| mal | @b100dian some mount errors in that log, looks like fixup-mountpoints is missing some things | 00:03 |
|---|---|---|
| mal | Timed out waiting for device dev-block-bootdevice-by\x2dname-modem.device and similar for several other droid mounts which means fixup did not replace the device nodes | 00:04 |
| T42 | <b100dian> Maybe this time I am the one not having the update from rinigus. The fixup mountpoints is in rootfs though, right? Except boot and data which are needed by hybri- boot? | 00:06 |
| mal | fixup is done during hybris-hal build and droid-hal packaging | 00:07 |
| mal | hybris-hal build does it for init-script and droid-hal packaging for the rest https://github.com/mer-hybris/droid-hal-device/blob/master/droid-hal-device.inc#L561 | 00:08 |
| T42 | <b100dian> so I did have the fresh hybris-boot repo, this is for me to look into tomorrow then.. I may build the hals locally tomorrow though. Thanks! | 00:09 |
| ecrn | I'm trying to port ubuntu touch to a samsung device, is this somewhat relevant to this channel? | 00:31 |
| ecrn | I created udev rules from ueventd.rc files, but they can't take effect because the groups do not exist in the ubuntu rootfs | 00:36 |
| ecrn | https://pastebin.k4be.pl/view/raw/32462264 | 00:36 |
| mal | not sure how ubuntu touch does things but I would assume there is some script to create the users and groups needed | 00:37 |
| ecrn | where are android user groups defined? /system/etc/group and /vendor/etc/group are empty | 01:20 |
| ecrn | well, the vendor not quite, but there are still no such groups as per ueventd.rc | 01:21 |
| ecrn | https://android.googlesource.com/platform/system/core/+/refs/tags/android-cts-14.0_r6/libcutils/include/private/android_filesystem_config.h | 01:42 |
| ecrn | apparently there | 01:42 |
| ecrn | 11-12 02:54:47.743(1731380087743) PrimaryDisplay 0: commit error | 02:59 |
| ecrn | 11-12 02:54:47.743(1731380087743) PrimaryDisplay 0: setActiveDrmMode:: Failed to commit pset ret=-13 in applyDisplayMode() | 02:59 |
| ecrn | but the device permissions are valid now | 02:59 |
| rinigus | @b100dian: nice - at least you have telnet working. something is missing over here in making it work... | 05:13 |
| rinigus | as for fix mountpoints - maybe we need also to add modem in addition to modem_a? I think some devices had it. I have added based on what was reported in adb, i.e. only _a and _b variants | 05:15 |
| ecrn | the hwcomposer then segfaults at exynos_open | 05:54 |
| ecrn | having properly listed the modes and formats before | 05:55 |
| T42 | <elros34> rinigus but you do not add fixup mountpoints with _a/_b or without based on what was reported on device but based on android fstab | 06:48 |
| T42 | <elros34> @ecrn I am prietty sure there must be ubuntu channel somewhere | 06:49 |
| T42 | <birdzhang> It's https://t.me/ubports_porting | 06:53 |
| T42 | <elros34> rinigus: of course now when you have rootfs you can just check not working mount units to figure it out | 06:56 |
| T42 | <b100dian> rinigus: I have telnet with manually built hybris-recovery as in https://github.com/sailfishos-sony-nagara/main/wiki/Hybris%E2%80%90boot-without-full-SFOS-SDK point 4, and our hybris-boot repo. | 07:41 |
| T42 | <b100dian> and formatting sda94 as ext4 | 07:43 |
| ecrn | yes, I figured both halium and ubports have only telegram and matrix channels | 07:48 |
| rinigus | @elros34: will have to check fstab then | 08:42 |
| rinigus | @b100dian: will try tonight that | 08:42 |
| ecrn | hm, seems there are lots of vendor specific things on the original system parition in the samsung phone | 09:15 |
| ecrn | hmmm | 14:17 |
| ecrn | I got the wayland compositor to display image for a while | 14:19 |
| ecrn | and it even set up hw acceleration | 14:19 |
| ecrn | but the thing is, the hwcomposer service tries desperately to use DRM_IOCTL_MODE_ATOMIC without being a DRM master | 14:22 |
