Umeaboy | Can someone please tell me why this happens? | 01:52 |
---|---|---|
Umeaboy | https://pastebin.ubuntu.com/p/Yd8ZJVgCCT/ | 01:52 |
Umeaboy | I DID change the values and still it says that it's unset. That should NOT happen. | 01:53 |
Umeaboy | I'm off to bed while waiting for a reply. I'll be back in 8 hrs or so. | 02:01 |
T42 | <A_T_R> Hi, Anyone who successfully compiled hybris-17.1 with 4 GB of RAM (host)?? | 06:33 |
Thaodan | why would you take so much pain? | 07:37 |
Thaodan | you could try make -j1 but even then it could be hard | 07:38 |
*** Daaanct12 is now known as Danct12 | 07:49 | |
rinigus | took sailfish-fpd-community as the next for aarch64/aosp10 on sony tama. in this case, fpd is 64 bit and all compiled fine. | 08:15 |
rinigus | it fails to load android.hardware.biometrics.fingerprint@2.1.so which is at /system/lib64/vndk-29/android.hardware.biometrics.fingerprint@2.1.so | 08:15 |
rinigus | after adding the link from /usr/libexec/droid-hybris/system/lib64 all seems to work | 08:16 |
rinigus | here "all" is the -test program | 08:16 |
rinigus | piggz mal ^ | 08:17 |
rinigus | what would be correct way to enable search in /system/lib64/vndk-29/ by hybris linker? | 08:18 |
Thaodan | mount bind might help on this too we use symlinks atm | 08:18 |
Thaodan | see droid-config-sony-seine | 08:18 |
rinigus | Thaodan: let me check | 08:18 |
Thaodan | magisk achieves similar things by using bound mount bind | 08:20 |
rinigus | Thaodan: cannot find it in https://github.com/mer-hybris/droid-config-sony-seine/tree/master/sparse/usr nor in the common sparse-10 | 08:20 |
Thaodan | I meant https://github.com/mer-hybris/droid-config-sony-seine/tree/master/sparse/system | 08:21 |
Thaodan | we link system for some libs so e.g. camera works | 08:21 |
Thaodan | So that could work here to | 08:22 |
rinigus | Thaodan: I see. but fingerprint works for you without it? | 08:22 |
rinigus | we have those ICU bits as well | 08:22 |
rinigus | piggz and mal : with the symlink suggested by Thaodan , fpd works on aarch64/aosp10 based port. | 09:32 |
T42 | <adampigg> Woop | 09:32 |
mal | rinigus: what symlink? | 10:37 |
mal | oh, that vndk thing, so similar that audio needs | 10:37 |
rinigus | mal: yes, in vndk. I wonder whether seine needs the same one or you have some different solution for it | 11:24 |
rinigus | as droid-config-sony-seine does not have this symlink | 11:25 |
mal | rinigus: I should test that on fp3, it has non-working fpd, it used to work on hybris-16.0 | 12:18 |
T42 | <ankaos> PlatformSDK [enes@enes hadk]$ sb2 -t $VENDOR-$DEVICE-$PORT_ARCH -m sdk-install -R zypper dup | 12:48 |
T42 | <ankaos> which ı use? https://paste.ubuntu.com/p/vxw9hBW6MF/ | 12:49 |
T42 | <ankaos> 1/2/3/c which? | 12:56 |
T42 | <b100dian> Context: | 13:10 |
T42 | <b100dian> I noticed that if I LD_PRELOAD the libselinux_stubs.so for vndservicemanager my device crashes some 60-80 seconds after booting with adsp failure bringing down the kernel. | 13:10 |
T42 | <b100dian> If I _dont_ then it is not rebooted, and logcat works and it spams me with `E /vendor/bin/adsprpcd: vendor/qcom/proprietary/commonsys-intf/adsprpc/src/adsprpcd.c:48:adsp daemon will restart after 25ms.` | 13:10 |
T42 | <b100dian> New change that gpt me 'past' this: | 13:10 |
T42 | <b100dian> Copying device's /vendor/ueventd.rc to out/target/product/${device}/root/ueventd.${device}.rc made it not reboot anymore. | 13:10 |
T42 | <b100dian> So the first simpler question would be: how can I make this change permanent (between builds)? | 13:10 |
rinigus | mal for me, 64 bit version worked. Let's see what it would be for fp3 | 13:11 |
T42 | <b100dian> 2/2 Also, even though the device does not reboot now, it kicks me out of telnet after 42 seconds since booting: Here's the console-oops from /sys/fs/pstore https://pastebin.ubuntu.com/p/T5nnHqnwJv/. | 13:14 |
T42 | <b100dian> Can someone tell me when does the usb0 disconnect and more importantly why can't I not get telnet to it anymore :( ? | 13:14 |
T42 | <b100dian> s/when/why | 13:14 |
mal | rinigus: what kind of error did you see in logs? | 13:15 |
mal | @b100dian which android base? | 13:22 |
T42 | <b100dian> mal: lineage 17.1 | 13:22 |
mal | that should not be using libselinux_stubs | 13:22 |
mal | 17.1 is done so that it assumes selinux is enabled | 13:23 |
T42 | <b100dian> However the vndservicemanager would crash otherwise. | 13:24 |
T42 | <b100dian> That doesn't mean I should actually enable selinux in kernel, right? | 13:24 |
mal | how did you disable it? | 13:25 |
