Saturday, 2019-12-07

deathmistabout to test memless haptics on OP5, wish me luck (fingers crossed)00:10
deathmisthmm seems the vibrator driver didn't make it into kernel due to OF_SPMI not enabling => QPNP_VIBRATOR_MEMLESS is never built.00:29
deathmistchanging it to just SPMI worked in Kconfig, now then I don't have linux/qpnp/vibrator.h in my kernel tree includes that is used by the driver and simply changing it to qpnp-haptic.h did not work at all :/00:37
deathmistreading about QPNP_HAPTIC: "It uses the android timed-output framework." wonder if I could somehow hack QPNP_VIBRATOR into my kernel for memless haptics to become a possibility00:40
deathmistmain reason I'm even trying is that it seems Qt5 haptics native plugin seems to not work on specific boots now for me on 16.0 (e.g. pulley menus / mce config vibrate command doesn't work) while keyboard taps for example still function normally00:41
T42<elros34> afaik pulley menu use ngf plugin not qt08:59
deathmisthuh, could be the other way around then. the Qt5 component app haptics all also worked when others didn't so that would make sense actually09:06
deathmistfound this in journal: "ngfd: libprofile: session bus connection requested while blocked" that would explain it, any clue? journal logcat
deathmistbtw this is with all memless crap reverted, just using ngfd-plugin-native-vibrator & qt5-feedback-haptics-native-vibrator installed10:30
deathmistwriting msec values to /sys/class/timed_output/vibrator/enable makes the device vibrate. maybe jusa could take a look10:33
T42<elros34> start ngfd in more verbose mode, iirc -vvvvv10:35
deathmistwow so starting it manually now makes it work wth :(  guess I'll edit the service file to get verbose logs10:39
deathmistnew journal (ngf init logs) and followed journal during pulley menu pulls & app close from home screen
deathmisthold on new journal link was some logcat I didn't even look, just a sec10:56
deathmist <-- ngf init logs journal10:58
T42<elros34> maybe this is related to your kernel driver: tfa98xx and q6core_get_service_version: Failed to get service size10:58
deathmiststarting it manually after boot works tho :/11:00
T42<elros34> firmware for tfa98xx is loaded after ngfd is started so that could be related11:06
deathmistyep, I added a "ExecStartPre=/bin/sleep 1" to the service unit and after 4 consecutive reboots it has always worked... what else could I do?11:06
T42<elros34> maybe create udev rule for tfa98xx so ngfd will wait for it but I am not sure it's sane approach. I use such a rule for sensorfw11:10
T42<austroloid> hello, have a problem with media music playing along during notification sounds in loudspeaker i guess its due to output-speaker+wired_headphone where should it be changed12:30
riniguspketo: please register on the store side device "Xperia XZ3 (h9436 / h9436)". for some reason, it was able to access store already. imei (x2) detected correctly, as wifi and bt.17:13
riniguspketo: I would expect that XZ2 dual sim and XZ3 single sim would work as well. but those have not been tested in the wild yet17:14
riniguspketo: corresponding devices are "Xperia XZ2 (h8266 / h8266)" and "Xperia XZ3 (h8416 / h8416)"17:17
rinigusI expect that xz2c will come as well, as one user has already ordered it. but that will have to get working as well before (maybe touchscreen requires some extra tuning)17:18
KALUBECan anyone offer any theories on treble porting? I seem to be stuck dealing with Android Verified Boot17:59
KALUBEI've flashed a stock vbmeta image with --disable-verity --disable-verification which should let any old boot image work without verification, the phone hangs on the splash screen for 5 seconds or so before rebooting18:04
KALUBEThis is my defconfig, mer_verify_kernel_config is happy with it.
