**** BEGIN LOGGING AT Mon Mar 23 02:59:59 2015 Mar 23 08:38:30 regarding Nexus7: I found "could not open egl display" in log from luna-next. Increasing the log-verbosity through the luna-next environment file seems not to work. All services where luna-next depends on are active, even the services that should run before luna-next are active. Maybe someone has a tip for that problem? I will try to look into the source of luna-next this evening... Mar 23 08:45:12 sewziq: can you check /etc/luna-next/environment.conf Mar 23 08:45:19 if there is hwcomposer configured Mar 23 08:45:58 or better just paste me the content somewhere Mar 23 08:46:25 if it looks like this: https://github.com/webOS-ports/meta-webos-ports/blob/dizzy/meta-luneui/recipes-luneui/luna-next/luna-next-conf/environment.conf Mar 23 08:46:34 try replacing fbdev and eglfs with "hwcomposer" Mar 23 09:09:56 morphis: The file looks like the one in the github-link. With replacing fbdev and eglfs with hwcomposer I'm getting now "QPA-HWC: hw_get_module(HWC_HARDWARE_MODULE_ID, (const hw_module_t **)(&hwc_module)) in create returned -2" [lsmod shows no loaded modules] Mar 23 09:10:25 sewziq: hm Mar 23 09:10:34 btw. hwcomposer isn't about kernel modules Mar 23 09:13:04 sewziq: can you paste me the output of /system/bin/logcat? Mar 23 09:27:23 morphis: lsmod was just an idea, because I currently don't have much knowledge about the architecture... Mar 23 09:27:45 output of logcat is here: http://pastebin.com/Q7iCVXLk Mar 23 09:27:57 hope, that's ok... Mar 23 09:33:31 sewziq: can you give me the output of "dmesg" and of "ps axu" too? Mar 23 09:34:11 morphis: everything you want! Give me a few moments. Mar 23 09:34:17 np Mar 23 09:42:07 morphis: dmesg = http://pastebin.com/YybDnzSc Mar 23 09:43:43 ps axu = http://pastebin.com/9pZFwgsb Mar 23 09:45:55 hm Mar 23 09:46:00 the android side is up and running Mar 23 09:46:13 sewziq: lets stick back to fbdev and eglfs Mar 23 09:46:27 when you changed the conf file back to both reboot the device Mar 23 09:46:34 after that call: systemctl stop luna-next Mar 23 09:46:44 then do a test_egl and test_glesv2 Mar 23 09:59:11 morphis: both tests ends with same results: "test_egl.c:52: main: Assertion `eglGetError() == 0x3000' failed" and "test_glesv2.c:104: main: Assertion `eglGetError() == 0x3000' failed" Mar 23 09:59:33 hm Mar 23 09:59:42 we had this in the past, not sure what i solved in the end Mar 23 10:00:23 sewziq: can you try the following: Mar 23 10:00:33 EGL_PLATFORM=fbdev and QT_QPA_PLATFORM=hwcomposer Mar 23 10:02:28 morphis: wasn't this 0x3000 error related to a non supported graphic configuration (i.e. 32bpp for instance) Mar 23 10:03:18 possible Mar 23 10:03:42 Tofe: I was just checking what the Mer guys have configured but there isn't anything special Mar 23 10:03:46 https://github.com/mer-hybris/droid-hal-device/tree/hybris-10.1/device-asus-grouper-configs Mar 23 10:19:08 morphis: QT_QPA_PLATFORM=hwcomposer shows the same error as with both entries = hwcomposer :-( Mar 23 10:20:17 tried it also with QT_QPA_EGLFS_DEPTH=24 and 16 (fbdev and eglfs) same result, both ends in "could not open egl display" Mar 23 10:21:06 hm Mar 23 10:21:38 sewziq: can you join #sailfish-porters and ask the guys there if they saw similar problems with the N7 Mar 23 10:55:25 Just to have more background: the hardware (esp. the display) is used via libhybris? Mar 23 10:55:45 sure Mar 23 10:55:49 right Mar 23 10:56:00 we utilize the EGL, GLESv2 libraries from Android through hybris Mar 23 10:56:13 basically we have also a tiny android system running in an lxc container Mar 23 10:56:35 which does device specific initializations without us having to replicate those under LuneOS Mar 23 10:56:54 hwcomposer is a hardware composer which implements compositing with hardware acceleration Mar 23 10:57:07 we have a qt5 QPA plugin which uses the hwcomposer HAL from Android Mar 23 10:57:26 see http://source.android.com/devices/graphics/index.html Mar 23 11:00:00 would it be ok to mention the luneos project? Mar 23 11:04:37 for sure Mar 23 11:04:51 basically we're using the same components they use Mar 23 11:05:29 Thanks, I'll try! Mar 23 11:46:18 morphis: does e.g. test_hwcomposer pick up the settings from /etc/luna-next/environment.cfg or do I have to provide a paramter to test_hwcomposer? Mar 23 11:58:35 sewziq: just set it as environment variable Mar 23 11:58:42 with export EGL_...