T42 | <b100dian> boot cmdline + param enabled in kernel config | 13:26 |
T42 | <b100dian> selinux=0 in cmdline, that is | 13:26 |
mal | and yes, it should be enabled and not disabled at kernel commandline, then some configuration is needed in droid-configs repo | 13:26 |
T42 | <b100dian> you mean like sparse/etc/selinux? | 13:27 |
mal | rinigus: it seems my system is not seeing "android.hardware.biometrics.fingerprint@2.1::IBiometricsFingerprint/default" in manifest | 13:27 |
mal | @b100dian yes | 13:28 |
mal | something like this https://github.com/mlehtima/droid-config-fp3/tree/master/sparse/etc/selinux | 13:28 |
T42 | <b100dian> thanks mal. let me try that | 13:29 |
mal | note that the files are copied from other places | 13:29 |
mal | https://github.com/mer-hybris/droid-config-sony-seine/tree/master/sparse/etc/selinux/minimum if you check here you can see where | 13:30 |
mal | https://github.com/mer-hybris/droid-config-sony-seine/blob/master/sparse/etc/selinux/minimum/contexts/files/file_contexts | 13:30 |
mal | https://github.com/mer-hybris/droid-config-sony-seine/blob/master/sparse/etc/selinux/minimum/policy/policy.30 | 13:30 |
mal | those are the symlinks but we have to copy the actual files | 13:30 |
rinigus | mal: sorry, stepped away from kbd | 13:31 |
rinigus | mal: error was that the sailfish-fpd-community.service crashed with the core dump | 13:31 |
rinigus | then when I ran it in terminal, it just stated that it cannot find the library | 13:32 |
mal | rinigus: I have the service in /vendor/odm/etc/vintf/manifest.xml but still I get the manifest error in logcat and then "Unable to get FP service" | 13:32 |
rinigus | mal: I haven't checked logact, but have andoid daemon running (biometry one), as shown by binder | 13:33 |
rinigus | so, I guess it should be in the manifest | 13:34 |
mal | rinigus: no idea why it's not seeing it in the manifest | 13:36 |
rinigus | mal: no idea, no such errors over here... | 13:38 |
rinigus | but does it work in aosp? | 13:38 |
mal | this is lineage based build | 13:38 |
mal | and I assume it should be working there | 13:38 |
mal | it did work in 16.0 based sailfish also | 13:39 |
rinigus | yeah, for me it was rather straightforward when based on your PR for 64bit support. | 13:40 |
mal | looks like my issue is something in android side maybe | 13:42 |
mal | not sure what should be loading that /vendor/odm/etc/vintf/manifest.xml | 13:42 |
rinigus | mal: yes, sounds like it. maybe test in lineage? | 13:42 |
mal | I hate to reflash it | 13:43 |
rinigus | mal: can you use some backup/restore? twrp? | 13:44 |
rinigus | I presume it is not using lvm | 13:44 |
mal | I have no backup of lineage | 13:44 |
rinigus | yes, but that can be flashed after making backup of sfos. then it is easy to get back. but I can understand the reluctance to do so. | 13:46 |
rinigus | although, if FP is not working, probably someone has complained about it | 13:46 |
rinigus | on xda | 13:46 |
mal | rinigus: it seems this one is not lvm based build for some reason, I did use it for 16.0 base | 13:47 |
mal | probably because hybris-17.1 is hacky and won't fully work without manual hacks | 13:47 |
mal | I mean unless using packaged /system | 13:48 |
mal | I need to figure out how to patch libhybris to make it work | 13:48 |
rinigus | mal: early days :) . for aosp10 it is not too bad, although there are issues here as well. but not with flashing and building | 13:49 |
mal | rinigus: main issue is camera | 13:49 |
rinigus | on lineage? | 13:50 |
mal | on any device not using packaged /system | 13:50 |
rinigus | packaged - as provided by us via RPM and used to build full image, right? since sony tama camera is just fine :) | 13:51 |
mal | yes, without patching libhybris camera won't really work if /system is mounted from lineage or aosp | 13:52 |
rinigus | pity. and probably non-trivial to fix | 13:53 |
mal | rinigus: it might not be too difficult, needs to be able to load ld.config*.txt from /usr/libexec also | 13:56 |
mal | and then some custom ld.config*.txt with suitable search paths | 13:56 |
mal | rinigus: hmm, managed to fix fpd, needed ln -s /vendor/odm /odm | 13:58 |
mal | no idea why | 13:58 |
rinigus | I guess /vendor/odm was not searched... | 14:00 |
mal | rinigus: wondering why the file was there | 14:01 |
rinigus | so, you don't have /odm otherwise? | 14:01 |
mal | nope | 14:02 |
rinigus | I have /vendor/odm, but it has only one file: etc/build.prop | 14:02 |
mal | rinigus: I still have another issue with fingerprint on fp3, it doesn't obey the fingerprint data path set in fpd | 14:03 |
