Friday, 2021-02-26

piggzmal: that one line?08:18
malpiggz: you messed up, you probably had all patches applied when you commited changes08:49
malpiggz: in the latest commit08:49
piggzfeck08:50
piggzyes, i had -a ... i need to stop using -a08:50
piggzone moment08:50
piggzmal: check now pls08:55
malpiggz: the issue I complained about i.e. first part of changes in gfx/gl/GLLibraryEGL.h is still there08:58
piggzmal: ok, i see it .. how do i remove that without breaking the diff, and without re-creating it fully?09:00
malpiggz: just manually remove that part of the patch file09:04
piggzdoes that not breakthe markers above?09:04
malno09:04
malpiggz: just remove lines 116-124 from the diff09:05
piggzah, the whole hunk09:06
T42<zinstack625> @swanux, is your device system_on_root?09:11
T42<zinstack625> Or whatever it's called09:11
piggzmal: force pushed again09:16
T42<zinstack625> My device bootloops the same way as swanux's btw, any ideas on what might be causing this?09:19
T42<zinstack625> It is as if it's skipping initramfs all anew, but I know for sure it doesn't, because I patched the kernel for that09:20
malif you used hybris-17.1 did you run "export TEMPORARY_DISABLE_PATH_RESTRICTIONS=true" before building hybris-hal09:33
T42<zinstack625> Not yet, but would like to know what that does09:36
T42<zinstack625> Like, is that a new requirement for building, can I somehow adapt device sources for avoiding to export that every time09:38
T42<zinstack625> Or is it just a workaround for a not yet implemented feature09:38
malhybris-17.1 brought some new limitations and we use command which is not allowed during build so we need disable the restrictions09:50
T42<zinstack625> Ah, I see, was wondering whether those errors were just a warning09:51
piggzmal: is 4.0.1 on obs yet?09:51
mal@zinstack625 most are but the cpio error is critical09:55
malit prevent creating a working kernel initrd09:55
malpiggz: I couldn't see it there yesterday09:55
T42<zinstack625> Also, what's the deal with aarch64, there seems to be no repo for community adaption commons?09:56
malyes, aarch64 is not supported in community OBS at the moment09:57
malthe repo needs to be removed manually09:57
deathmist_@zinstack625 @swanux for me the cause of new reboots on 17.1 was a few TTY related commits, see e.g. https://irclogs.sailfishos.org/logs/%23sailfishos-porters/%23sailfishos-porters.2021-02-17.log.html#t2021-02-17T13:18:5209:58
deathmist_find them in "git log" and then "git revert" by commit hashes09:59
T42<edp_17> @elros34 [@edp_17 https://github.com/libhybris/libhybris …], The TARGET_NEEDS_PLATFORM_TEXT_RELOCATIONS is set to true in BoardCommonConfig.mk. Seemingly this was not enough. Although, I don't see why commenting out that line will help to find the location of  AKM  library.10:31
rinigusmal: with the newer binder, audiosystem-passthrough starts. but I get waiting for media.camera messages as a result. something I was not getting before without MINIMEDIA_AUDIOPOLICYSERVICE_ENABLE:=110:36
rinigusand camera doesn't start either10:37
rinigusI wounder whether newer gbinder and without that audioservice would work.10:38
rinigusbut it would have to wait till later today...10:39
rinigusre OBS: tried to enable something like https://build.opensuse.org/projects/Fedora:32/meta for SFOS in opensuse OBS. that failed as it needs admin rights10:40
rinigusso, no simple workaround and we have to get 4.0.1 on sfos obs. same goes for aarch64 support10:41
rinigusmal: how is aarch64 implemented in Jolla OBS (non-public)? is it possible to use the same config in public one?10:42
malrinigus: which gbinder config do you have now?10:46
rinigusthe default one, with api level 2910:46
rinigusdrop to 28?10:47
rinigusmal ^10:47
malrinigus: is mini* working well?10:47
mal29 should be correct10:47
rinigusmal: mini* started; sound if playing10:47
