all, on how to plan this migration on both SDK and device. May need to ask pvuorella.11:36
malwe don't want to depend on libhybris for sdl so probably either mesa or wayland should be used for, probably wayland is better11:37
dcalisteYes, I fully agree, but in my opinion, it's the migration that requires to be planned, because currently libwayland-egl is provided either by hybris (on device) or mesa (in SDK).11:38
dcalisteAnd to make this a bit more tricky, libhybris-libwayland-egl on device is not only providing libwayland-egl but also the plugin for libhybris called eglplatform_wayland that we need to keep.11:40
malyes, need to check that everything is correctly done11:42
dcalisteyeh, sorry for raising more work to be done...11:43
maldcaliste: no problem11:46
dcalistemal: I agree, the crash can be easily reproduced. Need to investigate now.12:08
maldcaliste: just a moment, I'll find the hack for it12:26
dcalisteNo hurry, it's fine.12:26
maldcaliste: commenting out this line seems to prevent crashing
malthis is the commit that brought the change
maldcaliste: my guess is that either the too old qtwayland is causing the issue or libhybris wayland code has a bug, backtrace showed sometimes things from libhybris12:37
dcalisteThanks for the pointers. I'm testing on device on which I've a modified wayland_window.cpp from libhybris to avoid race conditions in calling dequeue().12:39
dcalisteIt's still crashing though, and gdb backtrace indicate an issue in wayland-client.c from wayland package with a double free.12:39
dcalisteI'm trying to install Valgrind but have network issues from phone at the moment...12:40
dcalisteBesides, I'm also looking at the modifications between wayland 1.6 and 1.15 in wayland-client.c:dispatch_event() which is calling the closure that double free.12:41
maldcaliste: I suspected some race condition in libhybris that would cause double free12:41
dcalisteInterestingly the free of the closure has been moved out of the locked section, for valid reasons, see
maldcaliste: libhybris indeed is missing some things, it doesn't use the existing functions in case cases which are available in newer libhybris12:42
dcalisteSo the bug may already exist before but not triggered thanks to the locked section in dispatch_event().12:42
dcalisteWhat do you mean newer libhybris ? I'm using the master from github, is it not the latest ?12:43
maldcaliste: I mean newer wayland12:46
dcalisteAh, I understand :)12:46
maldcaliste: some stuff that is implemented in libhybris wayland code has been upstreamed12:46
maldcaliste: and the code has changed12:47
maldcaliste: I have some stuff ready for it, need to create PRs12:47
malmaybe tonight12:47
dcalisteFine, I'm continuing reading all this code, trying to figure out a flow somewhere before adding fprintf everywhere... or finally succeed in Valgrind installation.12:48
maldcaliste: I think at least this was in upstream
malneed to check after work12:51
T4<adampigg> ah,/so emu uses mesa? i wonder if browser works ok there13:47
T4<eugenio_g7> Tried, it didn't13:50
T4<eugenio_g7> Not on the latest version of the SDK/emulator though13:50
T4<eugenio_g7> I rememember it working back in 1.x13:51
T4<eugenio_g7> <--- it could be due to this13:56
T4<eugenio_g7> (the TJC question is unrelated)13:57
Thaodanwhen will the plattform sdk be updated to
T4<elros34> have you tried this one
*** ChanServ sets mode: +v T414:55
Thaodan@elros34 I looked inside
Thaodanthe symlink wasn't updated14:59
T4<adampigg> Theodan, you do an irc client? Do you have an i486 build?16:20
r0kk3rzThaodan: usually once the release goes out after EA16:21
Benclark006when i run EGL_PLATFORM=hwcomposer test_hwcomposer i get test_hwcomposer: test_hwcomposer.cpp:179: int main(int, char**): Assertion `err == 0' failed.17:40
Benclark006does anyone know a fix for this?18:42
T4<elros34> I'm trying backported bluetooth drivers on kernel 3.4. When I modprobe hci_smd ift fails because of unknown symbols. Any ideas?
