**** BEGIN LOGGING AT Wed Jul 11 03:00:03 2018 Jul 11 05:51:44 Morning! Jul 11 06:07:23 Morning! Jul 11 09:18:48 o/ Jul 11 16:48:54 * elvispre wonders if Tofe is in a good mood for some reason :) Jul 11 16:54:31 * Tofe wonders if elvispre will be in a good mood in a couple of hours ;) Jul 11 16:55:34 elvispre: but to be franck, I'm not a huge fan of football in general... though, ok, I'll watch the final alright :) Jul 11 16:56:27 I'm not much into it at the moment either. It seems quite popular around here lately though. Jul 11 16:57:56 elvispre: the alarming issues with beer deliveries (because of the lack of CO2) are solved, I hope ? Jul 11 16:59:48 Tofe: All the factories that produce it do so as a by-product. Apparently, they all decided to close down for maintenance at the same time. The problem was caused by a coincidence! At least one of them came back online a week or so ago. Jul 11 17:00:11 pfew ! ok :) Jul 11 17:14:43 morning Jul 11 17:15:41 Tofe: should luna-next RDEPEND on luna-next-cardshell? or have you ever tried some alternative shell with luna-next? Jul 11 17:16:29 I see that it should respect LUNA_NEXT_SHELL env variable, so maybe it's not so hard dependency, but I haven't seen many alternative implementations yet :) Jul 11 17:16:48 JaMa: we never tried any alternative shell, though it should be possible Jul 11 17:17:13 It will probably be even tried with webOS-OSE's cardshell Jul 11 17:17:57 JaMa: you mean, we should drop RDEPENDS and add the cardshell in our packagegroup ? Jul 11 17:18:48 or have a cardshell provider ? Jul 11 17:22:30 it's currently in packagegroup, so when I've added just luna-next to the luneui-example-image (to test vboxvideo again) I was surprised that it exits before drawing anything Jul 11 17:22:44 just after warning: Jul 11 17:22:45 WARNING: 17:17:14.457: Could not find manifest for shell "card" Jul 11 17:23:03 which should probably be fatal, because it's followed by exit Jul 11 17:23:28 now I'm trying to find out how to properly configure EGLDisplay output with vboxvideo Jul 11 17:23:32 yes, just a warning for a compositor failure is quite light... Jul 11 17:24:12 now it's seeing just VGA1: Jul 11 17:24:16 DEBUG: 17:17:14.439: Selected mode 0 : 800 x 600 @ 60 hz for output "VGA1" Jul 11 17:24:16 DEBUG: 17:17:14.439: Physical size is QSizeF(1600, 1200) mm for output "VGA1" Jul 11 17:24:19 DEBUG: 17:17:14.439: Output VGA1 can use 1 planes: 25 Jul 11 17:24:22 DEBUG: 17:17:14.439: No EDID data for output "VGA1" Jul 11 17:24:24 DEBUG: 17:17:14.440: Creating new global GBM mouse cursor Jul 11 17:24:25 but then when trying to use it it fails in drm Jul 11 17:24:27 DEBUG: 17:17:14.440: Sorted screen list: QVector(OrderedScreen(QPlatformScreen=0x8090758 ("VGA1") : 2147483647 / QPoint(0,0) / primary: false)) Jul 11 17:24:30 DEBUG: 17:17:14.440: Adding QPlatformScreen 0x8090758 ( "VGA1" ) to QPA with geometry QRect(0,0 800x600) and isPrimary= false Jul 11 17:25:00 JaMa: I can add the RDEPENDS for now, if it helps Jul 11 17:25:21 CRITICAL: 17:25:07.543: Could not queue DRM page flip on screen VGA1 (Invalid argument Jul 11 17:26:07 that's ok, I've already added the cardshell to image for testing Jul 11 17:26:38 ok Jul 11 17:26:44 http://code.qt.io/cgit/qt/qtbase.git/tree/src/plugins/platforms/eglfs/deviceintegration/eglfs_kms/qeglfskmsgbmscreen.cpp?h=5.11#n283 <-- comes from here Jul 11 17:26:58 the same is with VBoxTouch plugin Jul 11 17:27:14 what is that "drmModePageFlip" doing... Jul 11 17:28:18 age-Flipping Jul 11 17:28:19 A call to drmModeSetCrtc(3) is executed immediately and forces the CRTC to use the new scanout buffer. If you want smooth-transitions without