**** BEGIN LOGGING AT Mon Jan 11 03:00:17 2021 Jan 11 07:53:34 Morning! Jan 11 07:56:02 Tofe: Couldn't sleep in the end so was busy with another image... Results are the same it seems, though I get less errors in dmesg and logcat. Still the same issue with the sound bits it seems. And I'm missing libwpa_client.so, not really sure how to get that one included, tried to add it to device.mk but no luck Jan 11 08:08:55 Morning! Jan 11 08:09:08 Herrie: the firmware missing file didn't help, then Jan 11 08:09:22 Well it helped a bit Jan 11 08:09:35 But didn't solve it Jan 11 08:09:52 The kgsl errors are gone Jan 11 08:10:01 Which is needed for hwcomposer to work Jan 11 08:10:11 So it's a step in the right direction Jan 11 08:10:29 OpenCL test works now as well Jan 11 08:10:36 At least it displays me that it runs successfully Jan 11 08:10:41 Which it didn't before Jan 11 08:11:09 https://paste.ubuntu.com/p/kRy2y7fsY2/ Jan 11 08:11:31 Seeing it mentions A506 I guess it didn't like the missing firmware files Jan 11 09:02:13 It could be I disabled a bit too much in BoardConfigCommon.mk causing the libwpa_client.so not to build. I saw that in UBPorts they only disabled 3 lines, I disabled all. Now retrying. Jan 11 09:02:34 I.e. They only have these 3 disabled: https://github.com/ubuntu-touch-lavender/android_device_xiaomi_lavender/commit/ad640e121d0dc75017f31d860332327f8acd5d36#diff-247ee86229a709a6e2eedfc1d3c4a557825aee073e20b0112ab76f4ca8e4bc4e Jan 11 09:56:34 Tofe: OK got libwpa_client.so now ;) Jan 11 09:56:47 So logcat doesn't complain about that anymore Jan 11 09:56:53 test_egl_configs now works as well Jan 11 09:57:05 Doesn't segfault anymore Jan 11 10:07:21 So some progress here and there, but not there yet still Jan 11 10:07:27 Nothing outputting on screen Jan 11 10:15:21 logcat: https://paste.ubuntu.com/p/vwrqTYrZxN/ Jan 11 10:15:41 ServiceManager: service 'display.qservice' died and ServiceManager: service 'SurfaceFlinger' died might be some hints? Jan 11 10:15:47 Maybe we need minisfservice as well? Jan 11 10:16:54 maybe yes, I'm not sure about that one Jan 11 10:17:46 I didn't see anything special about minisfservice in GSI commits Jan 11 10:19:39 Tofe: Well they don't need it in UBPorts because it conflicts with Mir from what I understood Jan 11 10:20:11 P.s. Stracing surfaceflinger tells me my libGLESv2_adreno.so is in the wrong location Jan 11 10:20:34 https://paste.ubuntu.com/p/Q4czjv8zxw/ Jan 11 10:20:52 I have it in /android/vendor/lib64/egl/libGLESv2_adreno.so Jan 11 10:22:57 Herrie: you have strace in your android image ? lucky you Jan 11 10:23:07 Yes Jan 11 10:23:10 I can strace a lot ;) Jan 11 10:23:12 And logcat Jan 11 10:23:26 ah no, it's from luneos, isn't it Jan 11 10:23:34 My mido is just on hold with a black screen Jan 11 10:23:42 Sometimes dimms but well Jan 11 10:23:46 That's about it Jan 11 10:23:51 Herrie: welcome to my world Jan 11 10:23:55 So I can do lots of logcat, strace etc Jan 11 10:24:06 But I think these EGL binaries at wrong location might be a clue Jan 11 10:24:47 yes, worth a try, just a bit weird that lineage works well Jan 11 10:25:19 Well that's a different story :P Jan 11 10:25:27 We use libhybris Jan 11 10:25:36 Could be something like LD_LIBRARY_PATH or something Jan 11 10:26:33 right Jan 11 10:35:28 OK just duplicated all the EGL .so's to lib(64) now, let's rebuild Jan 11 14:09:51 Tofe: Still no luck... Jan 11 14:09:54 Same status.. Jan 11 14:10:00 I'll try to add these bits in simply: https://github.com/Herrie82/android_device_xiaomi_mido/commit/dc8bcf694e242c60791d4e8da905773c4d15624a Jan 11 15:07:04 Nope no luck still Jan 11 15:07:50 Did see this in logcat: openat(AT_FDCWD, "/data/vendor/display/qdcm_calib_data_nt35596_tianma_fhd_video_mode_dsi_panel.xml", O_RDONLY) = -1 ENOENT (No such file or directory) so trying to get that in the right location