**** BEGIN LOGGING AT Sat Jan 16 03:00:36 2021 Jan 16 09:28:54 Morning! Jan 16 10:05:08 Morning! Jan 16 10:41:54 Herrie: I think the tissot linker isn't looking in the system/lib64/vndk-28 folder for finding .so Jan 16 10:44:03 Tofe: Isn't that simple adding it to ld_library_path? Jan 16 10:44:23 well, it seems to work without that on other devices Jan 16 10:55:26 Then it's probably some flag in boardconfig.mk that differs between Tissot and Mido/Rosy? Jan 16 10:55:36 At least that's my hunch Jan 16 10:55:54 NotKit might have a clue for that Jan 16 10:59:35 maybe Jan 16 12:53:18 Tofe: Seems I might have built Tissot at my end without those VNDK flags in the BoardConfig.mk from common Jan 16 12:53:23 Retrying now including those Jan 16 13:01:51 Tofe: I got this hint from another dev to test: PRODUCT_USE_VNDK_OVERRIDE := false Jan 16 13:01:51 And BOARD_VNDK_RUNTIME_DISABLE := true Jan 16 13:49:06 mmm I'd like to finally understand all those flags :D Jan 16 14:02:30 Herrie: on tissot we have /android/system/lib64/vndk-28/android.hardware.wifi@1.0.so whereas on rosy we have /android/system/lib64/android.hardware.wifi@1.0.so Jan 16 14:03:55 let's retry with explicitely putting the dependency in device.mk Jan 16 14:06:40 the .so are still in the same place Jan 16 14:07:18 so probably a failed attempt Jan 16 14:15:19 nope, didn't help a bit Jan 16 14:22:40 Herrie: I have "openat(AT_FDCWD, "/dev/adsprpc-smd", O_RDONLY|O_NONBLOCK) = -1 EPERM (Operation not permitted)" , maybe it's just that ? Jan 16 14:29:10 Tofe: I noticed that in the common for Tissot didn't put the wifi fix Jan 16 14:29:50 and adsprpcd isn't disabled Jan 16 14:30:07 I disabled that at my side on device.mk I think Jan 16 14:30:18 I see https://github.com/Herrie82/android_device_xiaomi_msm8953-common/commit/7b659fd5f661501f70d250ff44586c20e09c7862 Jan 16 14:30:59 Yeah Jan 16 14:31:10 But seems I have some commits in another branch as well :S Jan 16 14:31:11 https://github.com/Herrie82/android_device_xiaomi_msm8953-common/commit/4bb14aafc640431ae9287e746c9f530c1f2af4c7 Jan 16 14:31:18 revision="halium-9.0-extended" yes Jan 16 14:31:44 shall I try to merge them in halium-9.0 ? Jan 16 14:33:33 Not sure we want all Jan 16 14:33:37 Was testing a few things Jan 16 14:33:47 The extended one doesn't build for Mido for example I noticed Jan 16 14:34:35 ok so I'll just cherry-pick a bit for the moment Jan 16 14:35:45 I mainly want this adsp spam gone, to analyze the rest more quietly Jan 16 14:36:04 Yeah that one is really annoying Jan 16 14:36:36 ok I cherry-picked adsprc and time_daemon Jan 16 14:37:07 now how to have these vndk-28 libs found, or moved Jan 16 14:39:34 Tofe: came across this: If PRODUCT_USE_VNDK_OVERRIDE is false, the ro.vndk.lite property will be automatically added to /vendor/default.prop and its value will be true. Consequently, the dynamic linker will load the linker namespace configuration from /system/etc/ld.config.vndk_lite.txt, which isolates only SP-HAL and VNDK-SP. Jan 16 14:39:35 To upgrade an Android 7.0 or lower device to Android 9, add PRODUCT_TREBLE_LINKER_NAMESPACES_OVERRIDE := false to BoardConfig.mk. Jan 16 14:39:38 From https://source.android.com/devices/architecture/vndk Jan 16 14:40:56 Here some details on VNDK and how it works: https://source.android.com/devices/architecture/images/vndk_design_android_o.pdf Jan 16 14:53:10 Thanks, this will help Jan 16 14:54:41 I don't think I've put