tearing, you probably use double-buffering. You need to create one framebuffer object for each buffer you use. You can then call drmModeSetCrtc(3) on the next buffer to flip. If you want to synchronize your flips with vertical-blanks, you can use Jul 11 17:28:25 drmModePageFlip(3) which schedules your page-flip for the next vblank. Jul 11 17:28:45 will need to add -dbg symbols to see what exactly it's doing in this case Jul 11 17:28:56 mmh ok, but then do you see an error in dmesg ? Jul 11 17:29:28 lots of "[ 55.359268] vbox_deferred_io: Calling dirty update: 0, 503, 800, 97" but those aren't from this Jul 11 17:29:38 these are from running weston on vboxvideo before Jul 11 17:29:43 ok Jul 11 17:32:01 which drm driver are you using ? Jul 11 17:33:02 vboxvideo Jul 11 17:33:03 I'm so uncomfortable with DRM/KMS... I never really looked into it Jul 11 17:33:34 I'll try with virtglrenderer next Jul 11 17:34:21 I wanted to start with something which we currently use to eliminate other changes, but might be that DRM implementation in vboxvideo isn't complete or is strange in some aspect Jul 11 17:34:44 the vboxvideo included in staging vanilla kernels was crashing the kernel as soon as it was loaded.. Jul 11 17:35:47 and it's possible that weston works with vboxvideo only because it's using only the fbdev implementation from it Jul 11 17:36:09 which brings me to next test to configure weston to use drm backend to see if it fails the same as QCompositor Jul 11 17:37:54 you could try enabling DRM_DEBUG in the kernel too, though I don't know how exactly Jul 11 17:39:06 drm.debug=0x3f in the kernel cmdline Jul 11 17:39:33 (or just drm.debug=0x1 if it's too much) Jul 11 17:40:14 (see https://github.com/torvalds/linux/blob/6b25e21fa6f26d0f0d45f161d169029411c84286/include/drm/drmP.h#L121 ) Jul 11 17:40:46 ok, will try, for now I was trying to change all DRM in kernel to modules e.g. kms_helper, but linux-yocto is fighting with me Jul 11 17:40:57 oooh it's even possible at runtime "echo 0xf > /sys/module/drm/parameters/debug" Jul 11 17:41:18 even if I set it to modules in our .bbappend it's overriden from 2 .scc files in kernel-meta, which are included by default by some linux-yocto KMACHINE Jul 11 17:41:45 ah, I see... Jul 11 17:41:47 so I cannot just easily override them in own defconfig, <- I hate linux-yocto and might be using it incorrectly :) Jul 11 17:42:52 I guess it's documented somewhere in the 500 pages pdf for kernel build in yocto Jul 11 17:44:18 :) Jul 11 17:45:43 another strange thing is that I have 2 similar (now should be identical AFAIS in VirtualBox UI) machines and one of them shows the syslinux menu immediately and the other one draws it terribly slowly line after line Jul 11 17:48:14 with the drm debug enabled I see Jul 11 17:48:15 [ 65.933429] [drm:drm_ioctl] pid=678, dev=0xe200, auth=1, DRM_IOCTL_MODE_MAP_DUMB Jul 11 17:48:18 [ 65.949808] [drm:drm_ioctl] pid=676, dev=0xe200, auth=1, DRM_IOCTL_MODE_PAGE_FLIP Jul 11 17:48:21 [ 65.949814] [drm:drm_ioctl] ret = -22 Jul 11 17:49:11 checking vboxvideo code Jul 11 17:56:39 hmm -EINVAL can have couple reasons in DRM_IOCTL_MODE_PAGE_FLIP https://github.com/torvalds/linux/blob/6b25e21fa6f26d0f0d45f161d169029411c84286/drivers/gpu/drm/drm_plane.c#L764 Jul 11 18:00:26 that's with debug=0x3f ? Jul 11 18:07:42 with 0xf Jul 11 18:07:49 but looks the same with 0x3f Jul 11 18:09:54 .driver_features = Jul 11 18:09:54 DRIVER_MODESET | DRIVER_GEM | DRIVER_HAVE_IRQ | DRIVER_IRQ_SHARED | Jul 11 18:09:57 DRIVER_PRIME, Jul 11 18:10:56 https://github.com/torvalds/linux/blob/master/drivers/staging/vboxvideo/vbox_drv.c#L225 Jul 11 18:11:12 ah, ok :( Jul 11 18:11:35 that's