mal | so it tried to write those to wrong path in /data/ and doesn't have permission to do so | 14:03 |
rinigus | ohh, we were warned with piggz that for some devices the path was not supported well. | 14:04 |
rinigus | at least it rings some bells. but don't remember exactly what were the limitations | 14:04 |
mal | rinigus: it writes it to /data/vendor_de/0/fpdata/ not to /data/vendor_de/100000/fpdata/ | 14:05 |
rinigus | let me try to find where it is for me | 14:06 |
rinigus | mal: I have it under /data/system/users/100000/fpdata/user.db | 14:09 |
rinigus | ro.product.first_api_level is 26 for tama | 14:10 |
mal | yeah, so it follows what fpd sets | 14:10 |
rinigus | but why it tries to write under 0 and not 100000 for you? | 14:10 |
mal | rinigus: no idea | 14:12 |
rinigus | mal: looking at the code, we have 100000 hard coded in FPDCommunity::FPDCommunity() | 14:13 |
mal | so that is not really multi-user friendly :) | 14:13 |
mal | rinigus: I see in journal "setActiveGroup to /data/vendor_de/100000/fpdata" | 14:14 |
rinigus | mal: no, not really. there is a limit regarding number of fingers that can be stored and nemo/defaultuser is expected to have more than that already. | 14:14 |
rinigus | but we just postponed multiuser implementation, to make it future us problem | 14:15 |
rinigus | mal: that journal record is correct. do you have such file then? | 14:16 |
mal | yes, but it's empty | 14:16 |
mal | the files still go to /data/vendor_de/0/fpdata/ | 14:16 |
rinigus | should have user.db or something like it for fingerprints. | 14:17 |
mal | rinigus: wondering something somehow check the pid of the process using fingerprint and sets the pid from it | 14:17 |
mal | it has : fids fp2.bin uid | 14:17 |
rinigus | mal: so, fpd daemon works but crashes while you try to add fp? | 14:18 |
rinigus | just trying to figure out what's an issue | 14:18 |
rinigus | mal: re pid - don't know | 14:18 |
mal | rinigus: no, adding fingerprints works fine if I give /data/vendor_de enough permissions so anyone can write to it and then the /data/vendor_de/0/fpdata/ appears with the fingerprints | 14:19 |
mal | for now the only way could be to make fpd systemd service run some chmod in pre start to give that correct permissions | 14:20 |
mal | that would be a quite ugly workaround | 14:20 |
rinigus | mal: here are the permissions that I have over here - https://dpaste.org/GhaK | 14:21 |
mal | yes, I need to make sure system pid/group can create /data/vendor_de/0 | 14:23 |
rinigus | mal: I wonder how is it in lineage ... but it is strange that it pushes it to 0 path. which means that it ignores setActiveGroup call and still sets it to zero | 14:25 |
rinigus | probably setting permissions would be good enough | 14:26 |
mal | rinigus: checking strings of the vendor blob it has "/data/vendor_de/@/fpdata" and I assume that @ is replaced with something | 14:28 |
rinigus | mal: seems so. question is whether it takes that something from env and setActiveGroup is in this respect ignored | 14:30 |
mal | very difficult to say what it does | 14:31 |
mal | rinigus: maybe I should do a new build with those hacks in place, still need to figure out the camera hack | 14:32 |
mal | of course it would be easier to just patch /system | 14:33 |
mal | although doing it in libhybris would remove the need for having the ugly hack in /system | 14:33 |
rinigus | mal: but then we have ugly hack in hybris. | 14:39 |
rinigus | ... which is a big hack anyway :) | 14:40 |
mal | rinigus: is it really ugly hack in libhybris, we have done search path changes in there already | 14:41 |
rinigus | speaking of hybris, I have an odd problem with flatpak support for aarch64/aosp10. when using similar approach as for earlier aosp versions, I cannot enforce hybris linker loading | 14:42 |
rinigus | as a bg: flatpak support is similar to porting itself, it is just that hybris bits have to be injected into flatpak runtime via extension path. in this case, mounted under /usr/lib/aarch64-linux-gnu/GL/host | 14:42 |
rinigus | so, some hybris bits are copied under there (linker, lib64) and I have set some env vars to point there (HYBRIS_EGLPLATFORM_DIR, HYBRIS_LINKER_DIR, HYBRIS_LD_LIBRARY_PATH) | 14:43 |
rinigus | when checking strace, it does load libhardware, but not proceeding to /dev/socket/property_service | 14:43 |
rinigus | strace bits available at https://github.com/sailfishos-flatpak/main/issues/21 | 14:43 |
rinigus | question is: how can I check why linker is not loaded? | 14:43 |