T42<elros34> does it use system as root?18:21
KALUBE@elros34 yes18:32
KALUBEI've just been reading into this
KALUBEI'm guessing hybris-boot is building a "legacy" boot image that doesn't work with system as root18:33
T42<elros34> did you reverted commit in kernel which adds skip_initramfs?18:33
T42<elros34> so revert f26d3c76d3764c2d3b542eb8f0aa9ee7dc47233c and c3c2e99fcc3ed20ab49b482bdc102ebb8037ebd6 otherwise hybris-boot script will be not used at all18:35
KALUBEin my kernel?18:36
T42<elros34> yes:
KALUBE@elros34 that fixed it! Thank you so much :)18:54
KALUBEnow we get to fix telnet18:55
T42<elros34> so how log now it takes before it reboots? do you have anything in host dmesg18:57
KALUBE@elros34 stole some defconfig options from Deathmist, have telnet working now getting system/vendor mounted and we're off!19:01
T42<elros34> great, for system you need %define makefstab_skip_entries / in droid-hal-$DEVICE.spec before last %include19:03
KALUBE@elros34 how can I make droid-hal generate systemd mount units for /system and /system_root? Those folders don't exist on my device by default19:19
T42<elros34> it should be done automatically, check droid-hal-$DEVICE*rpm if not show you droid-hal-$DEVICE.log19:22
malshow the fstab file which is somewhere in your android device repo19:26
KALUBE@elros34: here's my droid-hal-device log
KALUBEmal: this is fstab I have added an entry for boot so that fixup-mountpoints would work19:28
T42<elros34> use make modules | grep fstab and then build it if fstab is found19:30
KALUBEthat command has no output19:33
T42<elros34> ok so you copy your fstab* to /out/target/product/$DEVICE/root and build droid-hal again19:34
T42<elros34> you need also custom vendor.mount so add /vendor to makefstab_skip_entries and add this to sparse
T42<elros34> use correct partition for your device and add symlink to it in
KALUBEGot it19:42
KALUBE@elros34 wouldn't duplicating vendor_a as vendor in fixup-mountpoints make it generate the right mount unit?19:46
KALUBEDoes seem to be a bit less clean though19:47
T42<elros34> what do you mean? you should have partitions without _a/_b in fixup. You need custom vendor.mount because it needs to be mounted before /vendor/firmware_mnt /vendor/dsp  and so on19:52
T42<elros34> makefstab helper script just doesn't create correct mount unit19:53
KALUBEah ok thanks19:55
malKALUBE: you need also "%define makefstab_skip_entries /" to droid-hal spec20:01
malmaybe, need to verify that20:01
KALUBEYeah I have that too, thanks20:01
malKALUBE: hmm, have you tried removing it?20:02
malKALUBE: I mean if you don't get the proper system_root mount then maybe you need it20:03
KALUBEI have it added and system.mount and system_root.mount both look good20:04
malKALUBE: you really shouldn't need to add those at all, both are added automatically if you have correct things in droid-hal spec, in this case makefstab_skip_entries / probably prevents the automation from working20:07
mallike shown in the link I just gave20:08
malnow that I think of it "%define makefstab_skip_entries /" should only be used in official device which have rpm packaged android system libs20:08
KALUBEI haven't defined them manually they are being built automatically and end up in droid-hal-device.rpm20:08
malKALUBE: so what is then the problem20:09
piggzmal: porting a large qqc2 app to sailfish is quite an undertaking!20:10
malpiggz: heh20:10
KALUBENo problem, I removed / from %define makefstab_skip_entries and it still generated the mount units so all is well20:11
malok20:11 -i fails with: repo problem: nothing provides droid-hal-enchilada-detritus needed by pattern:jolla-hw-adaptation-enchilada-1-1.noarch20:12
vknechtmal, what about suzu with pie and droid-system, should it skip_entries  for / or /system (like official port), or neither ? currently fighting with that... with aosp8 base I did use neither, but didn't have droid-system (it used aosp /system)20:12
malvknecht: I assume that doesn't use / for system mount in fstab? or does it?20:13
vknechtit uses /system, afaics
malso if you use rpm packaged system partition then you of course skip the mount20:15
vknechtok thanks20:18
KALUBElipstick is failing with EGL Error : Could not create the egl surface: error = 0x300122:11
KALUBEAny ideas22:13
KALUBE?* I have also tried test_hwcomposer, I will upload the log for it too22:14
KALUBEThis is the result of strace test_hwcomposer:
malwhich android base?22:16
malmaybe you need to try test_hwc2 if the device is very new?22:16
maldid you use it correctly "EGL_PLATFORM=hwcomposer test_hwcomposer" or same with test_hwc222:17
KALUBEyes I did, will try test_hwc2 thanks22:21
KALUBEtest_hwc2: test_hwc2.cpp:288: int main(): Assertion `eglGetError() == EGL_SUCCESS' failed.22:21
KALUBEI forgot to add WANT_ADRENO_QUIRKS could that be the issue?22:21
malnot at this level yet22:22
KALUBEhmmm ok22:22
malso maybe something is wrong with the android side, did you do things according to the android base specific instructions in faq?22:22
malwhat ever android base you use22:22
KALUBEI'm using LOS16 base, this is my manifest
KALUBEsystem and vendor are both mounted properly22:25
KALUBESorry I just pushed, that local_manifests repo was out of date22:27
maldid you do what faq told to do when porting 16 base?22:41
KALUBEyeah, I have these:
malwasn't there already a cheeseburger port?22:42
deathmistmal: he's doing OnePlus 6 (enchilada)22:43
KALUBEthe hadk says to copy the files I linked22:43
maldeathmist: maybe you can help him?22:44
deathmistmal: we've been trying ever since he got telnet working ;)  we got ld.config mount unit, DHI starting and stuff in fstab mounted including proper /system_root and /system22:51
KALUBEminimer outputs Cannot find EGLConfig, returning null config23:22
KALUBEand then the same EGL Error : Could not create the egl surface: error = 0x300123:22
T42<elros34> did you mask user@100000.service and reboot before trying minimer or test_hwc2? you should also check /usr/libexec/droid-hybris/system/bin/logcat to see whta fails23:30
KALUBE@elros34 logcat is full of this:
KALUBEI will try masking user and test hwcomposer23:37
maldeathmist: KALUBE what do logs say?23:37
T42<elros34> @KALUBE: reboot and and try again so logcat will contains also early boot messages23:42
malalso dmesg23:42
KALUBEdmesg: logcat:
KALUBEthe permissions of /dev/ion were wrong, and my ueventd.rc file is nowhere to be found in my source tree (except in out/target/product which is very odd) I modified the udev rules based on IRC logs which removed that error from logcat23:53

Generated by 2.17.1 by Marius Gedminas - find it at!