assuming that latest driver in staging has the same features as vboxvideo from 5.1.18-r0 vboxguestadditions Jul 11 18:11:55 should probably try there or test with 4.17 kernel Jul 11 18:12:15 kernel debugging is always very annoying Jul 11 18:13:00 will switch to qemu and virtio which I know that it works from OSE Jul 11 18:13:19 and will do the test with weston as well Jul 11 18:13:47 it's annoying that there are too many moving parts and I cannot change them one by one :) Jul 11 18:14:29 new mesa, uvesafb->vboxvideo->virtio-gpu, example image instead of full one, weston instead of lune-next etc :) Jul 11 18:14:46 but at least I've "returned" to pyro release to be a bit closer :) Jul 11 18:15:34 before I was testing with latest greatest including gcc8 the vboxvideo from latest kernel etc, which added even more possible roads to hell Jul 11 18:27:17 yes it's hard to know what's causing the trouble then Jul 11 18:45:45 hmm weston works ok with drm backend on vboxvideo even when it shows the same error from page flip https://paste.pound-python.org/show/MSNRgwuY12nbfKk1vdHd/ Jul 11 19:36:40 Tofe: at least the slow syslinux UI is resolved, it was because of enabled uart->tcp in vbox Jul 11 19:52:19 Tofe: there is another runtime dependency, which is a bit more complicated Jul 11 19:52:36 ^, file:///usr/palm/luna-next/shells/card/qml/LunaSysAPI/Preferences.qml:20:1: module "LuneOS.Service" is not installed Jul 11 19:52:39 import LuneOS.Service 1.0 Jul 11 19:52:42 ^) Jul 11 19:53:01 and LuneOS.Service I've found in luneos-components, but this depends on qtwebengine which I would rather not build yet for these tests Jul 11 19:53:17 would it make sense to split luneos-components a bit more? Jul 11 19:57:07 hmm luneos-components just built ok without qtwebengine Jul 11 19:58:19 JaMa: The LuneOS Components should have a rdepends I guess, not a hard depends Jul 11 19:58:55 It's mainly for https://github.com/webOS-ports/luneos-components/blob/master/modules/LuneOS/Components/LunaWebEngineView.qml if I recall correctly Jul 11 19:59:32 And https://github.com/webOS-ports/luneos-components/blob/81d9a85aff74d8b3acc367c659743f29f9083168/modules/LuneOS/Components/PermissionDialog.qml Jul 11 20:06:17 And the LunaWebEngineVIew is used by FirstUSe, Browser and luna-webappmanager Jul 11 20:13:00 ok, then I'll drop it Jul 11 20:14:08 there is another missing RDEPENDS in luna-cardshell Jul 11 20:14:09 Jul 11 20:13:20 qemux86 luna-next[604]: ^, file:///usr/palm/luna-next/shells/card/qml/BootLoader.qml:22:1: module "QtMultimedia" is not installed Jul 11 20:14:12 Jul 11 20:13:20 qemux86 luna-next[604]: import QtMultimedia 5.2 Jul 11 20:15:32 ah right Jul 11 20:16:23 I didn't remember we had that much trouble, I'll fix that Jul 11 20:17:05 I'm already fixing it Jul 11 20:17:14 oh, ok, thanks :) Jul 11 20:17:35 JaMa: Weird these never popped up before Jul 11 20:17:58 because normally it's pulled by packagegroups Jul 11 20:18:07 but I'm doing the minimal luneui-example-image Jul 11 20:18:10 Ah ok Jul 11 20:18:23 where I've added just lune-next and now I'm finding out what's missing to make it usable :) Jul 11 20:40:39 interesting I have pulsing LuneOS logo on background with few rendering artifacts but luna-next.service didn't fail this time (still complaining a lot about page flips) Jul 11 20:45:50 ah, good Jul 11 20:53:42 I've pushed current messy changes to jansa/pyro-vbox branch, will build full image overnight and we'll see Jul 11 20:54:12 next week wife will probably be away with kid on vacation with her parents, so hopefully I'll have more time for LuneOS :) Jul 11 20:54:42 :p **** ENDING LOGGING AT Thu Jul 12 03:00:01 2018