rinigusis playing10:47
malanything other in logcat10:47
riniguswill look in the evening, away from device. just wanted to give an update10:48
rinigusmal ^10:49
T42<edp_17> Hi all, I am trying to build hybris-hal and getting this error: https://paste.ubuntu.com/p/HMn5vjSf6M/11:22
T42<edp_17> Any idea what the problem is and how to fix?11:23
T42<elros34> @edp_17 even if it's defined in *.mk file I do not see how libhybris could know about that so defining it in libhybris might be required11:29
T42<edp_17> @elros34 [@edp_17 even if it's defined in *.mk file I do …], How? Simply hardcoding the location of that libakm.so?11:33
T42<elros34> libakm.so is found, you can see it in sensorfwd log you provided but it's not loaded.11:34
T42<elros34> you can see 2 errors from linker in logs: has text relocations11:38
T42<edp_17> @elros34 : I see, so if I comment out the line that drops that error, it will load the libakm.so. I give it a try.11:54
T42<edp_17> I guess need to comment out the 'return false' line too. right?11:56
T42<edp_17> Well, I don't as I want it to return as no error.11:56
T42<edp_17> @elros34 : I commented out that line (4218) rebuilt libhybris and installed al packages on device. After reboot the same error is still there in the log "AkmSensor: unable to load AKM Library, dlopen failed"12:13
T42<elros34> 4219: you don't want that function to return false so obviously comment out that return false. Whether it prints error or not is irrelevant12:29
rinigusmal: just to mention as it could be missed in details: mini* service required symlinks from /system/lib to /usr/libexec/.../lib for few 32 bit libs. issue was discussed 1-2 days ago12:31
T42<zinstack625> Someone mind checking my logs, I can't seem to logcat and android half is broken: https://pastebin.com/kmpsYMu712:32
T42<zinstack625> Here's full dmesg, if necessary: https://pastebin.com/tr3RCgFa12:34
malrinigus: ah, so you lost the changes12:35
rinigusmal: not really lost - the changes were the same before and I restored them after reflash. this is in the case if I use latest libgbinder and MINIMEDIA_AUDIOPOLICYSERVICE_ENABLE:=112:37
rinigusmal: to recap, that allows mini* services to start and audiopassthrough. but media.camera is waited for. when I get back from office, I will check logcat if there is something more regarding it12:38
T42<zinstack625> And, kconfig, might be interesting: https://pastebin.com/9upgeKym12:38
mal@zinstack625 is that hybris-17.1 based build? it's not very simple yet to get those working12:40
T42<zinstack625> mal: oh, ok then, if it's out of my league12:41
T42<zinstack625> Was very tempting, considering los 16 was discontinued more than a year ago for my device12:42
malwell it's going to be get easier soon once I manage to write some documentation hopefully this weekend12:42
T42<zinstack625> Awesome12:45
swanuxJust a quick question: Can I use the same tooling & target (everything else is different ofc) to build for 16.0 base which I used to build for 17.1?12:50
T42<zinstack625> swanux: I can't see why not12:51
T42<zinstack625> And by the way, about that installation issue you were having, is your device system_on_root?12:52
maljust make sure you clear any packages from droid-local-repo before building new adaptation packages with build_packages.sh12:52
swanuxT42 I thought the same, I just wasn't sure if I understood its role completely. Thanks for the confirmation.12:54
swanuxI guess yes, as it's required for android 10 (17.1)12:54
swanuxmal That shouldn't be a problem, I use an other $ANDROID_ROOT for it12:55
malswanux: then you need to make sure you change the url of the droid-local-repo in the target to point to the new location12:56
swanuxmal I checked and it builds the packages to the proper location (even though I don't remember changing anything in the target)12:58
