**** BEGIN LOGGING AT Tue Aug 21 02:59:58 2012 Aug 21 12:20:56 Alex[sp3dev], hi Aug 21 12:35:19 morphis, did you merge fso support for GTA04A4 (the scnearios posted on the ml) Aug 21 12:37:04 GNUtoo: to which ML? Aug 21 12:37:11 let me look Aug 21 12:38:17 [Shr-Devel] SHR hw audio routing Aug 21 12:38:25 I don't have A4 so I can't test Aug 21 12:42:40 GNUtoo: ok, but I don't have the time to validate them Aug 21 12:43:32 ok, if you can push them it would be nice Aug 21 12:56:59 !seen ayaka Aug 21 12:57:03 :P Aug 21 13:08:37 Alex[sp3dev], xmpp disconnected Aug 21 13:11:24 GNUtoo: I said writing raw image also failed. I suppose IOMMU is playing tricks on us. Probably when something touches powervr, the memory is remapped in some weird way and framebuffer is not contiguous Aug 21 13:11:38 oh Aug 21 13:11:49 GNUtoo: I'm now exploring omap4 gralloc in IDA Aug 21 13:11:55 ok Aug 21 13:12:50 what is vrfb? Aug 21 13:13:00 maybe it's activated? Aug 21 13:13:07 where have you found it? Aug 21 13:13:14 vrfb is probably virtual fb Aug 21 13:13:19 or powervr fb :) Aug 21 13:14:21 no Aug 21 13:14:24 GNUtoo: I usually build kernel in-place so .o files are placed where .c files lie Aug 21 13:14:30 I don't have vrfb.o so it is not used Aug 21 13:14:39 it's for rotation and such but if I remember well it's not contiguous Aug 21 13:14:41 and it's in kernel Aug 21 13:14:56 it's part of DSS Aug 21 13:15:17 hmm.. a wild guess. maybe framebuffer is expecting YUV for some reason? Aug 21 13:15:31 ah? Aug 21 13:15:35 possible Aug 21 13:15:46 do you know the values to put? Aug 21 13:15:55 let me look at the headers Aug 21 13:16:04 well, let me study kernel code. if xorg works, it is a bug in android userland Aug 21 13:17:31 HAL_PIXEL_FORMAT_YV12 ? Aug 21 13:17:31 ok Aug 21 13:17:40 Xorg and recovery Aug 21 13:17:41 recovery is at: Aug 21 13:18:06 bootable/recovery/minui/graphics.c Aug 21 13:18:22 and the setting of the Android.mk is in the android board file Aug 21 13:18:27 (android not kernel) Aug 21 13:22:36 GNUtoo: where is your gralloc code? Aug 21 13:22:43 I pastebined it Aug 21 13:22:59 could you do it again? if not, I'll look up the logs Aug 21 13:23:13 yes I'm finding it Aug 21 13:23:22 ok, I found it already Aug 21 13:23:27 pastie.org/private/gpvdgfpjopyxivb8u2kmrg Aug 21 13:23:34 basically I added that in: Aug 21 13:23:46 hardware/libhardware/modules/gralloc Aug 21 13:23:56 it's framebuffer.cpp Aug 21 13:24:08 #cat /sys/class/graphics/fb0/stride 2944 Aug 21 13:24:19 2944 doesn't look like a nice number Aug 21 13:24:22 yes I tried removing the divider too Aug 21 13:24:23 ok Aug 21 13:24:47 stride should be 720? Aug 21 13:25:02 stride should be bytes_per_pixel * width Aug 21 13:25:34 ok Aug 21 13:28:32 so 2880 ? Aug 21 13:29:20 GNUtoo: yeah, should be that. but that value is read-only, I guess it is set up via fb ioctl Aug 21 13:29:31 yes Aug 21 13:29:40 I just tried and it failed with sysfs Aug 21 13:31:46 now it hangs Aug 21 13:31:49 I mean adb dies Aug 21 13:31:54 *died Aug 21 13:32:20 ah it's back Aug 21 13:32:47 hmmm Aug 21 13:33:22 but stride is 2944 Aug 21 13:33:29 *is still Aug 21 13:33:46 I'll boot into Xorg and look Aug 21 13:40:39 GNUtoo: I've connected the hdmi cable and the screen is black, although it shows that the connection