| ecrn | and then decides to restart | 14:22 |
| ecrn | how is that supposed to work? | 14:23 |
| ecrn | the compositor is supposed to dynamically load the hwcomposer so that both are used in DRM master context? | 14:24 |
| mal | does restarting hwcomposer help? | 14:31 |
| mal | we have setup that, if configured correctly, will restart hwcomposer in certain situations during boot | 14:32 |
| *** phlixi is now known as Guest9196 | 15:03 | |
| *** phlixi_ is now known as phlixi | 15:03 | |
| ecrn | mal: I don't have working shell outside initrd yet, but your suggestion caused me to check if it is the order of the services that causes the hwcomposer to not become DRM master | 15:25 |
| ecrn | and indeed it is | 15:25 |
| Mister_Magister | mal: what's the latest base sfos can run on? | 16:24 |
| mal | aosp14 or hybris-21.0 | 16:27 |
| Mister_Magister | as i thought thanks! | 16:27 |
| Mister_Magister | mal: we randomly got this bonehead idea of porting a tablet to make jolla tablet v2 xd | 16:28 |
| Mister_Magister | i wonder if sfos removed tablet ui at this point lol | 16:41 |
| mal | nope | 16:47 |
| Mister_Magister | noice | 16:47 |
| mal | we actually have a setting to allow landscape orientation on phones also coming in 5.0.0, even for homescreen | 16:51 |
| mal | I mean dynamic orientation of homescreen | 16:52 |
| T42 | <Mister_Magister> nice | 16:52 |
| rinigus | @b100dian: as I was on stock, thought to look into gsi that you tested (lineage). in my hands it keeps on logo spinning during boot and not really proceeding anywhere | 18:22 |
| rinigus | adb is full of Access denied finding property "sony.screen_brightness" | 18:22 |
| rinigus | used the same version as you did | 18:22 |
| rinigus | will try TD version and then will go back to aosp and sfos booting | 18:23 |
| rinigus | and td variant booted | 18:29 |
| T42 | <b100dian> What is TD? Also, did you clear data between boots? | 18:41 |
| rinigus | @b100dian: did not clean originally, cleaned later. but that didn't help. | 18:45 |
| rinigus | td: https://xdaforums.com/t/gsi-14-lineageos-21-trebledroid-based.4654132/ full TrebleDroid patchset | 18:46 |
| rinigus | that one worked and is very impressive (after aosp). | 18:47 |
| rinigus | but OK, will try to see if I can get rndis messages and get to telnet using your approach | 18:48 |
| T42 | <b100dian> Ok. ThebTD as you call it has a chance of working if we would be able to skip bundling system and vendor in the img - do you think thats possible? | 18:49 |
| T42 | <b100dian> *the TD | 18:49 |
| rinigus | I guess that is this "classical" way for SFOS porting, without LVM. | 18:51 |
| T42 | <b100dian> rinigus: Not necessarily. My intuition says that we could still package in a lvm partition over /data and mount through dynparts other partitions such as system and vendor | 18:53 |
| T42 | <b100dian> mal: is that possible? ^ | 18:53 |
| rinigus | I suspect it is easier to make in this case a classical port first and later tune it with lvm | 18:54 |
| T42 | <b100dian> Yes, but in our case it could be just manually unpacking & testing, for now. As I said, /data ext4, adb push, tar -xvzf .stowaways.. Let me know If I should detail this on the wiki too | 18:57 |
| T42 | <b100dian> rinigus: btw, I still have to build stuff locally.. do you have any changes to fixup I should try or should I try to make them? Have you extracted fstab (from device I think?) on stock? | 19:03 |
| T42 | <b100dian> I have something like this in build output https://pastebin.ubuntu.com/p/g4rT6CjWZS/ | 19:03 |
| rinigus | @b100dian: no, I don't have any changes for fixup, sorry. | 19:05 |
| rinigus | please do make the changes | 19:05 |
| T42 | <b100dian> ok, since I have usb-moded masked by chrooting inside recovery, I can try changes "live" first and fix later | 19:05 |