T42<zinstack625> swanux: the bunzip2 is broken on twrp for system_on_root devices, so you'll have to do some ugliness like LD_PRELOAD=/system_root/system/lib64/libbz.so PATH=$PATH:/system_root/system/bin/ before the tar in an installer13:01
T42<zinstack625> Also, T42 is not a person, it's a telegram bridge13:01
swanuxOh, good to know, thanks! I went with a workaround, but this one seems better.13:03
swanux(I know, just couldn't mention you, so I just indicated the bridge)13:03
T42<zinstack625> irc-like mentioning works fine btw13:04
swanux@zinstack625 thanks13:06
T42<edp_17> @elros34 [4219: you don't want that function to return f …], Oh, okay. I thought needed to keep the return false. Knowing, that I do not need that, it is no wonder I got the same error in the log. 😃 I try again.13:08
T42<edp_17> @elros34 : All sensors (except the lightsensor) do works now. Thanks for your help!13:23
swanux@zinstack625 How does exactly that bunzip2 thing work? How to proceed after declaring the variables?13:29
T42<zinstack625> swanux: I'm not sure how to do it nicely using default installer, but you can write that in hybris/hybris-boot/updater-unpack.sh13:32
T42<zinstack625> Personally, I'm using the custom installer: https://github.com/sailfishos-oneplus5/hybris-installer13:33
T42<zinstack625> My modified version is https://github.com/sailfish-motog7/hybris-installer, can use as a reference13:34
swanux@zinstack625 Thank you!13:44
T42<elros34> @zinstack625 wouldn't be easier to mount bind /system_root/system to /system same as in sailfish instead some LD_PRELOAD and PATH changes?13:44
T42<zinstack625> @elros34, actually twrp on those devices provides its own /system, which doesn't contain bzip, and bind mounting /system_root/system to /system kills twrp, like you can't ls anymore kills13:46
T42<zinstack625> At least on my device that is, but seems to be related to system_on_root13:47
T42<elros34> ah now I get it, thanks.13:48
T42<PizzaG> Hmm, been following for a bit, sounds like a lot of work to get this going on a device.  Was wondering if it would work on a moto e 2020 with msm895314:00
T42<adampigg> @PizzaG [Hmm, been following for a bit, sounds like a l …], If there is a LOS port, you should be good to atleast make a start14:13
T42<PizzaG> @adampigg [If there is a LOS port, you should be good to …], ya there is a device tree and I've built 20-30 roms for it so far14:30
T42<PizzaG> android 10 and 11, it shipped with 1014:30
T42<PizzaG> and dynamic partition14:31
T42<PizzaG> is there a general build guide somewhere i could take a look at?14:32
T42<PizzaG> does SailfishOS supporty dynamic partitions?14:35
piggzmal: abranson: https://git.sailfishos.org/mer-core/qtmozembed/merge_requests/7914:38
piggzmatching part14:38
malpiggz: remove the .o file from the MR14:47
swanuxSo. I rebuilt everything on 16.0 base and flashed everything (manually at the end) and now it's a bit different, but not much:14:51
swanuxNow it stays on vendor logo for a long time (~1-2mins), then a short black screen (~1s), and vendor logo again. No telnet according to dmesg.14:51
deathmist_swanux: check your kernel RNDIS configuration; for example I had to set CONFIG_USB_CONFIGFS_RNDIS=y & CONFIG_USB_CONFIGFS_F_GSI=n before it worked14:52
piggzmal: good nitpick, done14:55
swanuxdeathmist_: I checked, and the first one didn't exist for me, and the second one was y. Should I try with your modifications?15:04
T42wiktorek140 was added by: wiktorek14015:18
swanuxOkay. After modifying the kernel config accordingly and rebuilding hybris-boot.img here's the situation:15:35
swanuxOn my device I see the lineage os boot animation (in infinite loop), but I can turn on and off the display with power button in the meantime (battery led also works)15:35
swanuxRegarding telnet, here's the relevant dmesg output (https://pastebin.com/DtZUeqeX)15:35
malswanux: that suggests it had telnet for 5 seconds15:39
