Wednesday, 2022-05-11

voidanix[m]@Thaodan so qt5-qpa-hwcomposer-plugin was the only thing missing?08:43
T42<Verevka86> @eugenio_g7 I enable binderfs mount10:54
malabout libhybris, if anyone wants to use vulkan support in Sailfish OS (it was merged to libhybris about two weeks ago) then people also need to update the libhybris used during hybris-hal build and rebuild hybris-hal and then repackage droid-hal, that will add one missing handling of a android side property needed for vulkan supportr11:05
T42<edp_17> mal : Thanks!11:07
maland another note, vulkan is supported in android versions 8 or newer, android 7 has some support but I haven't tested if it actually works there or not, need to find some device with android 7 base11:15
T42<NotKit> mal, is something compiled Android-side?11:22
T42<edp_17> @elros34 : I've managed to collect some logs for the device that has only telnet but no gui and doesn't stuck on boot logo. (Boot logo doesn't appear at all.) The screen stays black and blue let blinks until I plug the cable in, then eventually (few minutes later) it reboots. Android base 17.111:24
T42<edp_17> logs:11:24
T42<edp_17> mounts:
T42<edp_17> init log:
T42<edp_17> journal:
T42<edp_17> dmesg:
T42<edp_17> (I've set persistent journaling.) I've seen that selinux starts in permissive mode.11:25
mal@NotKit hwc2 compat lib11:32
T42<edp_17> @elros34 : I've seen in journal dbus reported selinux support enabled.11:40
T42<edp_17> I use "EGL_PLATFORM=hwcomposer". Should I try "EGL_PLATFORM=fbdev"?11:41
T42<edp_17> Nope, using fbdev didn't help.11:51
malwhat are you trying to fix?11:54
mala bit of an warning for libhybris also, for local builds with latest libhybris I suggest making sure that the qt5-qpa-hwcomposer-plugin source tree is clean (either running make clean or removing the sources or git clean in the source folder) after updating libhybris, there have been cases that qt5-qpa-hwcomposer-plugin has not been properly rebuild after libhybris update which breaks UI12:18
T42<Herrie1982> mal: Shouldn't this be taken care of by your build system properly somehow? At our end in LuneOS using Yocto it would automatically do this.12:20
malusually it does but in this case for some reason it doesn't12:26
malusually changes cause a rebuild if anything important changed12:26
malobs will probably handle this just fine but local builds seems to have issues12:27
T42<edp_17> mal : I am trying to fix this:
T42<edp_17> mal : I have logs: mounts, init log, journal and dmesg12:34
Thaodanvoidanix[m] and others: we found the commit that broke the qt qpa hwcomposer plugin when not rebuild cleanly against new libhybris:
mal@edp_17 has it worked before?12:46
T42<edp_17> mal : No, this is new port I am trying to do.12:49
mal@edp_17 droid-hal-init seems to be failing12:55
T42<edp_17> mal : Thanks. Do you know why and how to fix?12:59
T42<edp_17> mal : the /dev/input/event0 was wrong in config. On device I've updated that to be /dev/input/event3 which is the touchscreen.13:02
malI see this in your log "nohup: can't execute '/sbin/droid-hal-init': No such file or directory" I doubt that file is missing? could it be some selinux issue?13:09
malwhich android base is that?13:09
T42<edp_17> mal : los17.114:23
malhow have you handled selinux?14:48
Thaodanmal @herrie1982: On obs builds are always done in a clean builder while locally you reuse a builder/build target and the source folder. That fact and obs just rebuilds whenever a dependency in the chain changes workarounds the bug in this case.15:02
T42<Herrie1982> Ah OK. Yocto reuses build but when a dependency changed it will rebuild what's required (re @SailfishFreenodeIRCBridgeBot: <Thaodan>mal @herrie...)15:03
T42<Herrie1982> We'll be testing the libhybris bumps shortly too because it seems to fix some issues we worked around with some gl headers15:04
ThaodanWe found that symbol was changed was the source issue and the build system didn't got the header change because it is not aware of the header and thus doesn't treat it as a dependency.15:04
Thaodan@Herrie1982: well local builds are of dependency changes besides missing symbols in the dependency chains (e.g. after an so ver bump) because the don't act on a specific repository but just on the specific package you are building.15:07
