Additional information:
We build XBMC against Mer in COBS, and this XBMC can work in meego 1.2.0
rootfs(boot from hi3716m, chroot to meego and run XBMC).
2012-01-04
zhanghui
发件人： zhanghui
发送时间： 2012-01-04 14:58:33
收件人： mer-general
抄送：
主题： Re: Re: [mer-general] cannot start Mer-XBMC on hi3716m
Hi Carsten,
Thanks for the reply!
(1) /proc/cpuinfo is:
Processor : ARMv7 Processor rev 2 (v7l)
BogoMIPS : 1199.30
Features : swp half thumb fastmult vfp edsp vfpv3 vfpv3d16
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x2
CPU part : 0xc09
CPU revision : 2
Hardware : godbox
Revision : 0000
Serial : 0000000000000000
(2) Before further disscussion, I explain something first:
(a) The reason we test a platform without 3D hardware is: A lot of STBs
will be free to family users, and the government will pay for it. Since the
STBs are free, the cost should be low. So the hi3716m comes.
Of course, many TV chips and other STB chips do have 3D hardware.
(b) Since MeeGo-TV has chosen XBMC as its reference UI, We are considering
it for our TVOS reference UI.
(3) We need suggestion now. whether we wait for LLVMpipe, or just give up XBMC
on platform without 3D? Would you please suggest another UX we can try?
(4) Can we just comment out some code(or change some configuration) that
heavily depends on 3D hardware, then XBMC can be acceptable on platforms
without 3D?
Thanks a lot!
2012-01-04
zhanghui
发件人： Carsten Munk
发送时间： 2012-01-04 14:27:34
收件人： mer-general
抄送：
主题： Re: [mer-general] cannot start Mer-XBMC on hi3716m
4. jan. 2012 03.35 skrev zhanghui <[email protected]>:
> Hi everyone,
> We are testing Mer-XBMC on our hi3716m platform(this platform is for
> STB/TV), and we cannot start XBMC for the moment. Our steps are:
> (1) download Sage's
> http://sage.kapsi.fi/Mer/images/0.20111120.1.CE.2011.11.20.1/mer-tv-armv7l-pandaboard/,
> and test it on Pandaboard, everything is OK.
> (2) Copy rootfs from mer-tv-armv7l-pandaboard Image to our nfs server:
> 192.168.129.76/home/zhangh/nfs/mer-tv-rootfs
> (3) Boot our hi3716m(ARM Cortex A9, with 2D hardware, without 3D hardware)
> platform.
> (4) copy some necessary files from hi3716m-rootfs to mer-tv-rootfs,
> including /lib/modules/2.6.35.7_hi3716m etc...
> (5) In order to use hi3617m's framebuffer, delete
> mer-tv-rootfs/usr/share/X11/xorg.conf.d/99-pvr.conf, create
> mer-tv-rootfs/etc/X11/20-hifb.conf:
> Section "Screen"
> Identifier "Default Screen"
> Device "Configured Video Device"
> DefaultDepth 16
> SubSection "Display"
> Depth 24
> # Modes "1280x720"
> EndSubSection
> EndSection
>
> Section "Device"
> Identifier "Configured Video Device"
> Driver "fbdev"
> Option "fb" "/dev/fb0"
> EndSection
>
> (6) execute following script to prepare for CHROOT:
> MEEGO_ROOT=/mnt/
> mount -t nfs -o nolock -o tcp 192.168.129.76:/home/zhangh/nfs/mer-tv-rootfs
> ${MEEGO_ROOT}
> mount -o bind /proc ${MEEGO_ROOT}/proc
> mount -o bind /sys ${MEEGO_ROOT}/sys
> mount -o bind /dev ${MEEGO_ROOT}/dev
> mount -o bind /dev/pts ${MEEGO_ROOT}/dev/pts/
>
> (7) #chroot /mnt /bin/bash
> (8) Uninstall all pvr drivers of pandaboard, and Install all mesa releated
> drivers.
> mesa releated drviers are downloaded from
> http://releases.merproject.org/releases/latest/builds/armv7l/packages/
> glx-utils are downloaded from
> http://repo.pub.meego.com/CE:/Utils/Mer_Core_armv7l/
Hi,
This is where you go wrong :) GLX is for OpenGL, and on ARM devices
it's typically OpenGL ES instead. When running glxgears, you're in
fact using the software rendering for GL.
I'd try to use http://www.daimi.au.dk/~cvm/glestest.cc if I were you
instead - http://www.daimi.au.dk/~cvm/glestest is a armv7l binary,
chmod +x it and ./glestest
I'm a little unsure if the EGL/GLESv2 -> GLX -> swrast works properly
in Mesa, which may be why it's crashing. I know that Sage's XBMC is
built for EGL/GLESv2 for sure.
We have some work ongoing to support Mesa with LLVMpipe, which is a
kind of fast software rendered EGL/GLESv2, utilizing the CPU instead
of a GPU, but the LLVM JIT has problems on ARM chips currently, while
it works fine on X86.
XBMC would also require EGL/GLESv2 support and is very slow with
software rendering, an actual chip is recommended for that - it was
even slow on my 2.3ghz X86 with swrast.
>
> (9) Execute the following commands:
> #Xorg &
> #export DISPLAY=":0.0"
> #glxgears
> glxgears can work fine!
