Monday, 2021-05-03

riniguspiggz: did it work last night?05:59
DeckedHello everyone, I've been having some trouble porting SailfishOS to an old  LG G2, is this the right place to ask?07:02
DeckedI'm not sure if this is even working, I've never used IRC to begin with07:02
T42<Herrie1982> @Decked [I'm not sure if this is even working, I've nev …], It is working, yes this should be the right place07:10
DeckedGreat, thanks.07:11
T42<Herrie1982> G2 is quite similar to Nexus 5 in terms of specs07:11
T42<Herrie1982> There might be something done for it already07:11
DeckedI see, can I ask a small thing anyway?07:13
T42<Herrie1982> Sure you can ask, I'm not an expert on SFOS porting, but might be able to give you an answer, otherwise the other people here could.07:17
DeckedSo I was trying to port Sailfish os from Lineage 17.1, but I'm now stuck on part "7.2.1 Building the droid-hal-device packages". When I use the command `rpm/dhd/helpers/ --gg` the script can't build gst-droid.spec because it says it's missing gstreamer-photography-1.0. I've checked and it's gstreamer1.0-plugins-bad is already installed on both the PlatformSDK and on the The Target_SDK07:20
DeckedI tried removing the dependency on since I'm sure it's already installed but it fails somewhere else because there is still the dependency defined somewhere else07:22
T42<Herrie1982> I'm not sure how well 17.1 is working for SFOS port. For most other OS-es (UBPorts, LuneOS, Mobian etc) using Halium project) they're using either a 12.1, 14.1 or 16.0 based CM/LOS). I haven't done any SFOS ports myself, so cannot really help you there.07:23
DeckedI'm pretty new on SailfishOS but I saw that they added Android 10 support as of lately so I thought to give it a chance. From my understanding this problem isn't even related to the Android base and I can't find anything on the FAQs or IRC logs07:26
DeckedAnyways, thanks a lot either way, I'm sorry to trouble you07:29
rinigusDecked: I wonder whether you use too new gst-droid. check the version that you try to build against sfos 4.0.1 changelog07:30
DeckedHmm other than the fact that Gstreamer will be replaced by MW I can't find anything. I'll try building it without it and report back07:37
rinigusDecked: the repos are checked out under hybris/mw/, so see the version of gst-droid over there. for sony tama, I had to build gst-droid manually, without using -gg07:51
DeckedLooks to me as 1.0.008:00
riniguscheck with git log08:01
rinigusI am using 0.20210304.008:01
DeckedI see 0.20210430.008:01
rinigusthat could be too new. checkout older version and build using rpm/dhd/helpers/ --build=hybris/mw/gst-droid08:03
rinigusDecked: otherwise it will checkout the newest version again.08:04
DeckedSeems to have worked! thanks a lot08:06
rinigusmaybe this will help a bit - . grep for gst- and there is an instruction next for gmp-. those replace -gg build08:06
DeckedYou suggestion worked perfectly, I changed to configuration to exclude gstreamer1.0-droid, ran --gg, after that I included it again and built gst-droid on 0.20210304.0 as you suggested08:09
DeckedI'm now building the rootfs (I think)08:10
DeckedFor some reason my PC is complaining about my Wacom Tablet being disconnected and reconnected when building it O.o but alright08:10
DeckedTime to see if it boots08:12
DeckedUgh failed to extract filesystem, what now08:15
rinigusDecked: maybe someone can chip in with fs extraction, my port is not based on lineageos and has different mic packing08:17
rinigusI suggest to get a log with the error, paste on pastebin-type site and send URL over here. maybe someone can look and help you08:18
DeckedI'm using the lineageos recovery, could that be the reason?08:18
Decked[  268.902183] tar: exec bunzip2: No such file or directory08:24
DeckedHere's the problem08:24
rinigusDecked: you mean to flash it? no idea, we used to use twrp for lineageos based ports... don't know how it is now08:26
DeckedYeah, I might try flashing TWRP first08:27