is active. Will enable fbconsole and boot with it. If it's the framebuffer config problem, fbconsole should set the correct display settings Aug 21 13:42:13 ok but that's a sysfs away Aug 21 13:42:26 echo "lcd" > /sys/devices/platform/omapdss/manager Aug 21 13:42:46 GNUtoo: no, I don't have fbconsole compiled to kernel Aug 21 13:46:40 ok Aug 21 13:56:16 GNUtoo: huh, my rebuilt kernel doesn't boot. you'll have to wait a bit Aug 21 13:56:47 I guess we need to add fastboot to uboot one day. too complicated to switch kerne without it Aug 21 13:57:56 ok Aug 21 13:59:15 If we add fastboot and omap4 dss init, serial number reading, uboot would be a direct replacement for samsung bootloader Aug 21 14:00:36 wow Aug 21 14:00:56 but I'm more interested in people doing it for sgs2 because of the lack of signature,... Aug 21 14:01:03 even if I've no sgs2 Aug 21 14:07:08 wonderful. it can't mount ramdisk (according to kmsg from uart). Aug 21 14:10:41 GNUtoo: so, fb console is OK, but android is still corrupted Aug 21 14:10:52 ok Aug 21 14:11:03 hmmm Aug 21 14:11:10 I'll try to boot in SHR again Aug 21 14:24:57 hmmm Aug 21 14:25:00 root@tuna:/sys/class/graphics/fb0# cat stride Aug 21 14:25:00 2944 Aug 21 14:25:06 under Xorg Aug 21 14:25:40 GNUtoo: I've made strace of surfaceflinger with omap4 and default graloc Aug 21 14:26:04 ok Aug 21 14:26:11 I tried too and it's strange Aug 21 14:26:18 omap4 gralloc doesn't set fb params, it just queries fb info Aug 21 14:26:18 it loads gralloc.default each time Aug 21 14:26:24 ok Aug 21 14:26:37 it does a lot of ioctls via /dev/pvrsrvkm Aug 21 14:26:41 ah ok Aug 21 14:39:22 how do we proceed now? Aug 21 14:40:37 dunno. by decoding pvr ioctls? Aug 21 14:42:23 hmmm Aug 21 14:42:34 what if we remove pvr from the kernel? Aug 21 14:43:14 GNUtoo: then we won't be able to use proprietary drivers in ubuntu :( Aug 21 14:43:36 no, I mean in replicant Aug 21 14:43:42 Xorg already works Aug 21 14:44:09 with the pvr drivers but without the non-free pvr userspace that uses theses shim drivers Aug 21 14:44:44 GNUtoo: it would be better to try to set it up properly. not that I'm expecting powervr to be reversed soon. though we should implement proper gralloc and hwcomposer Aug 21 14:45:23 what do you mean by properly? Aug 21 14:46:04 GNUtoo: I mean, using pvr ioctl or whatever without modifying kernel side Aug 21 14:46:09 ok Aug 21 14:46:23 but we're not obligated to modify the kernel, again Xorg + recovery works Aug 21 14:46:34 so I suspect an userspace problem in android Aug 21 14:46:45 GNUtoo: the question is who 'spoils' the framebuffer setup Aug 21 14:47:02 I wonder why fbconsole works, but /dev/graphics/fb0 is broken Aug 21 14:47:19 maybe trace recovery ioctls? or xorg Aug 21 14:48:08 yes Aug 21 14:48:14 recovery may be easier Aug 21 14:48:27 how do you trace recovery though? Aug 21 14:48:38 I would run the recovery program in android Aug 21 14:48:53 like minui Aug 21 14:49:07 with surfaceflinger disabled Aug 21 15:05:45 ok I straced it Aug 21 15:05:49 let's pastebin the result Aug 21 15:07:16 http://pastebin.ca/2197068 Aug 21 18:55:22 Alex[sp3dev], hi again Aug 21 18:55:29 GNUtoo: hi Aug 21 18:55:42 did you have a new idea for the gralloc? Aug 21 18:55:53 I'm testing one idea now Aug 21 18:55:56 ok Aug 