**** BEGIN LOGGING AT Fri Jan 25 02:59:58 2019 Jan 25 08:26:22 Sam Sundar was added by: Sam Sundar Jan 25 10:18:25 Sam Sundar this https://github.com/sailfishos-kenzo could help you, working sailfish OS based on halium Jan 25 16:31:27 @Daniel, Thank you Daniel Jan 25 20:59:18 @WaseemAlkurdi Hey I just applied this patch, it may fix hwcomposer for us. Can you test? :) … https://github.com/Halium/android_kernel_oneplus_msm8996/pull/1/commits/3e019b8700b36aef379846a748e3a447a8c3dbe7 Jan 25 20:59:50 (It's for test_hwcomposer failing on aarch64 devices with armhf rootfs) Jan 25 21:01:59 (`repo sync -c` and recompile `halium-boot`) Jan 25 21:12:28 @Sergobot Will do! Jan 25 21:13:18 @NotKit The errors happen in both rootfs'es, CAF and non-CAF Jan 25 21:14:15 @Sergobot I'm using a vendor image from July IIRC Jan 25 21:15:04 You need the one that came with Nougat build we're based on Jan 25 21:15:14 @Sergobot, O. u. c. h. Jan 25 21:15:37 NEVER occured to me😢 Jan 25 21:15:52 No problem ;) Jan 25 21:16:24 Grab NOF27B from here … https://developers.google.com/android/images#marlin Jan 25 21:16:37 (It's March 2017, just in case) Jan 25 21:17:33 Doing that now Jan 25 21:17:46 Aaaand that's not only vendor, it's the whole factory image Jan 25 21:18:25 So, you'll need to unzip it and flash only vendor.img Jan 25 21:18:34 It's a little heavy (1.8 GB) Jan 25 21:18:41 I'm low on data Jan 25 21:18:53 If you have the vendor image on hand Jan 25 21:18:57 Can you send it to me? Jan 25 21:19:27 I think I have it downloaded somewhere Jan 25 21:19:30 Let me check Jan 25 21:19:42 I think you're looking for a different page Jan 25 21:20:05 https://developers.google.com/android/drivers#marlin Jan 25 21:20:19 The vendor image can be downloaded from here Jan 25 21:20:51 Thanks a lot! Jan 25 21:21:06 Oh thanks, didn't notice that page before @UniversalSuperBox Jan 25 21:21:18 It's great for almost everything Jan 25 21:21:30 If you want to run a custom ROM they normally complain if you don't have the right vendor image Jan 25 21:21:43 Exactly Jan 25 21:21:50 They'd be stuck on that Google logo Jan 25 21:21:51 This is much more convenient than getting the full factory image and wiping your device with it Jan 25 21:22:00 @UniversalSuperBox, Totally agree! Jan 25 21:22:31 It's a heeeuuuge waste of data and time to download full flash images simply to pluck out a vendor image😅 Jan 25 21:39:22 I've downloaded the first link Jan 25 21:39:28 Vendor image Jan 25 21:39:43 Ran the (100 MB?) script after giving its code the once-over Jan 25 21:39:52 It extracted a bunch of files Jan 25 21:39:58 Yeah, blobs in scripts. What a world. Jan 25 21:40:39 @UniversalSuperBox, Precisely! … It even gets weirder when you read the script, like damn what? Is the encoding all messed -up? Jan 25 21:40:46 Then you du -h the script Jan 25 21:41:05 You should have a single `vendor.img` in the extracted files. Jan 25 21:41:07 to see a hundred-megabyte hell of a .sh Jan 25 21:41:11 I do Jan 25 21:41:17 Is that all Jan 25 21:41:18 ? Jan 25 21:41:19 Flash it up and you're on your way Jan 25 21:41:43 Fastboot should handle it. `fastboot flash vendor vendor.img`... unless there's something different with A-B slots Jan 25 21:42:26 With A/B, one flashes vendor_a, reboot bootloader, then vendor_b, then a reboot. Jan 25 21:43:39 @UniversalSuperBox, But on Google's site, there's this other link that includes drivers for "GPS, Audio, Camera, Gestures, Graphics, DRM, Video, Sensors". Jan 25 21:43:44 Do we need any of that? Jan 25 21:44:09 I think those get put into an Android build tree if you desire to link against them or something... Jan 25 21:44:52 @UniversalSuperBox, Which most people don't, as ROMs mostly use LineageOS trees, right? Jan 25 21:45:06 Sounds right to me Jan 25 21:58:04 Good sign: didn't need to manually mount /vendor this time Jan 25 21:58:13 test_vibrator worked without a hitch Jan 25 21:58:20 Ah damn! Jan 25 21:58:32 root@localhost:~# EGL_PLATFORM=hwcomposer test_hwcomposer … cannot locate symbol "atrace_end_body" referenced