rinigus | (sorry for wall of text, had to describe bg) | 14:44 |
mal | hmm, need to check that a bit later, I need to get some food | 14:45 |
rinigus | mal: would be great to get ideas. thanks a lot and enjoy the food | 14:46 |
rinigus | mal: but no rush, could be pushed further into the future | 14:49 |
rinigus | ... as you are in the middle of FP fixes | 14:49 |
Mister_Magister | mal: got some time? i got g2 with 3.4 update and broken camera | 15:09 |
mal | logs as usual, have you updated anything? | 15:12 |
Mister_Magister | i updated to 3.4 :D https://build.merproject.org/project/show/nemo:testing:hw:motorola:thea every package is updated i think… dcd and dvd submodules too | 15:13 |
Mister_Magister | what logs do you require? | 15:13 |
Mister_Magister | GST_DEBUG the camera? | 15:14 |
rinigus | mal: my issue could be that qt 5.15 loads plugins bit differently. will investigate and come back on flatpak | 15:20 |
Mister_Magister | minimedia is running just fine | 15:23 |
rinigus | no, read it wrong. plugins are probably ok - just hybris linker is not kicking in . but I'll try to dig in | 15:30 |
mal | Mister_Magister: that and maybe logcat, also how exactly does it fail? | 15:32 |
T42 | <A_T_R> make hybris-hal | 16:11 |
T42 | <A_T_R> loghttps://paste.ubuntu.com/p/dYQZkbY4T9/ | 16:11 |
T42 | <ankaos> @A_T_R [make hybris-hal | 16:12 |
T42 | <ankaos> log- https://paste.ubuntu.com/ …], line:139 | 16:12 |
T42 | <A_T_R> https://github.com/SailfishOS-sdm660/device_asus_X00TD/blob/hybris-17.1/recovery/Android.mk#L17 | 16:22 |
T42 | <A_T_R> @A_T_R [https://github.com/SailfishOS-sdm660/device_as …], @ankaos | 16:51 |
Mister_Magister | mal: i start camera, viewfidner works fine, i snap photo, viewfinder stars spazzing out but if u just move hand in front of camera it will be bacc to okay and recording freezes app on stop | 17:15 |
mal | logcat might be useful in that case | 17:27 |
Mister_Magister | okay | 17:58 |
Mister_Magister | mal: https://paste.opensuse.org/422f38e0 logcat | 18:07 |
Mister_Magister | there was like nothing when stopping | 18:07 |
Mister_Magister | gstdebug is better | 18:10 |
Mister_Magister | https://paste.opensuse.org/3d4f302a | 18:10 |
Mister_Magister | useful info | 18:10 |
mal | "FAILED to allocate omx component 'OMX.qcom.video.decoder.avc'" | 18:25 |
mal | that suggests something wrong in adaptation side, maybe some firmware loading failed or some symlink is missing | 18:28 |
mal | Mister_Magister: ^ | 18:33 |
Mister_Magister | oh | 18:33 |
Mister_Magister | hmm | 18:33 |
Mister_Magister | not sure what that could be mal | 18:34 |
mal | check dmesg and journal for any signs of firmware loading | 18:34 |
Mister_Magister | okay | 18:34 |
mal | I assume that older device so maybe also the xml symlink in /etc/ | 18:35 |
Mister_Magister | it is older yeah | 18:36 |
Mister_Magister | mal: well /etc didn't change in 3years | 18:37 |
mal | so it worked before 3.4.0? | 18:38 |
Mister_Magister | yeah | 18:38 |
mal | where is the adaptation repo for that? | 18:38 |
Mister_Magister | got (semi) same device on 3.3 and its fine | 18:38 |
Mister_Magister | https://github.com/VerdandiTeam/droid-config-thea/tree/master/sparse/etc | 18:38 |
T42 | <b100dian> Ok, set selinux=1, backed out LD_PRELOAD-ing selinux_stubs, copied vendor_file_contexts to file_contexts and precompiled_sepolicy to policy.30, now it's radiosilence about *servicemanager crashes, but I assume they just don't start (not in ps -ef) and droid-hal-init fails running atrace etc: https://pastebin.ubuntu.com/p/PcZQQFqzN5/ | 18:42 |
Mister_Magister | mal: can't find anything in dmesg https://paste.opensuse.org/15a32314 | 18:46 |
mal | Mister_Magister: how about journal | 18:47 |
Mister_Magister | waaait [ 1.791679] 1.791679,1] No soundcards found. | 18:48 |
mal | @b100dian do you have the ld.config things (assuming your device needs those) and this line in spec https://github.com/mlehtima/droid-config-fp3/blob/master/rpm/droid-config-fp3.spec#L18 | 18:48 |
Mister_Magister | should it be like that | 18:48 |
mal | Mister_Magister: I think that can be normal | 18:49 |
Mister_Magister | ye audio work | 18:49 |
T42 | <b100dian> I have /usr/libexec/droid-hybris/system/etc/ld.config.29.txt with added `dir.system = /usr/libexec/droid-hybris/system/bin/` and that line in spec yes | 18:50 |
Mister_Magister | mal: lemme correct myself prevous dmesg wasn't from root https://paste.opensuse.org/2014431c | 18:51 |
mal | @b100dian just to be sure, do you have /system/etc/ld.config.29.txt ? | 18:53 |