DeckedOh this is a problem08:42
DeckedTWRP worked but got out of memory before finishing08:42
T42<b100dian> Decked: I don't have bunzip2 in twrp either. I am flashing the boot img with fastboot or with twrp then use something along these lines for pushing the unbzipped tar and exctracting it on device
DeckedWell my TWRP has bunzip2 installed with it but now the installation fails because the phone runs out of memory08:52
DeckedUsing tar manually on adb shell also fails because it runs out of memory09:03
T42<Herrie1982> We had some issues with tar on LuneOS from TWRP09:23
T42<Herrie1982> I think we use the busybox one now to do the unpacking09:23
T42<Herrie1982> @Herrie1982 [I think we use the busybox one now to do the u …], Yes see
DeckedI see09:33
DeckedReplacing busybox in TWRP did the trick thanks a lot09:51
T42<Herrie1982> Some of these funny quirks...09:57
DeckedHmm the phone won't boot after flashing, something something boot verification failed.10:02
T42<elros34> Decked, I guess you are booting android now. Follow 2 kernel related points from: That wont be enough to get gui on 17 base but maybe at least telnet.10:25
Decked@elros34 sorry for the late reply, I'm not booting from Android, it's the phone itself that refuses to boot anything after flashing boot.img12:43
T42<elros34> If You didnt follow advices mentioned in that url then your are probably missing sailfish specific initramfs in hybris-boot.img and your kernel might be trying to boot from /system partition so you are basically booting android12:46
T42<elros34> also flashing boot.img is wrong, you must use hybris-boot.img12:47
DeckedYeah I did that12:47
DeckedSorry I mispelled12:48
DeckedBut I'm 90% sure I'm not booting android, in fact I don't think I'm booting anything at all12:48
DeckedWith the phone in that state plugging it via USB shows me the various MMC partitions as if I'd plugged in an hard drive and the recovery is unaccessable anymore12:49
T42<elros34> most important is to check how device shows in host dmesg: as android or mer boat loader12:50
Decked[ 6735.911416] scsi 6:0:0:0: Direct-Access     Qualcomm MMC Storage      1.00 PQ: 0 ANSI: 212:53
DeckedNothing else12:53
Deckedlsusb shows Bus 001 Device 019: ID 05c6:9006 Qualcomm, Inc. QHSUSB__BULK12:54
T42<elros34> is that first messega after you flashed hybris-boot or several reboots later?12:54
DeckedAfter I flashed hybris-boot, but the phone refuses to boot it whatsoever12:54
DeckedI'm 100% sure that the kernel didn't even load12:55
Deckednor the bootloader12:55
T42<elros34> so did you needed to revert skip_initramfs12:55
T42<elros34> ?12:55
DeckedThe phone is a LG G2 from 2013 so I skipped that part12:56
DeckedI don't think that applies to it12:56
T42<elros34> so it doesn't use system as root?12:56
DeckedI don't know, I know it uses a separate recovery partition though12:58
T42<elros34> check in fstab where system partition is mounted, in / or in /system12:59
DeckedFrom android right?13:00
T42<elros34> in device repo you will find fstab13:00
DeckedSorry about my ignorance but I don't follow you13:02
DeckedI don't have any fstab files other for the ones for the initramfs13:04
DeckedOh my bad I searched for fstab only13:06
T42<elros34> maybe vbmeta then, don't you have such a partition?13:08
DeckedHmm the phone is basically bricked at this point let me try to fix it and I'll get back to you13:10
DeckedOk managed to get into the recovery, there's no vbmeta partition13:11
DeckedDDR         cache       dbibak      fota        grow        modem       mpt         pad2        recovery    sbl1        ssd         tz13:12
Deckedaboot       cust        drm         fsc         laf         modemst1    pad         persist     rpm         sns         system      tzbak13:12
Deckedboot        dbi         encrypt     fsg         misc        modemst2    pad1        rct         rpmbak      spare       tombstones  userdata13:12
DeckedRealistically I should have tried to port Sailfish from an older LineageOS version13:13