by "/vendor/lib/hw/gralloc.msm8996.so"... … test_hwcomposer: test_hwcomposer.cpp:171: int main(int, char**): Assertion `err == 0' failed. … Aborted Jan 25 21:59:31 Interesting. That symbol should be available in libcutils Jan 25 21:59:58 Is there something like good ol' Dependency Walker we can use? Jan 25 22:00:19 `ldd` does that, though how you use it I'm not sure Jan 25 22:00:42 I'll find out Jan 25 22:01:03 You're going across the hybris line here Jan 25 22:01:16 I'm getting a 'not a dynamic executable' Jan 25 22:01:39 That's for `ldd /vendor/lib/hw/gralloc.msm8996.so`? Jan 25 22:01:42 Possibly statically linked library? Jan 25 22:02:16 That seems unlikely. You could try running `file` on whatever you're trying to get linker dependencies on Jan 25 22:03:19 root@localhost:~# file /vendor/lib/hw/gralloc.msm8996.so … /vendor/lib/hw/gralloc.msm8996.so: ELF 32-bit LSB shared object, ARM, EABI5 version 1 (SYSV), dynamically linked, BuildID[md5/uuid]=1c7f8057c597012556e6724f9087cbca, stripped Jan 25 22:03:22 Double odd Jan 25 22:03:52 Nothing to do with the stripping though, because that's done for debugging symbols Jan 25 22:03:57 WAIT ... symbols? Jan 25 22:04:09 Isn't that what we're looking for here? Jan 25 22:04:39 No, debugging symbols are more "give a human friendly name to the things in the binary" Jan 25 22:05:14 The problem is likely that the `libcutils` shared object is not being linked in by Libhybris Jan 25 22:05:26 But I'm not sure how that's handled, so I must bow out after that. Jan 25 22:06:10 The Sailfish guys told me to "add a stub" to libhybris Jan 25 22:06:26 Sailfish OS, not sailfish the device Jan 25 22:07:22 I hope they knew what that means, because I have no idea. Jan 25 22:08:15 @UniversalSuperBox, At least we share the same libhybris, right? Jan 25 22:08:50 Because they told me to build hybris-boot first (and we don't do that here) Jan 25 22:09:14 I told them that it doesn't support A/B, they told me it does Jan 25 22:09:21 But we're using halium-boot Jan 25 22:09:28 So what do you think? Jan 25 22:09:31 @UniversalSuperBox Jan 25 22:10:55 Halium-boot is what you'll have to use if you want to boot Ubuntu Touch. Whichever boot image you use, it won't affect this far up the stack. Jan 25 22:11:32 @UniversalSuperBox, Exactly ... it's a boot image, after all. Dunno why they seemed to insist that I use hybris-boot Jan 25 22:11:43 They probably thought you wanted to boot Sailfish Jan 25 22:11:57 In which case it would indeed be folly to build halium-boot Jan 25 22:12:01 @UniversalSuperBox, I explicitly said I was booting Halium though Jan 25 22:12:30 Yeah, but they're the Sailfish OS porters. Normally we send people there who want to port Sailfish, I'm surprised they didn't just send you back Jan 25 22:13:24 It's their stuff, libhybris, so I guess they could help Jan 25 22:14:09 It seems to be a circular loop Jan 25 22:14:53 We're stuck here on adding a stub to libhybris, we go to them, they send us back, ad infinitum Jan 25 22:15:01 😅 Jan 25 22:16:50 They might mean that they want you to put `void atrace_end_body() { }` inside libhybris and rebuild it? Jan 25 22:17:22 I'll try Jan 25 22:17:23 @UniversalSuperBox, Yes I also feel its like that Jan 25 22:17:39 It's a folder Jan 25 22:17:54 I mean, the reference implementation of the library just has that blank. Jan 25 22:17:56 So Jan 25 22:18:14 Libhybris isn't built inside the Android source tree. It'd be rebuilding the deb package. Jan 25 22:18:15 hm, actually Jan 25 22:18:42 could you try HYBRIS_LD_PRELOAD=/system/lib/`libcutils`.so EGL_PLATFORM=hwcomposer ... ? Jan 25 22:19:07 I will Jan 25 22:19:13 Give me a moment Jan 25 22:19:41 @WaseemAlkurdi, Out of curiosity I just mounted vendor image from NOF27B and there's no `lib/hw/gralloc.msm8996.so` 🤔 Jan 25 22:19:56 There definitely is, in what I see here Jan 25 22:20:14 Aaand there's one in vendor from Pie that I'm running on the Pixel right now :P Jan 25 22:20:32 @Sergobot, I've just flashed it now ... it couldn't be Jan 25 22:20:40 Do I have to flash