T42 | <b100dian> mal: just checked, yes | 18:53 |
mal | @b100dian is the ld.config.29.txt mounted over that i.e. do you have the systemd mount service | 18:53 |
T42 | <b100dian> you mean `/dev/sda32 on /system/etc/ld.config.29.txt type f2fs` (line from mount | grep ldconfig) | 18:55 |
Mister_Magister | mal: and here's journal also couldn't find anything https://paste.opensuse.org/ff5cacbe | 18:56 |
Mister_Magister | mal: any idea? | 19:33 |
mal | Mister_Magister: try removing the gst cache from $HOME/.cache/gstreamer-1.0/ and then try to check GST_DEBUG and logcat | 19:36 |
Mister_Magister | mal: gst_debug is same thing https://paste.opensuse.org/ab0d2bb0 same is logcat i tihnk | 19:39 |
Mister_Magister | https://paste.opensuse.org/b774af69 | 19:42 |
Mister_Magister | oh i did'nt send logcat | 19:43 |
mal | can you check the permissions for /dev/video* | 19:45 |
mal | and do you have some symlinks in /dev/video/ | 19:45 |
Mister_Magister | https://paste.opensuse.org/a0cf0447 | 19:46 |
mal | Mister_Magister: just wondering what would happen if you change also those audio groups to camera, although there must have been a reason for the audio change | 19:49 |
Mister_Magister | my 5z has camera not audio but it doesn't have 3.4… | 19:49 |
mal | Mister_Magister: those rules are in https://github.com/VerdandiTeam/droid-config-thea/blob/master/sparse/lib/udev/rules.d/999-droid-vidc.rules | 19:50 |
Mister_Magister | oh | 19:50 |
Mister_Magister | its like 5yo or older file it could be a typo | 19:51 |
mal | maybe change on device the group and then reboot and see if anything changes | 19:51 |
Mister_Magister | yeah | 19:51 |
Mister_Magister | on it | 19:52 |
Mister_Magister | mal: nailed it | 19:54 |
Mister_Magister | that was the issue | 19:54 |
mal | make sure also playback works in gallery | 19:54 |
Mister_Magister | it does | 19:54 |
Mister_Magister | well camera still spazzes out after taking photo | 19:54 |
mal | ok | 19:55 |
mal | but at least it should be better now | 19:55 |
mal | also check logcat when playing in gallery to be sure it uses hw codecs | 19:55 |
Mister_Magister | mal: i mean it goes away pretty easy yeah | 19:57 |
Mister_Magister | https://paste.opensuse.org/e779a6fb heresthe logcat as i'm too dumb to know if it uses hw or not | 19:59 |
Mister_Magister | i think it uses | 19:59 |
mal | that seems to be showing video recording now playback, maybe | 20:04 |
mal | *not playbakc | 20:05 |
Mister_Magister | oh… there could be both but i was focusing on recording yeah | 20:05 |
Mister_Magister | https://paste.opensuse.org/6ff72cd9 | 20:06 |
Mister_Magister | omxcodec so i think hw…? | 20:06 |
Mister_Magister | gotta ask abranson later about how to get capabalities from gstdroid | 20:07 |
mal | I don't see anything about h264 decoding in that | 20:07 |
mal | just in case remove the gst cache like I mentioned earlier and try playback again | 20:08 |
Mister_Magister | okay | 20:08 |
Mister_Magister | https://paste.opensuse.org/40a99379 | 20:09 |
Mister_Magister | there is this | 20:09 |
Mister_Magister | OMX.qcom.video.decoder.avc | 20:09 |
mal | yes, that seems correct now | 20:09 |
Mister_Magister | nice | 20:10 |
Mister_Magister | are we doing something about pictures spazzing out (of which i had nothing in GST_DEBUG) or… | 20:10 |
mal | maybe not now, that is not so critical issue, let's see some other time | 20:11 |
mal | did that happen also before 3.4? | 20:11 |
Mister_Magister | no | 20:11 |
Mister_Magister | mal: okay i'm fine with that pics are semi working, not perfect but working :) i'll build 3.4 image and release to users, and hop on to fixing another phone or porting hybris17 | 20:12 |
Mister_Magister | thanks a ton though! | 20:12 |
mal | hmm, I remember there being some option somewhere for older devices with some issues | 20:13 |
Mister_Magister | oh? | 20:13 |
mal | Mister_Magister: add a new file to /var/lib/environment/nemo/ (similar way of naming as the existing files) and add just this in that file: QTMULTIMEDIA_VIDEO_TEXTURE_BACKEND_NO_RETAIN_TEXTURES=1 | 20:18 |
Mister_Magister | i was almost about to say it doesn't work but i automatically added it to compositor not nemo xd | 20:20 |
mal | not sure if there is difference or not | 20:21 |
Mister_Magister | welp that didn't help | 20:22 |
Mister_Magister | i would say made it worse but no big difference really | 20:22 |
mal | ok, then revert the change if it didn't help | 20:23 |
Mister_Magister | after couple photos now its just flashing wrong frame once | 20:23 |