now Jan 11 15:10:39 Also saw these: https://paste.ubuntu.com/p/fRGJpmDP2s/ Jan 11 17:48:44 NotKit said it might be kernel. Jan 11 17:48:54 I'm doing a last attempt now with the .xml at the right location Jan 11 17:49:10 If that fails I might revisit the other kernels again, seeing the rest in the image is now OK as wel know Jan 11 17:49:17 NotKit said that binder changes could cause issues Jan 11 17:54:32 Herrie: I have openat(AT_FDCWD, "/data/vendor/display/qdcm_calib_data_ft8613_csot_5p7_720p_video.xml", O_RDONLY) = -1 ENOENT (No such file or directory) Jan 11 17:54:39 followed by openat(AT_FDCWD, "/vendor/etc/qdcm_calib_data_ft8613_csot_5p7_720p_video.xml", O_RDONLY) = 12 Jan 11 17:55:12 so it looks like there's a fallback, and it also won't solve your issue I guess Jan 11 17:55:13 It's probably not critical anyway Jan 11 17:56:15 Tofe: While I was in talks with NotKit, I came across https://github.com/Halium/hybris-patches/blob/halium-9.0/build/make/0003-hybris-Reduce-vendorimage-build-size.patch Jan 11 17:56:23 Which I guess might be the reason for my vendor.img issues ;) Jan 11 17:56:39 We should build full vendor.img for our targets that are non-GSI Jan 11 17:58:37 oh, damn, right Jan 11 18:03:29 vendor rebuilt, pushing Jan 11 18:03:43 (no need to cleanup, we want more flies anyway) Jan 11 18:06:34 same Jan 11 18:07:02 ps: I alraedy have minisfservice Jan 11 18:10:13 Tofe: OK Jan 11 18:10:18 Well I added those Jan 11 18:10:29 Doesn't hurt probably Jan 11 18:10:55 it's frustrating, eveything seems to be up, but it doesn't see the composer HIDL service Jan 11 18:11:37 Yeah Jan 11 18:11:59 Besides the spam in logcat for the asdc or whatever it's called it's pretty OK Jan 11 18:12:49 the only other thing that fails is the vibrator Jan 11 18:12:59 I can't imagine this blocking the whole system... Jan 11 18:12:59 Yes not sure why but well Jan 11 18:13:06 And wifi test for me also Jan 11 18:13:23 But could be I need to write something to enable it Jan 11 18:13:40 there are some patches about it in GSI Jan 11 18:13:56 like "echo 1 > /dev/wcnss_wlan" or something like that Jan 11 18:14:05 Yeah that one Jan 11 18:22:20 Other 3.18 kernel still crashes with alignment error it seems Jan 11 18:24:22 on my side I hopefully have only one kernel, baunilla's Jan 11 18:24:29 though I use our 7.1 defconfig Jan 11 18:25:11 Yeah it's a bit tricky here with kernels Jan 11 18:32:13 comparing with lineageos's vendor, I miss: ./bin/hw/android.hardware.cas@1.0-service ./bin/hw/android.hardware.media.omx@1.0-service ./bin/vndservicemanager ./etc/init/android.hardware.cas@1.0-service.rc ./etc/init/android.hardware.media.omx@1.0-service.rc ./etc/init/vndservicemanager.rc ./lib/hw/vibrator.default.so , the rest is probably not so interesting Jan 11 18:32:51 omx would be for the medias, I'll look at it last; "cas" I'm not sure what it's about Jan 11 18:33:32 vndservicemanager I have it in system Jan 11 18:33:59 vibrator.default.so I can always try to get that one, that'll be one less thing missing Jan 11 18:41:09 "cas" is related to media too Jan 11 18:45:45 ok, including vibrator.default package did the trick for this one Jan 11 18:47:16 You also have the adsp one? Jan 11 18:47:23 Really flooding with me Jan 11 18:47:36 nope Jan 11 18:50:02 Hmmz Jan 11 18:53:41 ooh there is a /system/bin/lshal command Jan 11 19:02:02 Tofe: Any fixes you have for device repo, please push ;) Jan 11 19:08:43 done Jan 11 20:00:59 Thnx that saves me fiddling Jan 11 20:01:30 Really weird but doing the mka vendorimage separately (even without reversing the patch) gives me more proper vendor Jan 11 20:08:02 mmh the firmware was already loaded with my own vendor image, so it's not that Jan 11 20:10:45 Can you just put them side by side and compare contents? Jan 11 20:10:59 I.e. with WinMerge or something similar I would use on Windows