piggzrinigus: why would tbuilder not auto install a deb that is available?15:31
riniguspiggz: it caches the available dependencies. remove caches in the build folder15:33
riniguslet me write it into the docs as well15:33
riniguspiggz: caches described in the updated readme of tbuilder15:42
piggzrinigus: now another issue finding some packages15:47
piggzrinigus: ah, failed to install because of sig verification15:50
piggzyou must have had that? a workaround?15:52
riniguspiggz: you could either add repo with an option to disable sig check out run zypper refresh manually in target (not snapshot) and accept non signed repo16:06
piggzrinigus: now, a funky file path is missing Cannot find file: /parentroot/data/piggz/mer/android/tbuilder/amazfish-build/src/harbour-amazfish/parentroot/data/piggz/mer/android/tbuilder/amazfish-build/build/SailfishOS-
piggzthis folder exists and is the source /parentroot/data/piggz/mer/android/tbuilder/amazfish-build/src/harbour-amazfish/16:16
piggzbut there shouldnt be a parentroot/ in there16:16
riniguspiggz: ohhh, that's something unexpected16:19
rinigusdo you compile with "shadow_builds"?16:20
piggz  - harbour-amazfish.spec16:21
piggzi just copied that from the example16:21
riniguswould you mind to try without?16:21
rinigusjust remove that option. not sure if it will help16:22
piggzrinigus: yeah, that seems to be building now16:23
rinigusto see where that issue came, please check how symlink was defined in spec. ls -l spec?16:24
piggzlrwxrwxrwx 1 piggz piggz 49 May  2 20:12 harbour-amazfish.spec -> ../src/harbour-amazfish/rpm/harbour-amazfish.spec16:34
piggzlooks ok16:34
riniguspiggz: it does look fine, indeed16:43
riniguspiggz: but I wonder where is this /parentroot sneaking in? let me find check if I can reproduce it16:46
T42Tz %lastname% was added by: Tz %lastname%16:59
T42<Tz %lastname%>
riniguspiggz: just in case - is src/harbour-amazfish folder or a link somewhere further?17:51
T42<adampigg> A folder17:52
rinigus@adampigg: I tried to reproduce it with /parentroot/tmp/test and a project under it, but no luck. no such weird path.18:29
riniguswould you mind to paste build.log for failing case? just define using shadow_builds as before and touch the spec to start the rebuild18:29
riniguswould also help to have `pwd` from the shell where you run tbuilder18:30
T42<adampigg> rinigus: i warn you, my path is a funny one (tho it has always worked)18:57
riniguspiggz: not sure about "always", let me grep the logs18:58
T42<adampigg> PlatformSDK [piggz@linux-f1uu amazfish-build]$ pwd18:59
T42<adampigg> /home/piggz/mer2/android/tbuilder/amazfish-build18:59
T42<adampigg> the funny bit:18:59
T42<adampigg> PlatformSDK [piggz@linux-f1uu amazfish-build]$ file /home/piggz/mer218:59
T42<adampigg> /home/piggz/mer2: symbolic link to /parentroot/data/piggz/mer/18:59
riniguspiggz: OK, I will try with something similar. let's see19:02
rinigusif I can reproduce19:02
piggzrinigus: background....19:03
piggzall my builds happen on my spinning disk, as it has space, mounted in /data, in /data/piggz/mer19:03
piggzall the scripts/docs usually use ~/mer as a folder, so i made that a symlink19:04
piggzbut, once in the sdk, that symlink doesnt work, so, i made another mer2 which works from within the sdk19:04
piggzANDROID_ROOT uses that one19:05
rinigusYeah, I used to use something like it, but ended up settling with just home. Let me see if I can reproduce it with symlinks in the path19:12
riniguspiggz: nope, it still works here as well. if I make ~/tmp/test -> /parentroot/tmp/test and use ~/tmp/test/subdir as a project dir, all seems to be resolved fine19:28
rinigusit detects that build should be done in /parentroot/tmp/test/subdir/build/SailfishOS- (which will fail, but for other reasons). no double parentroot in the path19:29
rinigusas you have19:30
piggzok, i just wont use shadow build19:30
piggzfor now, until i can be bothered debugging19:30
riniguslet's stick with that :)19:30