=.. Mar 23 12:11:19 morphis: do there is a qt-app, which I can run for test? Mar 23 12:11:58 luna-next is one Mar 23 12:12:25 :-D ok Mar 23 12:17:42 need to find my N7, then I can have a look too ... Mar 23 12:28:20 sorry for not beeing really helpful Mar 23 12:29:37 sewziq: no problem Mar 23 12:29:41 such things are not really easy Mar 23 12:29:45 but I am sure we can solve it Mar 23 13:00:20 morphis: shouldn't I have hwcomposer libs in /system/lib/hw/ ? Mar 23 13:13:43 yeah Mar 23 13:13:53 but those can also be in /vendor Mar 23 13:14:03 do a find /system | grep hwcomposer Mar 23 13:35:09 morphis: there are no hwcomposer libs, even not in /vendor. I copied now a few from the build via adb push and now I end up with a segfault. I think I missed one... Mar 23 13:36:08 there was a question in sailfish, which cm / aosp combination we flash... I think it's cm 4.1, but I currently have no idea what aosp is... Mar 23 13:50:34 sewziq: I think we're on CM 10.1 for both N4 and N7 Mar 23 13:51:47 But morphis can confirm Mar 23 13:52:45 I don't understand this whole "on CM" thing for LuneOS and Sailfish OS. Can someone explain this to me? Mar 23 13:54:07 GodGinrai: We build our system on top of an ultra light Android base from CM team. This mainly takes care of the various firmware bits of components in the devices if I understood correctly. Mar 23 13:54:29 But my low level device knowledge is limited, I'm sure that morphis can explain in great detail. Mar 23 13:54:31 what I've learned so far... there is a cm running in a container to be used for some hw connections, that luneos doesn't need (in the first steps?) to implement the drivers by themselves... Mar 23 13:56:01 GodGinrai: There are currently on Android drivers for HW (legacy had proprietary webOS ones including dual frame buffer etc). In order to utilize this we use libhybris which is basically a bridge to use the Android bits in another kernel/OS the way I unde Mar 23 13:56:01 rstood it Mar 23 13:56:15 So we don't have the chicken/egg driver problem Mar 23 13:58:57 Ok, I was aware of libhybris, but I was under the impression it was just a wrapper for the closed-source binaries Mar 23 13:59:33 I'm sure morphis can elaborate more, all this low level stuff is over my head :P Mar 23 14:18:59 The system seems to miss the libEGL_tegra.so and libGLES_tegra.so though I'm trying to copy them from the source. But for libGLES I have libGLESv1_CM_tegra.so and libGLESv2_tegra.so, which one should I copy? Any ideas? Mar 23 15:01:38 morphis: +all other, thanks for your help today! I've now some homework regarding the N7 ... I will leave now my office and have a look to some things in the evening... Mar 23 15:04:27 sewziq: it does? it shouldn't. Mar 23 15:04:36 seems like there is something wrong with the Android side build Mar 23 15:12:26 morphis: to find the failure, I have to look into the "receipes" I think? The android-build for the N4 seems to work, maybe I could find some differences there...=? Mar 23 15:12:45 sewziq: we're building the android bits outside of OE Mar 23 15:12:55 let me look into that Mar 23 15:17:48 morphis: ok, I'm now away (and currently have no irc client on my portables... Mar 23 15:17:55 sewziq: ok Mar 23 16:42:59 sewziq: looks like I've found the problem Mar 23 16:43:15 atleast for why those files are missing Mar 23 16:44:59 morphis: ah! :-) that's great! Mar 23 16:45:25 however still the question if that fixes the problem we have with graphics Mar 23 16:47:05 I have the fingers crossed! if you have fixed it and you gave me a "Go!", I'll try to update and build and install it, to test if that was the one we are hunting... Mar 23 16:50:05 sewziq: I will give you a diff for meta-smartphone which you need to apply Mar 23 16:50:44 ok! Mar 23 18:39:38 sewziq: https://bpaste.net/show/1e02f52ca602 Mar 23 19:09:00 sewziq: sorry, that one doesn't work Mar 23 19:10:21 morphis: np, I had to start earlier, to backup my build directory... :-/ it will take a while to do a backup before I can try... Mar 23 19:10:46 have you tried it? Mar 23 19:10:47 ok Mar 23 19:10:57 sewziq: https://bpaste.net/show/437db3d0a3a6 Mar 23 19:11:06 sewziq: just building an image with it now locally Mar 23 19:11:10 still didn't found my device Mar 23 19:11:46 ... let me be your "remote executor"... ;-) Mar 23 19:12:19 :D Mar 23 19:13:13 sewziq: but atleast all files are present now Mar 23 19:18:25 morphis: great. I didn't expect a patch today! :-) If I couldnt get through it today, I take my laptop to office tomorrow! Mar 23 19:22:25 Allow me a question, if you have the time... are the directories in meta-smartphone all targets for LuneOS? (I dream of LuneOS for a nexus9 :-) But I doubt, I have the knowledge for it... beside of that, I didn't even have a nexus9 ... Mar 23 19:30:17 morphis: one important question: I do an update, then I do the patch and than bitbake the whole, am I right? Mar 23 20:17:30 sewziq: yeah, but don't need to update Mar 23 20:17:49 sewziq: search for conf/machine Mar 23 20:17:53 in there are targets Mar 23 22:04:23 morphis: good evening... sadly that wasn't all, like you doubt before... luna-next comes now with the error "EGL Error : Could not create the egl surface: error = 0x3009" :-( Mar 23 22:05:17 I'll have a look into it (with the tests I did today) tomorrow... Mar 23 22:19:42 short note: "EGL_PLATFORM=fbdev test_glesv2" creates a graphical output on the display. Mar 23 22:19:58 test_egl outputs: Mar 23 22:21:44 Found EGL_HYBRIS_native_buffer \n Found eglHybrisCreateNativeBuffer \n Stride is 320 \n Found eglHybrisLockBuffer \n locked at 0x41b0f000 \n Memsetting.. \n Found eglHybrisUnlockBuffer \n Found eglCreateImageKHR \n Found eglDestroyImageKHR \n Found eglHybrisReleaseNativeBuffer \n stop \n destroyed context \n destroyed surface Mar 23 23:20:16 sewziq: That looks like progress to me ;) **** ENDING LOGGING AT Tue Mar 24 02:59:58 2015