T42swanux was added by: swanux15:40
swanuxmal: That would explain the notifications trying to connect to it. Do you have any ideas?15:41
malyou probably need to somehow disable some service to prevent it from rebooting15:42
swanuxThe thing is it didn't reboot. It looks like this:15:44
swanuxVendor logo (that's that 5s when I got telnet) > battery led starts working > lineage boot animation15:44
malhmm, the animation should not be there, did you do the additional things mentioned in faq?15:45
malhttps://github.com/mer-hybris/hadk-faq/blob/master/README.rst#hybris-16-015:45
swanuxFrom here only the first one (cloned repo & patched)15:46
malthe second part about .rc files is very important15:47
swanuxI'll try it then, thank you. Only hybris-boot needs rebuild again, right?15:47
malno, you need to have those in rootfs15:48
malyou can manually add those via recovery15:48
swanuxMy bad... (asked before read...) So like adb push to proper location? (where the extracted rootfs is)15:49
malyes15:51
malin recovery the sailfish rootfs is at /data/.stowaways/sailfishos/15:51
swanuxOk, thanks I'll try it now15:52
swanuxrc files copied. Here's the new dmesg (https://pastebin.com/3fTq6vNw)15:58
swanuxNow it stays at vendor logo (definitely less weird), battery led works15:58
malswanux: you are sure it doesn't reboot?16:03
swanuxmal: I think so. It's a constant vendor screen all the time. When the switch happend (in dmesg, new name, new service, previous disconnects) that's when the led starts to work (so I guess that's when other services start)16:06
swanuxAnd at that point chnges this:16:06
swanux[30306.708748] usb 3-10: Product: init-debug in real rootfs16:06
swanux[30306.708749] usb 3-10: Manufacturer: Mer Boat Loader16:06
swanux[30306.708750] usb 3-10: SerialNumber: Mer16:06
swanuxTo this:16:06
swanux[30311.829459] usb 3-10: Product: Mi 816:06
swanux[30311.829461] usb 3-10: Manufacturer: Xiaomi16:06
swanux[30311.829464] usb 3-10: SerialNumber: 3852b33416:06
swanuxWoah. Rebooted (with poweroff before that) and now I'm back at the lineage situation16:09
malswanux: could you enable persistent journal log in /etc/systemd/journald.conf, there is the Storage setting, change it to Storage=persistent, after rebooting one after the change go back to recovery and try to look at the logs16:09
malhmm16:09
swanuxOk16:09
swanuxWhere should I look for the logs?16:14
swanuxNvm, found it.16:23
swanuxmal: Which parts are needed?16:23
mallook for anything related to usb16:24
swanuxmal: From the 3k line file these are the lines that seemed appropriate.16:36
T42Rotoffset %lastname% was added by: Rotoffset %lastname%16:38
T42<Verevka86> https://github.com/LineageOS/android_frameworks_av/commit/b3145b99959f4fc24c5fc43bac022e56026e062b16:52
T42<Verevka86> this breaks the hybris-16.0 compilation16:52
malwhat error do you see?16:53
T42<Verevka86> @mal [what error do you see?], https://paste.ubuntu.com/p/cnVRHPQZdq/16:55
swanuxmal Forgot to send the link.... https://pastebin.com/thYN0nSS16:58
rinigusmal: nothing really jumps out in the logcat https://dpaste.org/2G5G17:55
malswanux: try adding file /etc/usb-moded/usb-moded-configfs.ini with content https://pastebin.com/rh5aTfEF18:08
swanuxmal: I did it, nothing changed (at least on the surface).18:23
swanuxHowever some usb problems are no more there (commented out, here: https://pastebin.com/eQQEK4sM)18:23
swanuxHere's the whole file (just in case): https://pastebin.com/4GgVRfP918:23
malswanux: did you reboot the device using power button during the boot you got the log from?18:45
swanuxmal: I deleted the previous journal in TWRP, then pushed to config file (which you mentioned), and then rebooted to system (from TWRP menu), then waiting for the boot and at the end went back to the recovery to pull the new logs.18:49