Mister_Magister | yeah | 20:23 |
Mister_Magister | it doesn't matter much | 20:23 |
Mister_Magister | its like your viewfinder just doesn't work for second | 20:23 |
mal | ok | 20:23 |
Mister_Magister | it doesn't matter much to me as you can still make photos just fine | 20:24 |
Mister_Magister | lets see if advanced camera has same issue | 20:24 |
Mister_Magister | mal: it doesn't :O | 20:25 |
mal | hmm | 20:25 |
Mister_Magister | i take it back | 20:25 |
piggz | cos....its more advanced ;) | 20:25 |
Mister_Magister | it does | 20:25 |
piggz | oh :( | 20:25 |
Mister_Magister | yeah the issue is more advanced indeed it doesn't recover | 20:25 |
Mister_Magister | xd | 20:25 |
Mister_Magister | mal: i'll add that variable back, camera was recovering easier | 20:26 |
piggz | adv-cam will get some nice additions with the gst-droid in 4.2 | 20:26 |
Mister_Magister | yeah advanced camera doesn't recover, camera sometimes has difficulty too | 20:28 |
Mister_Magister | mal: to describe issue, after you make photo, it keeps flashing one frame and then viewfinder | 20:29 |
mal | ok, quite strange, not sure what could be causing that, is that one frame the one it normally shows for a moment after taking a picture i.e. the preview of the photo | 20:32 |
mal | at least jolla-camera does that | 20:32 |
Mister_Magister | mal: no its like if you moved phone sideways super fast, advanced camera jsut stays on that | 20:33 |
Mister_Magister | since i'm at udev i'm gonna remove those annoying udev rules | 20:34 |
Mister_Magister | oh i cant nvm | 20:35 |
piggz | mal: latte needed the systemdcmdline fix | 21:00 |
mal | piggz: what fix? | 21:07 |
piggz | for the old cgroups | 21:09 |
mal | ok | 21:11 |
piggz | mal: may aswell move all servicers to /usr while im updating! | 21:13 |
mal | yes | 21:20 |
T42 | <Алексей %lastname%> Hello, I am porting SFOS (Hybris 17.1) I am getting a black screen. droid-hal-init.service won't start. "can't execute '/ sbin / droid-hal-init': Too many levels of symbolic links" | 21:22 |
T42 | <Алексей %lastname%> I added /usr/libexec/droid-hybris/system/etc/ld.config.29.txt but that didn't solve the problem. Could you help me? | 21:22 |
T42 | <Алексей %lastname%> My journal log https://pastebin.com/dTwuhx5H | 21:22 |
T42 | <Алексей %lastname%> My ld.config.29.txt https://pastebin.com/EeQWWdba | 21:22 |
T42 | <Алексей %lastname%> And why i have error "Too many levels of symbolic links" other servise too? | 21:22 |
piggz | mal: after running migrate_patterns script ... is it ok to remove the left over pattern? | 21:26 |
T42 | <b100dian> hmm I used to have something like system/vendor linked to /vendor and /vendor linked back to system/vendor, check your rootfs for circular links Алексей | 21:27 |
T42 | <elros34> "system-etc-ld.config.29.txt.mount: What= setting is missing. Refusing". Also looks like you disable audit or am I wrong? | 21:29 |
T42 | <Алексей %lastname%> @elros34 ["system-etc-ld.config.29.txt.mount: What= sett …], yes, i have "# CONFIG_AUDIT is not set" My cofig https://pastebin.com/MZKa8qNd | 21:34 |
T42 | <elros34> so revert any selinux/audit changes you did in defconfig | 21:35 |
T42 | <b100dian> speaking of audit, I just enabled back selinux after he pointed out that lineage-17.1 would need it and got radio silence from servicemanager(s) - discussion above. Now I am seeing denied messages for droid-hal-init.. :hmm: https://pastebin.ubuntu.com/p/WYY4BxmSw9/ | 21:38 |
T42 | <b100dian> I enabled audit too, that's why it's so spammy | 21:39 |
T42 | <elros34> I guess audit=0 can be used , I was talking about CONFIG_AUDIT=n which is part of selinux dependency thus disable CONFIG_SELINUX* also | 21:41 |
T42 | <b100dian> Oh, I get it. But in my case I want audit to actually know what's wrong. | 21:43 |
T42 | <b100dian> Other than the fact that audit logs look like an unwanted strace, their ending in permissive=1 means they dont.. apply I guess? | 21:48 |
T42 | <b100dian> *looking | 21:48 |
mal | those audit messages are probably not critical, you can see at least some android side services starting in there, does logcat work? (/usr/libexec/droid-hybris/system/bin/logcat) | 21:51 |
T42 | <b100dian> no, it doesn't, read error | 21:53 |
T42 | <b100dian> I'm assuming for logcat to work, logd must be running. If I try stracing logd, I get `setuid(1036) = -1 EPERM (Operation not permitted)` and "failed to set AID_LOGD" | 21:54 |
T42 | <b100dian> Алексей you could also try strace, maybe you find what circular links you have | 21:55 |