>
> (10) #xbmc &
> XBMC cannot work with following error:
> /usr/bin/xbmc: line 133: 1064 Illegal instruction
> "$LIBDIR/xbmc/xbmc.bin" $SAVED_ARGS
> Crash report available at /root/xbmc_crashlog-19691231_161444.log (See
> attachment)
> In the crashlog, We can confirm that XBMC have used Mesa 3D drivers, Not pvr
> 3D drivers. What's wrong then??
I'd like to see /proc/cpuinfo since this is a new chipset we haven't
tested Mer on (we've tested on Tegra2, OMAP3/4, Marvell, ST-Ericsson,
Freescale, etc, but not this one).
>
> Any help or instruction is appreciated!
> Thanks in advance!
BR
Carsten Munk
>
>
>
> 2012-01-04
> ________________________________
> zhanghui
> ________________________________
> 发件人： Carsten Munk
> 发送时间： 2012-01-03 01:26:16
> 收件人： mer-general
> 抄送：
> 主题： [mer-general] Available tasks to do, 2/1/2012
> As promised, I will start posting weekly 'task' bug lists, If you see
> a bug/task
> that you'd like to work on, assign the bug to yourself within the Mer
> bugtracker. The bugtracker will bother you on email if you don't touch
> the bug within seven days. If you have any questions, this mailing
> list or #mer IRC channel is open to get help within :)
> Over time, we might start meetings to establish priority of tasks as well.
> The currently available tasks for taking are:
> bug - priority - short_desc
> http://bugs.merproject.org/show_bug.cgi?id=46 Low mpc has possible
> bashisms in it's RPM scriptlets
> http://bugs.merproject.org/show_bug.cgi?id=47 Low nano has possible
> bashisms in it's RPM scriptlets
> http://bugs.merproject.org/show_bug.cgi?id=48 Low openvpn has possible
> bashisms in it's RPM scriptlets
> http://bugs.merproject.org/show_bug.cgi?id=49 Low qt has possible
> bashisms in it's RPM scriptlets
> http://bugs.merproject.org/show_bug.cgi?id=38 Low binutils,
> binutils-devel has a possible bashism in RPM scriptlets
> http://bugs.merproject.org/show_bug.cgi?id=50 Low time has possible
> bashisms in it's RPM scriptlets
> http://bugs.merproject.org/show_bug.cgi?id=39 Low cvs has possible
> bashisms in it's RPM scriptlets
> http://bugs.merproject.org/show_bug.cgi?id=51 Low glibc-devel has
> possible bashisms in it's RPM scriptlets
> http://bugs.merproject.org/show_bug.cgi?id=41 Low libassuan-devel has
> possible bashisms in it's RPM scriptlets
> http://bugs.merproject.org/show_bug.cgi?id=42 Low libffi-devel has
> possible bashisms in it's RPM scriptlets
> http://bugs.merproject.org/show_bug.cgi?id=43 Low libtasn1-devel has
> possible bashisms in it's RPM scriptlets
> http://bugs.merproject.org/show_bug.cgi?id=44 Low m4 has possible
> bashisms in it's RPM scriptlets
> http://bugs.merproject.org/show_bug.cgi?id=45 Low make has possible
> bashisms in it's RPM scriptlets
> http://bugs.merproject.org/show_bug.cgi?id=29 Undecided Drop gamin from core
> http://bugs.merproject.org/show_bug.cgi?id=69 Undecided PAM should use
> system db4, not it's own
> http://bugs.merproject.org/show_bug.cgi?id=30 Undecided bash should
> link against system readline, not against it's own copy
> http://bugs.merproject.org/show_bug.cgi?id=70 Undecided e2fsprogs
> should be upgraded to 1.42
> http://bugs.merproject.org/show_bug.cgi?id=31 Undecided Move ntpdc and
> ntpq to ntp-utils package to save footprint
> http://bugs.merproject.org/show_bug.cgi?id=54 Undecided Include
> package build logs to the releases
> http://bugs.merproject.org/show_bug.cgi?id=35 Undecided pixman should
> be upgraded to 0.24
> http://bugs.merproject.org/show_bug.cgi?id=15 Undecided Switch from
> libjpeg to libjpeg-turbo
> http://bugs.merproject.org/show_bug.cgi?id=19 Undecided Drop GL
> support from Mer Core
> http://bugs.merproject.org/show_bug.cgi?id=52 Undecided Switch from
> DWARF-3 to DWARF-4 in order to save debuginfo space
> http://bugs.merproject.org/show_bug.cgi?id=20 Undecided Evaluate
> necessity of consolekit
> http://bugs.merproject.org/show_bug.cgi?id=53 Undecided Switch db4 to
> libdb, 5.2.36 and switch RPM to it
> http://bugs.merproject.org/show_bug.cgi?id=26 Undecided Evaluate dash
> as /bin/sh vs /bin/bash, disk/memory footprint/boot time/impact
> http://bugs.merproject.org/show_bug.cgi?id=57 Undecided dbus-glib is
> now merged to glib upstream so we should remove it
> http://bugs.merproject.org/show_bug.cgi?id=27 Undecided Evaluate
> benefits eglibc vs glibc, in terms of packaging, cross compilation and
> footprint/impact
> http://bugs.merproject.org/show_bug.cgi?id=58 Undecided nss should be
> upgraded to NSS_3_13_1_RTM
> http://bugs.merproject.org/show_bug.cgi?id=28 Undecided
> Drop gamin dependancy
> http://bugs.merproject.org/show_bug.cgi?id=59 Undecided nspr should be
> upgraded to 4.8.9
> BR
> Carsten Munk