rinigusmal: when using droidmedia without MINIMEDIA_AUDIOPOLICYSERVICE_ENABLE, camera works until switching it off after shooting video or switching from video to photo. logcat https://dpaste.org/Ns4W19:41
rinigusending on that ServiceManager: Waiting for service 'media.audio_policy'19:42
T42<elros34> @swanux: you need to disable selinux via bootparam for hybris-16. Unfortunately mer-kernel-check suggest something different so this is common issue for new porters. If you will still have no rndis then you can try masking usb-moded.service19:42
riniguskinda catch 22: either camera works till getting out of video mode; or media.audio_policy works but not camera19:42
malrinigus: what means minimedia is not working as it should be19:44
malrinigus: how did you rebuild droidmedia after changing env.mk?19:44
malrinigus: ah, sorry misread what you wrote19:45
rinigusmal: not the desire, but maybe I should grab logcat in AOSP while switching camera modes19:46
malrinigus: that issue suggests droidmedia might be starting things in wrong order, it starts audiopolicy before camera and if that somehow hangs then camera service later is not started19:47
riniguswhen grepping one logcat that I have from AOSP, there is no media.audio nor media.camera. if it is of any importance19:48
rinigusmal: wrong order - that is when I have MINIMEDIA_AUDIOPOLICYSERVICE_ENABLE=1?19:49
malrinigus: could you try moving AudioPolicyService::instantiate(); later in minimedia.cpp19:49
maland enabling audiopolicy again19:49
rinigusmal: will try. but it all feels a bit strange. as /system/lib did not have multiple audiopolicy libs that were only in /usr/libexec/droid../lib, it suggests that we are using a path that is not used normally19:51
rinigusor is it known and is the same on all devices?19:51
rinigusmal ^19:51
malrinigus: hmm19:56
swanux@elros34 You mean like this?20:10
swanuxCONFIG_AUDIT=y20:10
swanuxSELINUX_BOOTPARAM=y20:10
swanuxSECURITY_SELINUX_BOOTPARAM_VALUE=020:10
swanuxCONFIG_SECURITY_SELINUX=y20:10
swanuxCONFIG_SECURITY_SELINUX_BOOTPARAM=y20:10
swanuxCONFIG_SECURITY_SELINUX_BOOTPARAM_VALUE=020:10
malrinigus: one thing I have always wondered why only community ports need that audio_policy20:12
T42<elros34> swanux: yes, those CONFIG_*20:13
rinigusmal: note that this is also sony device with the port closely following the official one. so, I would expect it to work similar to the other sony's.20:13
malrinigus: yes, that makes it even more strange20:13
rinigusmal: but on seine, there is no such issue, right?20:14
rinigusofc, original release for xz2 was with older android20:14
malrinigus: hmm, only relevant difference is how droidmedia is built, in official builds hybris-hal build is separate and the droidmedia stuff is not in it20:14
rinigusandroid 8, when looking it up20:14
malrinigus: one idea, build hybris-hal, then package it, then build droidmedia and package it20:15
rinigusmal: sure, I can try. that is without the audio_policy service, right?20:16
malyes20:16
rinigusmal: OK, as I am restoring version with audio_policy from backup and have droidmedia ready for it, I will test moving AudioPolicyService::instantiate() towards the end of main() first. and later deal with the second approach, probably tomorrow in the first half of a day20:18
swanux@elros34 Thanks, I'll try to rebuild hybris-boot then with those beginning with CONFIG_* from there.20:18
swanuxShould I leave the added /etc/usb-moded/usb-moded-configfs.ini there?20:18
T42<elros34> sure20:20
rinigusmal: moving AudioPolicyService::instantiate(); to the end helped it. camera works as it should. diff at https://dpaste.org/kLRj#20:28
malrinigus: interesting20:29
rinigusmal: I will try the second idea with separate hybris-hal and droidmedia packaging. I guess I wipe "out",  remove droidmedia, build + package hal; add droidmedia and build + package it20:30
malrinigus: I can append that to the PR I already have20:30