T42<Herrie1982> Ah OK, @ Yocto end we build from Git SRCREV usually so if that changes, it will trigger rebuilds for dependencies15:07
T42<edp_17> mal : For selinux I followed hadk-faq for 17.1 base.16:02
mal@Herrie1982 did you have issues because of the old headers or something else?16:14
T42<Herrie1982> Yeah we had some conflicts while trying to use older libhybris with some wayland stuff from LG I think (re @SailfishFreenodeIRCBridgeBot: <mal>@Herrie1982 did...)16:24
T42<elros34> @edp_17 quite obvious from mount output, you do not have mount units generated16:58
T42<edp_17> @elros34 : That's why last time we checked fixup-mountpoints. That was correct, so I expected mount units generated correctly. How should I fix it?17:03
T42<elros34> not really, fixup only take care of correct block device, you do not have any mount units generated. Read: "Make sure that after making hybris-hal, you have fstab* ..." from (or whole section)17:05
T42<edp_17> Well, that is not clear to me what to do. :(17:06
T42<elros34> What is not clear in: "Make sure that after making hybris-hal, you have fstab* and ueventd*rc in out/target/product/$DEVICE/root/"?17:07
T42<edp_17> This is clear. :)17:09
T42<edp_17> As far as I remember didn't have those files there, I manually had to copy them.17:09
T42<elros34> so next: "Read $ANDROID_ROOT/droid-hal-$DEVICE.log to confirm that mount units for /system and other partitions are created"17:09
Thaodanelros34: Does the path used in fixup mount points match the paths in fstab?17:11
T42<edp_17> Thanks, will do once I am at my pc.17:11
T42<elros34> Thaodan: yes17:11
ThaodanDo they have What=/dev/sd* or What=/dev/mmc*?17:13
T42<elros34> do not remember17:14
ThaodanIf so the units are fine.17:15
T42<elros34> ah I misunderstand your latest question. Mount units are most likely not created at all17:17
Thaodanthen please check this line %{dhd_path}/helpers/makefstab --files $fstab_files  --skip auto17:19
ThaodanYou can add bash -x in front of it if you want to debug the script17:19
T42<elros34> checking droid-hal-device.log is easier and I have already told edp17 to do this just few minutes ago17:20
T42<edp_17> Thaodan : Thanks.17:21
T42<b100dian> Hi porters! I have one experience with porting SFOS to a Xiaomi. Now there's a flagship-like that's smaller and boots A11:
T42<b100dian> Then there's Xperia 5 ii which I believe work has started for. Both seem to have promising cameras and overall size. One has IR.. the other has audio jack. I'd like your opinion to help me decide which way to dig sometime this year.19:17
Thaodanb100dian: If you want to join in on work on Xperia 1/5 II you're welcome19:26
ThaodanI have the xperia 1 ii its a great device19:27
ThaodanThe main difference for me as a porter is the approach of porting lineage vs aosp.19:28
ThaodanXperia 1/5 has radio support on stock not sure if that can work for aosp.19:29
malafaik aosp hasn't had fm radio support from sony ever, at least not in a very long time19:32
T42<b100dian> Thaodan: thank you. I know that you started porting the 'mark 2' devices. Am just curious which would be a better choice for the community: a bigger team for a port? More devices available? Will 12x be more available to buy? Just trying to get the pulse, because I'm really torn:)19:39
Thaodanb100dian: Hm I don't know about device availability but Sony devices are easier for community to use because there are no legal issues over reusing stock /vendor.19:51
ThaodanI can just distribute a zip with the image that installs on its own besides Sony blobs that you need to download separately.19:51
ThaodanAlso the port is much closer to official devices which increases the quality and maintainability of the port.19:52
T42<b100dian> interesting.. I am also re-using the /vendor partition on Xiaomi Mi Note 10, but did no think there were any legal issues. Also can be downloaded separately..20:01
ThaodanWell you can't really fix any bugs with the mix of oss vendor and blob vendor that stock is.20:11
ThaodanUsually lineage just imports blob from stock. You can also distribute them separately  but it is harder for the users if they have to extract themselves.20:14

Generated by 2.17.1 by Marius Gedminas - find it at!