BOARD_VNDK_VERSION := current in Rosy, which explains where the .so are Jan 16 14:54:54 so we'd have to compare with mido Jan 16 14:54:59 It's pretty detailed in general and explains many flags Jan 16 14:55:11 Didn't have time to study it yet, but looks helpful Jan 16 14:56:04 https://github.com/Tofee/android_device_xiaomi_rosy-2/commit/c6ac7910420974059a880c31d18b4674e57461ac that's what I though: I did have the flags, but commented it out Jan 16 14:57:29 I'll try the same on tissot; it's not really what we want, but I'd like to see if it helps a bit Jan 16 15:01:57 reading the pdf, it looks like the linker *should* look into the vndk-28 folder... Jan 16 15:09:05 Herrie: when you'll have some time, it would be nice to make a little test on mido, where things work quite well: do you have a system/lib64/vndk-28 folder ? is android.hardware.wifi@1.0.so in there ? is it also in /system/lib64 ? if you take some vendor process and look at its /proc/$(pidof /maps , do you see some .so from vndk-28 folder ? Jan 16 15:09:51 without BOARD_VNDK_VERSION, without surprise, I now have out/target/product/tissot/symbols/system/lib64/android.hardware.wifi@1.0.so Jan 16 15:17:38 still a couple of issues (qseecom and camera mainly), but it looks like it's much better Jan 16 15:18:53 But now I realize that we have to make it run with BOARD_VNDK_VERSION, otherwise it's a bit hopeless for GSI Jan 16 15:18:53 Tofe: Seems I have a /system/lib64/vndk-sp-28 only on Mido Jan 16 15:19:25 Herrie: ok, interesting Jan 16 15:19:48 that one is also there on tissot, and is found Jan 16 15:20:30 it's the folder that is used from the "system" side processes to load dependencies specific to vendors Jan 16 15:20:47 vndk-28 would be the other way around, I think Jan 16 15:21:44 Herrie: where is your wifi.so service located ? Jan 16 15:24:40 On Mido? Jan 16 15:24:47 yes Jan 16 15:25:07 just in your out/ folder, should be enough to tell :) Jan 16 15:28:09 Tofe: Well that one is gone Jan 16 15:28:20 Let me reflash Mido Jan 16 15:28:30 Seems I had some bootloop Jan 16 15:40:43 Herrie: as long as tissot has not been prepared for treble at all, I'm not sure if we should seek to get an image working with a clear separation of vendor Jan 16 15:41:17 Tofe: Well I think there are some Tissot treble images out there Jan 16 15:42:06 https://forum.xda-developers.com/t/dev-treble-p-based-vendor-implementation.3918288/ like this ? Jan 16 15:44:40 Yes Jan 16 15:44:50 But the vendor is gone from there it seems Jan 16 15:44:54 https://github.com/Flex1911/proprietary_vendor_xiaomi-1/tree/lineage-16.0-treble Jan 16 15:45:44 Seems a vendor.img on: https://drive.google.com/file/d/1n1TcOXpVtore5M7-lVpHv7IP9Q3z00Sv/view Jan 16 15:46:30 https://github.com/ghpranav/proprietary_vendor_xiaomi-1 Jan 16 15:48:00 Ah ok and that would mean 4.9 kernel as well Jan 16 15:48:08 Would be nice to have if we can get it running Jan 16 15:48:43 I can give it a try Jan 16 15:49:00 if I just find Flex's vendor repo somehow Jan 16 15:50:30 TOfe: I pasted the vendor.img above Jan 16 15:50:36 The Google Drive link Jan 16 15:50:44 ah yes Jan 16 15:51:05 And this one was forked from Flex: https://github.com/ghpranav/proprietary_vendor_xiaomi-1 Jan 16 15:51:08 So should work too Jan 16 15:51:09 but the other repo is a fork from Flex's, so should be fine Jan 16 15:51:12 yes exactly Jan 16 15:51:35 well... let's go **** ENDING LOGGING AT Sun Jan 17 02:59:57 2021