| rinigus | as we stumble a bit on lvm, I would make a branch without it and then we can later go back when all works | 19:05 |
| rinigus | as it was more or less hunted down, should be simple enough to make. before making zip image | 19:06 |
| T42 | <b100dian> uh, I have no clue what pattern/spec change is switching lvm <—> classical, if you know, please go ahead. I can look it up otherwise | 19:08 |
| rinigus | @b100dian: tried something silly - to boot stock with aosp boot images (boot, then +dtbo, then +vendor_boot) and ended with the device that had no fastbootD anymore. fortunately green light works and now flashing stock once more. after that - aosp | 19:09 |
| rinigus | I'll check that lvm. if its fast then will proceed with it and would document as well | 19:10 |
| T42 | <b100dian> I think I've tried that too. There is definitely a problem with stock/gsi and older kernel that we need to take into accoung I guess.. | 19:14 |
| T42 | <b100dian> *account | 19:15 |
| rinigus | yes, bit spooky though :) | 19:17 |
| rinigus | back on aosp, will look into lvm business | 19:17 |
| T42 | <b100dian> interesting, why is metadata.mount trying to use /dev/block/by-name/metadata (missing bootdevice) and vendor-dsp.mount trying /dev/block/bootdevice/by-name/dsp ? | 19:39 |
| T42 | <b100dian> Does this mean the sed call in fixup needs to really match this https://pastebin.ubuntu.com/p/g4rT6CjWZS/ and all will be created under the appropriate /dev/ path? | 19:39 |
| mal | @b100dian that metadata is quite usual, see for example here https://github.com/mer-hybris/hybris-boot/blob/master/fixup-mountpoints#L53 | 19:46 |
| mal | same for oem_a | 19:46 |
| mal | @b100dian sed call must match what you have in fstab and/or the .rc files it parses, usually just fstab has mounts but sometimes also in some .rc files | 19:47 |
| T42 | <b100dian> mal: thanks, then I'll push my changes, this is what I've guessed, but it is better to have it confirmed :) | 19:47 |
| mal | you can also see that dsp partition being block/bootdevice/by-name/dsp not _a | 19:48 |
| mal | same for modem | 19:48 |
| T42 | <b100dian> thanks mal - can you also shed some light if the lvm builds can be made without including /system and /vendor folders? | 19:55 |
| rinigus | @b100dian: haven't tested it yet, but builds are on the way. see our wiki, last section of hadk | 19:57 |
| rinigus | now just have to find mic command | 19:58 |
| T42 | <b100dian> rinigus: I just pushed fixup mountpoints fixup | 19:59 |
| mal | afaik nothing really prevents lvm build which uses mounts for system/vendor | 19:59 |
| mal | but before doing lvm build it is good to make regular image which boot reliably to UI | 19:59 |
| T42 | <b100dian> mal: but are they included by default and you have to exclude them, or the other way around? | 20:00 |
| rinigus | @b100dian: that means rebuilds for hal with new fixup, I think | 20:00 |
| mal | which part includes them? | 20:00 |
| T42 | <b100dian> rinigus: the command https://github.com/sailfishos-on-sake/Release/blob/main/scripts/create-image.sh#L71 | 20:01 |
| mal | you can always add Provides: droid-system or something like that to some package like droid-hal if there is some hardcoded dependency somewhere | 20:01 |
| T42 | <b100dian> mal: so this https://github.com/sailfishos-sony-nagara/droid-hal-sony-nagara/blob/hybris-14-r67/rpm/droid-hal-common.inc#L23 needs to go ? What about vendor? | 20:04 |
| mal | why did you add it there? | 20:06 |
| mal | droid-system provides both | 20:06 |
| mal | so to use mounts you just need to remove the partitions you want as mounts from that makefstab_skip_entries part | 20:07 |
| rinigus | @b100dian: its gone already in nolvm branch | 20:08 |
| rinigus | mal: thanks - that info we were also looking for. missed that / in the list - while it wasn't there in sake | 20:09 |