T42 | <b100dian> mal, you're right the messages show droid-hal-init spawning android services, but shouldn't thy also show in process list if they were? | 21:55 |
T42 | <elros34> @b100dian that is only ~35s of log so if droid-hal-init crash later then logcat might not work | 21:57 |
T42 | <Алексей %lastname%> @b100dian [I'm assuming for logcat to work, logd must be …], Ok , I check now | 21:58 |
T42 | <b100dian> @elros34 agreed, and it does crash. here's the before-audit log, sligthly longer (1 minute), which stops right before "droid-hal-init.service: Main process exited, code=exited, status=6/NOTCONFIGURED", and in which you see several services started by droid-hal-init not being able to run. | 22:01 |
T42 | <b100dian> https://pastebin.ubuntu.com/p/PcZQQFqzN5/ | 22:01 |
T42 | <elros34> try to run logcat before droid-hal-init crash then | 22:05 |
mal | @b100dian also verify what files you have in /usr/libexec/droid-hybris/system/etc/init/ | 22:08 |
T42 | <b100dian> I have "disabled_services.rc, disabled_services_device.rc, ecclist.rc, hybris_extras.rc, overrides.rc, servicemanager.rc", where 2 of them are from https://github.com/b100dian/droid-config-tucana/tree/master/sparse/usr/libexec/droid-hybris/system/etc/init (which are from sagit, 17.1 too) | 22:15 |
T42 | <b100dian> @elros34 I couldn't come up with this idea in 1 month, but that actually worked | 22:15 |
T42 | <b100dian> I have 10k+ lines of 10 seconds of logcat before its gone https://pastebin.ubuntu.com/p/DKGnQfd3S4/ | 22:16 |
T42 | <b100dian> I should have dumped to a file, that was my terminal limit because I output to screen.. | 22:17 |
T42 | <elros34> set limit to inifinity | 22:19 |
mal | the journal log linked earlier didn't show the crash of droid-hal-init, would be good to see it | 22:21 |
T42 | <b100dian> I don't think it was more than " Main process exited, code=exited, status=6" but let me post a longer boot without audit | 22:24 |
mal | maybe some messages a bit earlier than that could show something | 22:26 |
T42 | <b100dian> let's see.. beginning of logcat https://pastebin.ubuntu.com/p/5s6FPkPMcs/ | 22:31 |
mal | @elros34 this looks interesting in that log "VendorServiceManager: add_service('display.qservice',1) uid=1000 - PERMISSION DENIED" | 22:34 |
T42 | <b100dian> 5 mins of journalctl https://pastebin.ubuntu.com/p/JrqqwD5Ygd/ | 22:34 |
T42 | <b100dian> sorry, I was selecting and scrolling all this time, should have cleared the backbuffer first, only now I am looking into logcat too:) | 22:35 |
T42 | <b100dian> there's an actual audit denied above that line, with permissive=0, and .. audit disabled in cmdline | 22:36 |
mal | @b100dian can you check "ls -l /dev/mdss_rotator" assuming there is such device | 22:37 |
T42 | <b100dian> there isn't | 22:37 |
mal | ok | 22:37 |
T42 | <elros34> yeah a lot of permission denied | 22:37 |
T42 | <b100dian> should I try enforcing=0 in cmdline ..? | 22:38 |
mal | do you have androidboot.selinux=permissive in kernel commandline, check "cat /proc/cmdline" | 22:40 |
T42 | <b100dian> yes I have it.. | 22:41 |
mal | hmm, strange, the log says "permissive=0" | 22:44 |
mal | I mean in logcat in some selinux messages | 22:44 |
T42 | <elros34> can you change mode to 644 here: https://github.com/b100dian/Xiaomi_Kernel_OpenSource/blob/tucana-hybris-17.1/kernel/power/power.h#L84. no idea whether this is relevant but you have 8 failures because of that in journal | 22:46 |
T42 | <b100dian> Sure, next reboot is 644. The entire /proc/cmdline is https://pastebin.ubuntu.com/p/cbjzpPGqJg/ btw | 22:47 |
T42 | <b100dian> That message was bugging me, you found the cause so fast :+1: | 22:49 |
T42 | <elros34> not so fast, it took me few minutes to figure out it but looks like it's only warning so it shouldnt cause anyy issue | 22:50 |
mal | @b100dian do you have this file /etc/selinux/minimum/contexts/dbus_contexts | 22:51 |
T42 | <b100dian> Not trying to minimize it, it's just that I had hours and din't find the minutes:) | 22:51 |
T42 | <b100dian> yes mal, it's basically a DTD + "<busconfig> | 22:52 |
T42 | <b100dian> <selinux> | 22:52 |
T42 | <b100dian> </selinux> | 22:52 |
T42 | <b100dian> </busconfig> | 22:52 |
T42 | <b100dian> " | 22:52 |
mal | ok | 22:52 |
T42 | <elros34> @b100dian move your /lib/systemd/ dir to /usr/lib/systemd/ | 22:58 |
T42 | <elros34> your mnt-vendor-persist.mount have incorrect symlink | 22:59 |
T42 | <b100dian> oh | 23:00 |
T42 | <b100dian> I had the link fixed locally but I think at one point I moved them to etc because they didn't started.. let me do this proper then, thanks for this | 23:02 |