21 18:57:44 GNUtoo: I have some questions about booting the tuna Aug 21 18:57:48 ok Aug 21 18:58:55 GNUtoo: the thing is, that if you comment out the fbioput call, it kinda works. if you just do a fbioget/fbioput without changing data, it fails. looks like framebuffer struct definition may have changed? Aug 21 18:59:48 ah? Aug 21 19:00:26 GNUtoo: just comment out modifications to the info structure in gralloc, just leave the ioctls and it still fails. comment out the fbioput - and it works, although a bit wrongly still Aug 21 19:00:36 GNUtoo: see PM Aug 21 19:03:55 Alex[sp3dev], ok, let me try Aug 21 19:17:39 Alex[sp3dev], can you pastebin framebuffer.cpp Aug 21 19:17:40 ? Aug 21 19:17:53 GNUtoo: I broke it again :) Aug 21 19:17:58 ah? Aug 21 19:18:04 because for me it doesn't work Aug 21 19:18:10 GNUtoo: I don't have it working atm Aug 21 19:18:28 ok Aug 21 19:18:39 do you have your old somewhat-working changes Aug 21 19:18:59 nope Aug 21 19:19:16 ok Aug 21 20:55:21 GNUtoo: http://rghost.ru/private/39924503/680b7ba734f33c75843e80c00eb4020d/image.png Aug 21 20:55:48 ahhhh ok Aug 21 20:55:57 I knew what android wanted Aug 21 20:56:02 but not what omap wanted Aug 21 20:56:18 and I had a figure like yours Aug 21 20:56:19 once Aug 21 20:56:24 when trying to extract the image Aug 21 20:56:37 GNUtoo: well, I kinda have it working, but it's using twice as much memory and the display is stretched 1/2 times horizontally Aug 21 20:56:44 ok Aug 21 20:56:48 can I try it? Aug 21 20:56:53 yeah, just a while Aug 21 20:56:58 ok Aug 21 20:59:40 http://pastebin.com/EeTsXPfv the only real change is to change xres_virtual. also, the framebuffer seems to ignore format setup, so we change it to bgra. I also changed memset to see which area of display is used Aug 21 21:02:12 thanks a lot!!!!! Aug 21 21:02:21 I guess mirage335 will also be very happy Aug 21 21:02:25 let's try the changes.... Aug 21 21:02:36 guess he'll not be happy Aug 21 21:03:01 first off, it's still broken. secondly, you won't get UI on ICS without GLES2 Aug 21 21:05:08 xres_virtual ... I would never have guessed that. Aug 21 21:06:53 GNUtoo: I've decided to stick with ICS and above, but I'm still looking forward to replicant for the Galaxy Nexus. Aug 21 21:07:12 mirage335, that's for ICS Aug 21 21:07:41 I know. If the UI requires GLES though, that's a problem. Aug 21 21:07:45 Alex[sp3dev]: Nice work! Aug 21 21:07:46 paulk and me will try to make ICS work without hardware GLES Aug 21 21:08:03 GNUtoo: how about reimplementing powervr and fix it once and for all?) Aug 21 21:08:33 that may be harder Aug 21 21:10:17 Alex[sp3dev], when it's ready could you push the commit somewhere so we have your --author and -s inside it? Aug 21 21:10:42 GNUtoo: probably. I want it to work properly first Aug 21 21:10:58 ok Aug 21 21:40:56 GNUtoo, shall I try xorg-video-msm-fbdev, I find X is trying the common fbdev driver Aug 21 21:41:15 ayaka, common fbdev is just slower Aug 21 21:41:27 but if you are able to compile xf86-video-msm why not Aug 21 21:41:34 you need a compatible kernel Aug 21 21:41:36 and the headers Aug 21 21:41:39 etc... Aug 21 21:41:42 it's hard to compile Aug 21 21:41:44 but X -configure will try it but does tell me a succeeful answer Aug 21 21:41:53 ah? Aug 21 21:41:55 