rinigusmal: yes, sure. but no idea if it disturbs anything else20:30
malrinigus: you can have droidmedia sources and everything, just run make hybris-hal before packaging droid-hal20:30
rinigus... and no need to keep old code as I commented it out :)20:30
malrinigus: so simple way is remove out/, make hybris-hal, build_packages.sh -d20:31
rinigusmal: understood. so no need to remove droidmedia sources20:32
T42<zinstack625> Tried to build sfos 4.0.1 with hybris 16.0 and this started happening: https://pastebin.com/jQM0hvJQ20:44
T42<zinstack625> Any idea how to fix, anyone?20:45
T42<swanux> @elros34 I flashed the new image, and something happened for sure, as now I'm not seeing any lineage logo at all (progress I guess?), just the vendor logo + battery light. However still no telnet. (but I got a connection now through Ethernet, which is new, and I also got an ip so it's a working host). But same as before, in dmeag I get20:49
T42 the mer stuff for around 5s, then the Google - xiaomi stuff replaces it.20:49
T42<elros34> @swanux do in twrp: touch /data/.stowaways/sailfishos/init_enter_debug2. This will stop botting at /init-debug script and  should allow you to telnet/ease debugging further20:52
swanux@elros34 It really stopped now (no battery light), but can't connect to telnet. I get this (iSerial                 3 Mer Debug telnet on port 2323 on rndis0 192.168.2.15 - also running udhcpd) but I can't establish a stable Ethernet connection.20:58
T42<elros34> are you trying at 2323 port? Does ifconfig shows new device with ip assigned?20:59
T42<zinstack625> Device might be dropping its interface and thus, computer's ip21:00
swanux@elros34 Yes 2323, and ifconfig only shows the usb interface but no ip. Probably what @zinstack625 says21:01
T42<zinstack625> "while (true) do ip addr add 192.168.1.9/24 dev *your_device_interface_name*; done" should do the trick21:01
T42<zinstack625> can throw a sleep in there if you want to spare your cpu from ineitable death21:02
swanuxwhile (true) do ip addr add 192.168.1.9/24 dev usb0; done Like this? I get these: RTNETLINK answers: Operation not permitted21:03
T42<zinstack625> oh, do it as root21:04
T42<zinstack625> it will start spewing that the interface already has that ip, occasionally that interface doesn't exist, which is fine21:04
swanuxRTNETLINK answers: File exists21:05
T42<zinstack625> Exactly that21:05
T42<zinstack625> Ignore and work in a different terminal21:05
T42<zinstack625> if your video is unbearably slow, you might consider silencing that by just > /dev/null21:06
T42<elros34> I think you don't need infinity loop unless you really lose interfece constantly21:06
T42<zinstack625> @elros34 it brings interface up faster and makes it stable though21:07
swanuxIt partially works (as it keeps me connected - nice idea!) But still no telnet though (it'll timeout again probably)21:08
swanuxI got now an 1.9 IP in ifconfig, and the device is on 2.1521:09
T42<zinstack625> my bad, that ip should have been 192.168.2.9/2421:09
T42<zinstack625> muscle memory with local adresses21:10
T42<zinstack625> Also, can someone please look at my problem just above?21:10
swanuxWorks like a charm, thank you! (even though it saw my Mi 8 as Google 8 previously, in the most recent attempt as Ethernet, and now it's Google Nexus 4 (fastboot), lol)21:12
mal@zinstack625 can you show output from mer-kernel-check21:59
T42<zinstack625> mal: sorry, it's getting late here, already shut down my build machine, brb online in about 9-10 hrs22:03
malok22:13
mal@zinstack625 do you happen to remember which kernel version your device uses?22:13
T42<mintphin> Unrelated to Sailfish porting, but is the Xperia 10 II going to be supported by Sailfish X?22:41
T42<mintphin> Oh, I edited the message. Not a terrible edit tho. Sorry anyways.22:42

Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!