both slots? Jan 25 22:20:50 I flashed one only Jan 25 22:20:56 depends on which get mounted Jan 25 22:21:02 ^ Jan 25 22:21:12 Only the active one Jan 25 22:21:21 That's slot A Jan 25 22:21:28 And I flashed B Jan 25 22:21:52 I'll switch them Jan 25 22:21:55 You can change slots in fastboot or twrp Jan 25 22:22:04 I did change the slot Jan 25 22:22:14 ah, ok :) Jan 25 22:22:54 I'll check what partition number is vendor_b Jan 25 22:23:57 (Document) https://irc.ubports.com/H8q9FXNY.txt Jan 25 22:24:01 May this assist you Jan 25 22:24:03 I'm in TWRP Jan 25 22:24:08 I did a mount command Jan 25 22:24:11 It's 32 as well Jan 25 22:24:16 Active slot is B Jan 25 22:24:23 Let me see your list Jan 25 22:24:50 @WaseemAlkurdi, Yeah, as it should be Jan 25 22:24:57 And it has the file inside Jan 25 22:25:12 Ohrly 🤔 Jan 25 22:25:32 How about flashing both slots? Jan 25 22:25:43 To eliminate possibilities? Jan 25 22:26:05 can you do in TWRP `cat /vendor/build.prop`? Jan 25 22:26:24 It should show your BUILD ID Jan 25 22:26:53 Whoops, too much caps Jan 25 22:27:12 ro.vendor.build.fingerprint=google/marlin/marlin:7.1.1/NOF27B/3687361:userdebug/dev-keys Jan 25 22:27:41 Then I'm confused :/ Jan 25 22:27:41 I'll do that in Halium now Jan 25 22:31:00 Surprise! Jan 25 22:31:01 ro.vendor.build.fingerprint=google/marlin/marlin:7.1.1/NOF27B/3687361:userdebug/dev-keys Jan 25 22:31:04 Same build ID Jan 25 22:32:53 😞 Jan 25 22:34:33 Okay Jan 25 22:34:38 Now I have no idea Jan 25 22:34:49 Guys ... it's doing something! Jan 25 22:34:57 I tried HYBRIS_LD_PRELOAD=/system/lib/libcutils.so EGL_PLATFORM=hwcomposer test_egl Jan 25 22:35:06 Got this: Jan 25 22:35:07 library "libgui.so" wasn't loaded and RTLD_NOLOAD prevented it Jan 25 22:35:13 And it's still blinking Jan 25 22:35:46 Wait Jan 25 22:36:09 It's supposed to be test_hwcomposer Jan 25 22:36:13 I'll try Jan 25 22:36:52 God, it's really doing something! Jan 25 22:37:00 HYBRIS_LD_PRELOAD=/system/lib/libcutils.so EGL_PLATFORM=hwcomposer test_hwcomposer Jan 25 22:37:06 It's a blinking cursor Jan 25 22:37:11 Ohhh Jan 25 22:37:16 Congrats! :D Jan 25 22:37:27 Nothing on the screen? Jan 25 22:37:56 Didn't check Jan 25 22:37:59 I'll try again Jan 25 22:38:24 No, nothing Jan 25 22:38:38 It's doing the nanosleep thing Jan 25 22:38:46 At this point, I'm really having a deja vu Jan 25 22:39:55 I'm attaching the strace output Jan 25 22:42:46 https://termbin.com/nxfk Jan 25 22:54:19 CAF rootfs, right? Jan 25 22:54:35 No, Pixel isn't CAF Jan 25 22:54:51 Actually Jan 25 22:54:54 ok Jan 25 22:54:56 I quite lost track Jan 25 22:55:06 I'm not sure which rootfs I'm on Jan 25 22:55:15 Do I try non-CAF or CAF? Jan 25 22:55:23 but vendor is correct now it seems Jan 25 22:56:33 Well, I suppose, but since we're not getting anything on the screen ... Jan 25 22:56:43 And it's doing the nanosleep() thing Jan 25 22:56:52 It's probably stalled somewhere Jan 25 23:03:03 @UniversalSuperBox Is initramfs fetched automatically when building halium-boot? Jan 25 23:03:35 @Sergobot From the Internet? Jan 25 23:03:39 I guess not Jan 25 23:03:44 @WaseemAlkurdi, Yea Jan 25 23:03:48 because it was relatively quick Jan 25 23:03:56 If there was a download involved Jan 25 23:04:10 the scrolling of commands would've stopped for a while Jan 25 23:04:11 It's only 4 megabytes Jan 25 23:04:38 And I see automatic releases … https://github.com/Halium/initramfs-tools-halium/releases Jan 25 23:04:50 @Sergobot, So the delay wouldn't be noticeable Jan 25 23:05:36 @UniversalSuperBox What do you say? Jan 25 23:11:10 Yes Jan 25 23:11:19 That was not a reply Jan 25 23:11:41 Yes, it is fetched. Jan 25 23:11:57 Cool Jan 25 23:12:14 So the patch must've been picked up, right Jan 25 23:12:14 @UniversalSuperBox, So it's safe to remove prebuild one from my device repo? Jan 25 23:12:48 Yes, I think I fixed the uploads from Travis Jan 25 23:12:49 @WaseemAlkurdi, Well, there's been an old version of it, unupdated since july Jan 25 23:13:00 ok thanks Jan 25 23:13:37 What do you guys think we should go now? Jan 25 23:15:15