T42 | <elros34> be aware that your sparse/system directory is most likely ignored because /system_root/system is bind mounted to /system | 23:09 |
bakulinm | hey everyone. I have a (probably) strange question. I need to run sailfish on an arm board, I can run debian for arm64 on it, have source for kernel including drivers, etc. do I still need to port Android on it first and then use HADK guide to run sailfish? | 23:13 |
mal | bakulinm: it is also possible to run sailfish on linux kernel assuming mesa has support for the graphics | 23:15 |
bakulinm | are there any howtos? where do I start? | 23:16 |
mal | bakulinm: you might want to talk to piggz on this channel for example, he has done a native linux port, you might even be able to use the image he made with some changes | 23:16 |
bakulinm | thank you, will try | 23:17 |
mal | bakulinm: just as an example of running sailfish on linux kernel https://twitter.com/calebccff/status/1390698859094061057 that used the image piggz had made for another native linux port | 23:18 |
T42 | <b100dian> new journal https://pastebin.ubuntu.com/p/XYchB8QFgt/ (I'd say not much changed) and logcat https://pastebin.ubuntu.com/p/pbsxD36sPC/ (if mal or @elros34 you're still around) | 23:33 |
T42 | <b100dian> overlay-service is active now, I tried adding back enforcing=0 to cmdline.. | 23:34 |
T42 | <b100dian> mnt-vendor-persist is active now too.. | 23:34 |
mal | @b100dian about the selinux configuration files, did you copy the files from your device or my repo? | 23:38 |
T42 | <b100dian> from a mounted vendor.img - let me check they're the same on the device though | 23:39 |
mal | @elros34 one bad looking thing in the journal "dbus.service: Main process exited, code=exited, status=1/FAILURE" | 23:40 |
T42 | <b100dian> (they are the same) | 23:41 |
T42 | <b100dian> hmm systemctl status says dbus - active | 23:41 |
T42 | <b100dian> I think it restarts.. | 23:42 |
T42 | <ankaos> rpm/dhd/helpers/build_packages.sh —gg output https://paste.ubuntu.com/p/2M6q6ZHDSp/ | 23:44 |
mal | @b100dian I see many dbus related messages in the log, those are probably quite serious issue, but it's quite late here already so more debugging some other time | 23:45 |
T42 | <b100dian> yes, thanks mal, I straced that and I have food for thought https://pastebin.ubuntu.com/p/qXM2rxTKr7/ | 23:46 |
T42 | <elros34> mal that journal is incomplete | 23:46 |
mal | @ankaos https://piggz.co.uk/sailfishos-porters-archive/index.php?log=2021-03-27.txt#line70 | 23:46 |
mal | @elros34 still dbus failure doesn't look good, there were several of those | 23:47 |
T42 | <ankaos> mal: ı use `since --gg didn't compile gst-droid, do it manually with build_packages.sh -b hybris/mw/gst-droid` | 23:48 |
T42 | <ankaos> mal: but I get the same error. | 23:48 |
mal | @ankaos did you downgrade gst-droid | 23:49 |
mal | as mentioned in the link | 23:49 |
T42 | <ankaos> mal: downgrade `gstreamer1.0-plugins-bad gstreamer1.0-plugins-bad-devel gstreamer1.0-plugins-base gstreamer1.0-plugins-base-devel gstreamer1.0-plugins-good` | 23:50 |
T42 | <elros34> mal: sure you are right but who knows when it start to crash in partial log | 23:50 |
mal | @ankaos did you even read the link I gave | 23:50 |
T42 | <b100dian> https://t.me/sailfishos_porters/112897 this would be the link to the telegram mesage @ankaos to downgrade gst-droid | 23:50 |
mal | @ankaos go to hybris/mw/gst-droid and run "git clean -x -f -d" and then "git reset --hard 0.20201104.0" and then check that the common/ folder in there has content, then back to $ANDROID_ROOT and run rpm/dhd/helpers/build_packages.sh -b hybris/mw/gst-droid | 23:50 |
T42 | <ankaos> @mal [go to hybris/mw/gst-droid and run "git clean - …], yea rebuild gst-droid | 23:52 |
T42 | <b100dian> @elros34 I think my journal is not persistent since I've re-built the image, that's why I don't get the first lines. I'll probably post them here tomorrow. Thank you both for the help! | 23:53 |
T42 | <ankaos> mal: https://paste.ubuntu.com/p/Hgxzrh9BtC/ | 23:53 |
T42 | <b100dian> Hmm does git work from PlatformSDK for you @ankaos ? | 23:55 |
T42 | <b100dian> > git: command not found | 23:55 |
T42 | <ankaos> zypper in git | 23:55 |
T42 | <ankaos> @b100dian [> git: command not found], "git submodule init" this is problem | 23:57 |
T42 | <b100dian> maybe issue that yourself from outside platformsdk, but it is worrying that you'd have no git in here | 23:57 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!