**** BEGIN LOGGING AT Tue Jan 17 03:00:03 2017 Jan 17 06:00:07 morning Jan 17 07:54:47 Morning! Jan 17 08:21:38 ok, kernel built Jan 17 08:26:11 granpc: you're building for an aarch64 target? Jan 17 08:34:14 Ah, I forgot to tweak nyx-modules Jan 17 08:34:44 That should have been in meta-smartphone, for the machine specific part Jan 17 09:39:18 Tofe: That's still solid progress :) Jan 17 09:39:27 Kernel already built :) Jan 17 09:51:22 same remark for luna-next-conf: the machine specific files should belong to .bbappend in meta-smartphone Jan 17 10:08:22 Herrie|Pre3, JaMa: is meta-smartphone supposed to be independant of meta-webos-ports? Jan 17 10:13:39 Tofe: Not sure I follow. JaMa gave some remarks? Jan 17 10:14:20 Tofe: AFAIK meta-smartphone could theoretically be used for multiple target OS-es not only LuneOS. Jan 17 10:27:26 ok, so my nyx-module bbappend thing might not be correct, in this case Jan 17 10:37:59 Tofe: Anything LuneOS specific should be in meta-webos-ports but then again kernel defconfig is also LuneOS specific and in meta-smartphone I think :P ? Jan 17 10:40:27 Well, the proposed defconfig could very well work for other OS-es. Jan 17 10:41:13 However, some other bits are very specific to us: fstab, initrd... Jan 17 10:46:49 Tofe: Yeah it's not set in stone, we're anyway the only ones using meta-smartphone atm afaik. Jan 17 10:47:36 I guess we should ask JaMa when he's around, imho all LuneOS specific bit are better suited in our own layer. Jan 17 12:02:47 I guess we probably shouldn't put any recipe dependency between the two layers; but even in that case, I'd like to rework nyx-module, to have the machine specific bits outside of the main source code Jan 17 12:02:54 A bit like luna-next-conf Jan 17 12:07:03 Tofe: Yeah that would make sense I guess. Jan 17 12:11:05 And it's completely feasible: I even did a .bbappend that worked (see bottom of https://gist.github.com/Tofee/f3a65d5ea391da20817d3746f0031357 ), so a simple .bb should be no problem Jan 17 14:08:54 tadaa, I've got a first rootfs Jan 17 14:09:08 Probably unbootable I guess, but still Jan 17 14:13:48 Tofe: Nice progress! Jan 17 14:17:20 All went very well, lucky me Jan 17 14:35:12 Tofe: Seems it's pretty straight forward! Jan 17 14:36:09 Tofe: Seems Nylas Mail Basic is now free & Open Source. We might want to give that a go sometime to see if we can get that working :) Jan 17 14:36:19 Has Exchange support too :) Jan 17 14:36:46 yes, pretty much; but I have a good knowledge of the thing too ;) Jan 17 14:39:35 Echange support, for free ? Isn't that weird :) Jan 17 14:42:23 Herrie|Pre3: do you have an url ? I only find the desktop version; or maybe I misunderstood? Jan 17 14:47:18 Tofe: Hmmz seems you might be right.... Jan 17 14:48:13 Well it uses Electron, Flux & React which is all web. So should be possible to pack it as an app :P Jan 17 14:48:55 They also seem to have nylas-nodejs which we might use as some Synergy connector for Email, Contacts and Calendar :) Jan 17 15:26:18 Tofe: yep, it's aarch64 Jan 17 15:26:21 which I thought would make everything a lot lot lot harder Jan 17 15:26:24 but it's actually just a bit trickier Jan 17 15:27:17 oh ok, I would have thought like you, but then it's good news Jan 17 15:28:03 I don't think we built an image for this architecture before Jan 17 15:28:03 something frustrating happened though Jan 17 15:28:03 let me show you real quick Jan 17 15:28:16 yep Jan 17 15:28:33 so with the hal-droid thing Jan 17 15:28:36 test_sensors works fine Jan 17 15:28:41 but test_hwcomposer broke Jan 17 15:28:52 if I "disable" hal-droid, test_sensors crashes again, but test_hwcomposer starts working? Jan 17 15:28:54 http://paste2.org/79FMCb6G Jan 17 15:29:07 I sorta suspect it's because of the get_tls_hooks thing, but I'm not sure Jan 17 15:32:55 well, hybris's job is to avoid conflicts between bionic and glibc implementations; if you don't have any program using the video driver, it might be that it just works Jan 17 15:33:30 but it's still a bit weird, yes Jan 17 15:34:32 anyway, I decided I'd try to start up luna-next anyway Jan 17 15:34:34 because, why not Jan 17 15:34:44 exactly :) Jan 17 15:34:59 but it doesn't seem like the mali drivers were too happy about whatever surface format qt is asking for Jan 17 15:35:13 W/mali_winsys( 0): EGL_BAD_MATCH:pixel format (surface:0x5 config:0x4) Jan 17 15:35:13 FATAL: 21:12:33.423: EGL Error : Could not create the egl surface: error = 0x3009 Jan 17 15:35:37 tried fiddling with eglfs_depth but that didn't help Jan 17 15:39:11 where is your repo for the libhardware part of CM12.1 ? Jan 17 15:40:29 seems to be mer-hybris/android_hardware_libhardware Jan 17 15:41:05 oh, ok Jan 17 15:45:25 it's probably a rgb/bgr thing Jan 17 15:45:26 when I ran test_hwcomposer, the spiral thingy was blue instead of orange Jan 17 15:45:26 HAL_PIXEL_FORMAT_RGB_565 = 4, HAL_PIXEL_FORMAT_BGRA_8888 = 5, it seems Jan 17 15:45:26 oh thanks, I was trying to find that yesterday! Jan 17 15:45:55 so supposedly if you use hwcomposer instead of fbdev, it'll automatically convert or whatever Jan 17 15:46:03 can you check what you have in your system/includes/graphics.h and compare to http://bazaar.launchpad.net/~ubuntu-branches/ubuntu/vivid/android-headers/vivid/view/head:/system/graphics.h ? Jan 17 15:46:36 I can't see Mediatek changing that, though. Jan 17 15:47:27 there's a few changes Jan 17 15:47:30 but it seems to be mostly for camera stuff Jan 17 15:48:33 ok Jan 17 15:48:52 is there any way to change the pixel format luna-next uses? Jan 17 15:48:56 I'm sorta curious to see if it would work Jan 17 15:49:00 even if the colors look wonky Jan 17 15:50:10 Here are some things you can tweak: https://github.com/webOS-ports/meta-webos-ports/blob/master/meta-luneui/recipes-luneui/luna-next/luna-next-conf/mako/environment.conf Jan 17 15:51:12 You can try setting QT_QPA_EGLFS_DEPTH to 16 Jan 17 15:51:22 I tried fiddling with these Jan 17 15:51:30 but it always seems to pass "surface:0x5 config:0x4" Jan 17 15:51:38 mmh Jan 17 15:51:48 I'm setting them in the command line to luna-next manually Jan 17 15:51:54 I suppose that should work the same Jan 17 15:52:50 yes, it should Jan 17 15:53:15 you are using hwcomposer? Jan 17 15:53:19 fbdev Jan 17 15:53:30 hwcomposer always crashes Jan 17 15:53:33 ah, ok Jan 17 15:53:41 not quite sure why Jan 17 15:53:55 in test_hwcomposer it doesn't crash, but it only shows a black screen Jan 17 15:56:49 I'll have to leave you here until tomorrow Jan 17 15:56:53 that's ok Jan 17 15:57:03 thank you for your help! hopefully we'll figure this out some time Jan 17 15:57:10 :) I'm sure we will Jan 17 18:36:27 neat Jan 17 18:36:29 https://imgroll.net/images/2017/01/17/photo_2017-01-17_19-18-54.jpg Jan 17 18:36:44 performance is pretty bad and r/b are swapped of course Jan 17 18:36:48 I wonder if hwcomposer would fix both things Jan 17 18:42:36 granpc: NICE! Jan 17 18:43:01 hey thanks! Jan 17 18:43:37 Yeah colors look swapped but it's progress Jan 17 18:44:41 yeah, I think if I can somehow get the hwcomposer implementation working Jan 17 18:44:48 colors should be ok Jan 17 19:24:03 granpc: great! Bravo! Jan 17 19:24:27 The performances are due to using fbdev, yes, which isn't accelerated at all Jan 17 19:24:54 for rgba, not sure... Jan 17 19:26:13 On my side, I've just flashed the rootfs Jan 17 19:26:26 there is no UI, but I can see the device on my host machine Jan 17 19:26:46 I cannot adb into it though; adb is probably not configured correctly Jan 17 20:12:52 Tofe: OK well it's not that bad, no UI could be plenty of things Jan 17 20:12:58 Like some small config Jan 17 20:13:08 For ADB you need some tweaking too I think let me check Jan 17 20:13:28 Herrie: no actually I think I spoke too soon Jan 17 20:13:44 It didn't boot, I was just in a very weird mode Jan 17 20:13:49 Ah OK Jan 17 20:14:05 I managed (after many retries) to reboot the device to bootloader Jan 17 20:14:23 and I think what happened is that the zImage I provided isn't correct Jan 17 20:14:40 It should have a "DTB" - I just learned about it :) Jan 17 20:14:59 So I have to make so that bitbake does make zImage-dtb instead of zImage Jan 17 20:22:21 Found it: KERNEL_IMAGETYPE="zImage-dtb" Jan 17 20:23:37 :D Jan 17 20:27:22 ooops, I think I erase the "aboot" partition... Jan 17 20:27:50 I'll do a full factory reflash, can't hurt. Jan 17 20:30:30 ok yes, I did erase the bootloader itself. I'm so clever! Jan 17 20:31:09 hopefully, there's a backup bootloader on the N5, and I found the key combination (Vol- + Vol+ + Power for some time) to boot into it Jan 17 20:35:19 Here's a guy as clever as me: https://forum.xda-developers.com/google-nexus-5/general/bootloader-bootloader-nexus-5-t2746343 Jan 17 20:52:42 ok, doesn't boot. no adb, nothing. :( Jan 17 21:01:22 Tofe: OK well but bootloader is back at least? Jan 17 21:01:30 Good thing is N5 is documented at least :P Jan 17 21:01:45 And it's not some obscure device LOL Jan 17 21:02:30 yes, all is back to "normal" now Jan 17 21:02:37 I've been lucky... Jan 17 21:04:17 Tofe: Well no ADB you can still get the log from CWM like you told me for Mako? Jan 17 21:04:21 Might give some ideas? Jan 17 21:05:16 no, I don't have any last_kmsg here Jan 17 21:05:27 which isn't normal at all, there must be a mistake in the defconfig Jan 17 21:05:53 But I'll check that tomorrow, I'm a bit tired tonight Jan 17 21:06:33 Tofe: Yeah sleep does miracles sometimes :D Jan 17 22:34:19 how does qt's "qeglfsdeviceintegration.cpp" thing work? Jan 17 22:34:38 I'm trying to figure out why it doesn't work if I use hwcomposer Jan 17 22:34:50 and it turns out that the "createNativeWindow" function always returns 0 **** ENDING LOGGING AT Wed Jan 18 03:00:00 2017