rinigusmany packages don't support it anyway, so it is easier to do without. I can build few of my own with the shadow builds, but not the port19:32
UmeaboyHi! Is there a fix for the python error when running the repo init command for hybris-17.1 ?21:09
T42<b100dian> You can run repo commands outside of the platformsdk shell21:10
UmeaboyI OK.21:11
T42<elros34> or use python2 compatible repo from same site you have downloaded python3 one21:11
UmeaboyThe HABUILD_SDK python version is 3.4 and 3.5+ is needed.21:16
UmeaboyIsn't it time to update the chroot to a newer version of Ubuntu ?21:17
UmeaboyIt would get rid of this issue as the python version would be more correct.21:18
UmeaboyJust saying.21:21
T42<elros34> for older devices is good to have old ubuntu chroot for compability reasons, for newer device you just don't need to use chroot at all21:23
T42<b100dian> I was also wondering, so one could just build the android bits in the host system and call it a day?21:28
T42<b100dian> So, I now have different crashing services (other than servicemanager) since i added LD_PRELOAD here for servicemanager.rc and then added the cheesburger init files (parent commit). Here is the sys/fs/pstore output: https://pastebin.ubuntu21:38
T42<b100dian> Maybe you can help with any of the questions:21:38
T42<b100dian> Q1: `drv2604_firmware_load, ERROR!! firmware not found` at about 12th second: this is not built as a module, however systemd seems to run the mounts before that, wondering why it doesnt find the firmware.21:38
T42<b100dian> Q2: There are now other services crashing, do they also need the LD_PRELOAD patch? Or do they need listing in disabled_services.rc? (grep for `>>>`. Example: /system/bin/hw/android.system.suspend@1.0-service)21:38
T42<b100dian> Q3: Reboot at second 88: `adsp subsystem failure reason: err_qdi.c:450:EF:sensor_process:1:dog_virtual_user.c:240:USER-PD DOG detects stalled initialization, triage with IMAGE OWNER.` Does this sound familiar? I've recently switched audio and wlan to modules, but they seem to not be loaded automatically from rootfs's /lib/modules, in21:38
T42 fact, some "disagrees about version of symbol module_layout" messages are around second 41 which indicate modules are loaded from .. /vendor/lib maybe? (the untouched vendor partition..)21:38
*** Sellerie3 is now known as Sellerie21:38
Umeaboyelros34: Would you happen to know how it works to patch a newer release of Lineage once it's released? I'm curious about how it works. I know we can see all the commits, but I mean, what decides that that file should be patched so that the Hybris image can be built? My phone uses a unofficial version of Lineage 18.0, but the latest official version of Lineage is 18.121:45
T42coddato was added by: coddato21:49
T42<coddato> Hi Is there a build for herolte ?21:50
T42<elros34> Umeaboy: latest patches are for lineage-17 not 18. Preparing patches for lineage 18 is not trivial so without having some knowledge and spend a lot of time you will not be able to do this. I only patched some vendor sdk android sources once using existing patches for older base and even that wasn't piece of cake21:55
Umeaboycoddato: Yes:
T42<elros34> @b100dian I am wondering why do you add hybris-16 rc files if you are using hybris-17?22:04
T42<elros34> dmesg is not enough to figure out  what is wrong, you should check logcat and journal.22:04
T42<b100dian> Not that clever, right? Just trying everything I did not try. So you're saying i should back off the cheesburger init stuff because it was lineage-16 specific22:05
T42<elros34> I thinks you should follow hybris17 device repos: xperia x10 II, fp3 and probably some other devices22:06
T42<elros34> here is another device (no idea whether this one works):
T42<b100dian> Thanks, I'll try to look more into those. The xiaomi sagit seems promising:) - I know you said you dont' know if it works, just cheering that I recognize some patterns22:20

Generated by 2.17.1 by Marius Gedminas - find it at!