| rinigus | as for why it was added - that's based on zambezi in its lvm form | 20:10 |
| rinigus | so, what are we expected to have in makefstab_skip_entries. something like in sake https://github.com/sailfishos-on-sake/droid-hal-sake/blob/master/droid-hal-sake.spec#L13 | 20:12 |
| rinigus | or empty? | 20:12 |
| mal | sake seems to have lvm build or custom mount services, not sure which | 20:14 |
| mal | custom mounts https://github.com/sailfishos-on-sake/droid-config-sake/tree/master/sparse/usr/lib/systemd/system | 20:14 |
| T42 | <elros34> rinigus you add entries to skip so custom mounts (dynamic partitions) or packaged partititions (as in hadk aosp build) | 20:15 |
| mal | probably if the partitions are in super partition some custom services are needed | 20:17 |
| rinigus | OK, so all dynamic partitions should go to custom mounts. I think we had several, including / | 20:17 |
| mal | @b00dian ^ | 20:17 |
| rinigus | @elros34 and mal : thank you very much, makes sense now | 20:17 |
| T42 | <elros34> according to https://pastebin.ubuntu.com/p/g4rT6CjWZS/ you don't have "/" | 20:19 |
| mal | I think ignoring / is because of the system_root stuff | 20:19 |
| mal | not sure if it's just habit to put it there | 20:19 |
| T42 | <b100dian> rinigus: sake has custom -> dynparts / super mount points in the script | 20:20 |
| rinigus | @b100dian: let's continue on "nolvm" branches then. | 20:21 |
| rinigus | so, we would have to populate https://github.com/sailfishos-sony-nagara/droid-config-sony-nagara/tree/nolvm/sparse/usr/lib/systemd/system with super mounts | 20:23 |
| T42 | <b100dian> rinigus: ok. This is the way https://github.com/sailfishos-on-sake/droid-config-sake/commit/40d0ab86a7aa0f236be282dbf8c9fcbdfc0f19a5. I'll do it | 20:23 |
| rinigus | @b100dian: thanks! just keep an eye on used branch :) | 20:24 |
| rinigus | I wonder if changes in fixup mountpoint require full hybris-hal rebuild or "dirty" one is fine. it looks to be called in hybris-boot/Android.mk . not sure if anywhere else in android builds | 20:26 |
| mal | nothing more should build if you just change that, only hybris-boot.img and stuff like that should get regenerated | 20:27 |
| rinigus | excellent! | 20:28 |
| rinigus | @b100dian: pushed hal images with your fixes (mountpoints) to obs. when you make your changes in config, just trigger rebuilds at obs. | 20:41 |
| rinigus | after that, mic create fs should work. | 20:41 |
| rinigus | I'll stop for tonight. will catch up tomorrow. good luck! | 20:42 |
| T42 | <b100dian> rinigus: thank you | 20:43 |
| T42 | <b100dian> rinigus: I pushed the exact commit from fp4/sake, because I don't think we need vendor_dlkm https://pastebin.ubuntu.com/p/x9PSHF7qXT/ and I think we are on system-as-root as mount has shown `/dev/block/dm-12 on / type ext4` (dm == dynpart) | 21:23 |
| T42 | <b100dian> So droid-config-sony-nagara updated and mw submodule in droid-hal-sony-nagara added | 21:25 |
| T42 | <b100dian> I wanted to start my first droid-hal packaging with this env https://github.com/sailfishos-sony-nagara/main/wiki/HADK-for-Sony-Nagara-devices#environment-variables-in-hadkenv (223->224, xqct->xqcq) but I'm running into | 21:39 |
| T42 | <b100dian> `rpm/dhd/helpers/build_packages.sh -d` -> `Fatal: 'sony-xqcq54-aarch64' is not a known build target` | 21:39 |
| T42 | <b100dian> too stupid to think what's wrong at this hour, gonna log out | 21:41 |
| ecrn | lomiri crashes with segmentation fault soon after starting and causes compositor crash on the ubuntu touch, I captured the core dump but now have no idea how to debug - gdb doesn't show much even when pointed to right sysroot | 23:16 |
| ecrn | trying to dump strace changed the behavior and it crashes or exits just when loading libraries | 23:17 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!