T4<elros34> Benclark006: provide logs: dmesg, journalctl18:43
Benclark006@elros34 journalctl dmesg (phone) dmesg (host)
T4<elros34> so confusing logs... I bet nobody will bother to read them. Reboot, wait a while and type: dmesg, and journalctl —no-page19:12
T4<elros34> amybe add -b0 to journal to get latest boot logs19:13
mal@elros34 you used my code?19:14
T4<elros34> mal: yes, instruction from faq19:16
Benclark006journalctl -b0
malBenclark006: we never need host dmesg anymore19:18
mal@elros34 just to be sure you haven't run make hybris-hal or hybris-boot after make backports? which you never do, make backports is always last19:18
malBenclark006: that is only when you don't have telnet19:19
Benclark006i see19:19
malBenclark006: you have masked droid-hal-init?19:20
Benclark006how do you do that?19:20
malyou shouldn't have it masked if you want graphics19:22
malso run systemctl unmask droid-hal-init on device and reboot19:22
T4<elros34> mal: I was aware of that, first hybris-hal next backports19:22
mal@elros34 are you sure kernel config changes are ok?19:23
Benclark006mal: doesn't look like that fixed it :(19:25
T4<elros34> I checked kernel configs in .config and they looks ok19:30
malBenclark006: new logs19:41
malBenclark006: dmesg and journal19:41
Benclark006sure, give me a sec19:41
malBenclark006: systemctl unmask user@100000 and maybe also systemctl unmask ofono19:48
T4<elros34> not sure about:  Warning: unable to open an initial console. but what mer-kernel-checker prints19:48
Benclark006well the graphics still don't work but hwcomposer works, i think20:06
Benclark006is it supposed to give out any output when it's running?20:06
malBenclark006: test_hwcomposer should show things in display20:16
Benclark006then it just hangs D:20:17
malBenclark006: so more log20:18
*** ChanServ sets mode: +v T420:20
malBenclark006: edit /init.rc and comment out line with init_user020:27
Benclark006ok so now it just bootloops to recovery20:35
Benclark006i'll see if i can very quickly get some logs20:38
Benclark006mal: journalctl:
Nokiusis there away to change to devel repos from testing image ? I don't recall the way21:12
r0kk3rzssu dr the testing repos21:16
r0kk3rzssu ar the devel repos21:16
Nokiusvi /usr/share/ssu/features.d/adaptation-community.ini and vi /usr/share/ssu/features.d/adaptation-community-common.ini is what I do now :-/21:17
Benclark006hey, i'm getting Sailfish jolla-startupwizard-pre-user-session[2589]: libminisf is incompatible or missing. Can not possibly start the SurfaceFlinger service. If you're experiencing troubles with media try updating droidmedia (and/or this plugin).== hwcomposer module ==21:18
Benclark006 in my system journal, is there a fix for this?21:19
malthat is normal, not an error21:22
Benclark006i see21:27
Nokiusr0kk3rz: thanks, cu soon mate :)21:27
Nokiusmi3/mi4 is on SFOS3 :D thanks to Mike22:06
T4<It_sMike> You're welcome :) Thanks to mal too :)22:06
Nokius^ OTA22:07
Nokiusyes mal thanks to you for all the great help!22:07
T4<It_sMike> :) <322:07
mal@It_sMike this is just wrong
malthose files are generated at build time22:48
malalso why did you even fork that?22:48
T4<It_sMike> just testing around :D22:49
T4<It_sMike> I'll remove that22:49
T4<It_sMike> so sorry !22:49
malfor OBS build you can specify the used commit so you can easily build the needed older version22:49
T4<It_sMike> yes I know....22:50
malalso you have a mistake in OBS22:50
T4<It_sMike> Hmmmm .... what ?22:50
malgeoclue package22:51
T4<It_sMike> so ..?22:51
mal this should be named geoclue-providers-hybris22:51
malyou are missing one s22:51
T4<It_sMike> guess this wasn't mine :)22:51
malit cannot build now because it cannot know which spec file to use22:51
malok, I can fix it22:51
T4<It_sMike> yeah !!!22:52
T4<It_sMike> Thanks mal!!!22:52
T4<It_sMike> :)22:52
T4<It_sMike> mal the revision should be set on 38ef0f1d8a56b56db5bfe99fdb88b9a0e684ca6d22:54
malyes, just noticed that22:55
T4<It_sMike> :)22:55
malI copied the wrong one accidentially22:55
T4<It_sMike> no problem :)22:56
T4<It_sMike> finally :D finished building22:58
T4<It_sMike> Thanks a lot mal. and sorry for that mistake ( o-o )22:59
malthe mistake has probably been there a long time but it didn't break until we added android 8 support22:59
T4<It_sMike> O-o23:00
malandroid 8 support added a second spec file to that repo and OBS decides the spec to use based on package name but in that project the package name didn't match either of spec files and therefore failed23:00
T4<It_sMike> Got it ....23:00
maljust so you know in future that package name should usually match the spec file just in case23:01
malespecially when there are more than one spec files23:01
T4<It_sMike> yes... of course :)23:01
malif the repo has only one spec file the package name doesn't matter but it's safer to name it correctly23:04
malat some point the testing projects should be fixed also23:05
T4<It_sMike> I see :) , Should always respect the naming rules !23:05
T4<It_sMike> Hmm... fixing what exactly? the sources?23:06
T4<It_sMike> I don't have any access to my device testing project, should ask saidinesh5 to give me that permission23:10
T4<It_sMike> :)23:11
T4<It_sMike> projects*23:11
T4<It_sMike> mal, I have to go now... will talk tomorrow :)23:17
T4<It_sMike> Thanks again :) good night porters <3 ---{-@23:19

Generated by 2.17.1 by Marius Gedminas - find it at!