xorg-video-msm-fbdev is in debian repos Aug 21 21:42:01 ok Aug 21 21:42:02 try it then Aug 21 21:43:01 X -configure will try to load xorg-video-msm-fbdev but it is can't done( I do it in ssh), it said no screen to configure Aug 21 21:45:08 JaMa, about libphone-ui, is there a config for switching when the headset is inserted (it would look in /dev/input/event*) Aug 21 21:45:27 ayaka, try to write xorg.conf manually Aug 21 21:45:40 else did you install xorg-video-msm-fbdev already? Aug 21 21:46:25 GNUtoo, yes I have installed it, but I wonder whether "no screen to configure" is caused by ssh Aug 21 21:46:39 I don't think so Aug 21 21:46:56 it's rather caused by xorg not finding your screen for some reason Aug 21 21:47:00 try to get the logs Aug 21 21:49:04 GNUtoo, wait a moment Aug 21 21:49:09 I boot it Aug 21 21:52:59 http://pastie.org/4564535 Aug 21 21:53:14 sorry it is No devices to configure. Configuration failed. Aug 21 21:53:48 ok Aug 21 21:53:54 try writing an xorg.conf by hand Aug 21 21:54:04 take an existing one in openembedded Aug 21 21:54:06 for instance: Aug 21 21:54:10 yes sir Aug 21 21:54:53 ah internet is slow here, one minute Aug 21 21:55:47 it doesn't matter, you have gived me before, I can try to search log Aug 21 21:55:59 http://git.openembedded.org/openembedded-core/tree/meta/recipes-graphics/xorg-xserver/xserver-xf86-config/xorg.conf Aug 21 21:56:10 change "intel" with "msm" Aug 21 21:56:15 or msmfb Aug 21 21:56:18 or something like that Aug 21 21:56:31 else there are other here: Aug 21 21:56:35 oh a different one from before Aug 21 21:57:16 http://git.openembedded.org/openembedded/tree/recipes/xorg-xserver/xserver-xorg-conf Aug 21 21:57:29 yes it's more generic Aug 21 21:57:34 more autodetection is done Aug 21 22:23:12 no screens found Aug 21 22:23:27 it seems I don't config correct Aug 21 22:23:45 how does openembedded know so much xorg.conf Aug 21 22:26:50 you mean have? Aug 21 22:27:00 because it's for embedded devices with specific needs Aug 21 22:27:36 for instance you must often tell to use msmfb instead of fbdev Aug 21 22:27:37 etc... Aug 21 22:27:38 I mean how does openembedded find out correct xorg.conf Aug 21 22:27:45 ah Aug 21 22:27:54 with trials and errors Aug 21 22:28:00 people like you or me do it Aug 21 22:28:07 *does it Aug 21 22:28:18 basically we base on previous xorg.conf Aug 21 22:28:23 and we modify to suit the device Aug 21 22:29:11 I see, sadness I know little about it Aug 21 22:30:14 sorry, I forge to use a dri kernel, it seems msm fbdev need dri Aug 21 22:42:52 ok Aug 21 22:43:05 I'll go and push later, bye Aug 21 22:43:11 bye Aug 21 23:35:26 mirage335: fixed the bitch Aug 21 23:38:59 mirage335: https://gitorious.org/i9100-proper-linux/omap4_gralloc/commits/master actually the only really useful thing was fixing stride. gralloc just ignored it. had to fix it to make it round up to a multiple of 32, like the kernel does Aug 21 23:51:49 mirage335: https://wiki.linaro.org/Platform/Android/ICSwithSoftwareGL this also helps Aug 21 23:52:15 UI, browser etc works without opengl binaries. but anything including animations/scrolling is dead slow. Aug 22 00:35:23 lhttp://michelebavaro.blogspot.com/2012/08/the-gps-tracker-to-beat-smallest.html **** ENDING LOGGING AT Wed Aug 22 03:00:00 2012