Jan 11 20:13:31 well I can remark that "/vendor/bin/vndservicemanager" is started, not the one in /system, so I have to check again if it's started with my build Jan 11 20:13:52 Herrie: I did the diff up there at 19:32:13 Jan 11 20:14:17 it was with the "cust" partition, but it's what is flashed by lineageos for vendor Jan 11 20:15:24 Well custom vndservicemanager could be it easily Jan 11 20:15:53 there is just out/target/product/rosy/vendor/bin/vndservice Jan 11 20:16:02 though it might be it Jan 11 20:18:07 aha, there is a hybris-patches/frameworks/native/0008-halium-include-vndservicemanager-into-system-partiti.patch Jan 11 20:18:59 .... which doesn't change the path in the .rc file ? Jan 11 20:20:36 Well I think the true or not flag determines which one to use Jan 11 20:20:52 Since we're not on GSI we probably want ot keep the true Jan 11 20:21:17 Or not, I'm confused :P Jan 11 20:21:22 it's just the target I guess; but the path in the .rc has to match Jan 11 20:21:35 let's see Jan 11 20:27:37 well, it's better ! Jan 11 20:29:35 Herrie: can you just change /vendor/bin/vndservicemanager to /system/bin/vndservicemanager in frameworks/native/cmds/servicemanager/vndservicemanager.rc and refresh the system image Jan 11 20:29:50 for me it did the trick Jan 11 20:30:30 of course, now, I still have some steps: "CRITICAL: 01:22:42.691: Cannot open keyboard input device '/dev/input/event6' (No such file or directory)" and "library "libstagefright_ccodec.so" not found" but nothing I can't handle Jan 11 20:32:36 Tofe: Aaah goood! Jan 11 20:33:09 Seems NotKit is merging the fixes, he removed the vendor minimizing already Jan 11 20:38:22 great Jan 11 20:39:17 I'm a bit baffled at how things could have worked so far... are we the first people to rebuild our vendor image ?... Jan 11 20:48:47 why doesn't it build libstagefright_ccodec ?... Jan 11 21:17:52 anyway, this was a satisfying evenig, I'll leave it there for tonight Jan 11 21:17:55 gn8! Jan 11 21:18:17 Tofe: Yeah we're the first ones it seems mainly Jan 11 21:18:30 There are a few UBPorts ports out there but they're slightly different Jan 11 21:18:38 I did another round of stairs painting... Jan 11 21:18:47 Mrs will be back on Friday so want to have some things done before that :P Jan 11 21:18:56 3 stairs that needed some work ;) Jan 11 21:19:01 So quite some layers LOL Jan 11 21:19:03 :) Jan 11 21:19:25 consequence of having a huge house ! I don't have this issue :p Jan 11 21:19:56 HYeah LOL Jan 11 21:20:18 Well I did some 2 years ago, but the silicone started to crack everywhere, so decided to ditch it all and do new Jan 11 21:20:26 And anyway it needed another layer of paint Jan 11 21:20:40 And while I was at it, started to replace all the other silicone that was bothering me :P Jan 11 21:20:43 Cracks everywhere LOL Jan 11 21:21:37 "A house is never done" Jan 11 21:24:02 True LOL Jan 11 21:24:12 Especially new ones, since cracks will appear in the first 2 years LOL Jan 11 21:32:59 Tofe: Just curious, why would we need fonts? Jan 11 21:35:13 Herrie: camera drivers need some fonts, at least on tissot, I remarked; it's also included in GSI config Jan 11 21:35:30 must be for some watermark Jan 11 21:36:10 Ah ok Jan 11 21:36:15 Yeah I remember now Jan 11 21:37:03 OK mido with all changes building as well Jan 11 21:41:22 I'll leave you there, I've got some book reading catchup to do :) Jan 11 21:48:34 Hehe Jan 11 21:48:45 I'm building, 1 more stairs and some testing, then sleep :P Jan 11 22:55:24 test_opengl crashes, test_glesv2, test_glesv3 and test_hwcomposer work Jan 11 22:55:36 test_vibrator still not Jan 11 23:04:13 LunaSysMgr seems to crash at my end (not really sure why) which causes luna-next not to start eventually it seems Jan 11 23:04:32 Could be vibrator nyx module (or something else).. **** ENDING LOGGING AT Tue Jan 12 02:59:57 2021