bitbake virtual/kernel fails for davinci

bitbake virtual/kernel fails for davinci

[Bjørn Forsman]

Hi,

I am a bit new to OpenEmbedded, but have successfully built angstrom
for davinci and qemu in the past. Today I tried to "bitbake
virtual/kernel" for a davinci (dm6446) system. The build fails during
unpack of linux-davinci kernel because git cannot find a given object
hash (see output below). Has anybody else experienced this? How come
OE references a non-existent git object?

The non-existent git object hash is specified in
"recipes/linux/linux-davinci_git.bb" and was lastly modified by commit
160616d5f2a57ba828ed5ec06e37fc1cd4713af7.

---------------------------------------------------------------
$ bitbake virtual/kernel

[snipped]

From git://git.kernel.org/pub/scm/linux/kernel/git/khilman/linux-davinci
 * [new tag]         v2.6.29    -> v2.6.29
 * [new tag]         v2.6.29-davinci1 -> v2.6.29-davinci1
 * [new tag]         v2.6.29-omap1 -> v2.6.29-omap1
 * [new tag]         v2.6.29-rc5 -> v2.6.29-rc5
 * [new tag]         v2.6.29-rc6 -> v2.6.29-rc6
 * [new tag]         v2.6.29-rc7 -> v2.6.29-rc7
 * [new tag]         v2.6.29-rc8 -> v2.6.29-rc8
 * [new tag]         v2.6.30    -> v2.6.30
 * [new tag]         v2.6.30-davinci1 -> v2.6.30-davinci1
 * [new tag]         v2.6.30-rc1 -> v2.6.30-rc1
 * [new tag]         v2.6.30-rc2 -> v2.6.30-rc2
 * [new tag]         v2.6.30-rc3 -> v2.6.30-rc3
 * [new tag]         v2.6.30-rc4 -> v2.6.30-rc4
 * [new tag]         v2.6.30-rc5 -> v2.6.30-rc5
 * [new tag]         v2.6.30-rc6 -> v2.6.30-rc6
 * [new tag]         v2.6.30-rc7 -> v2.6.30-rc7
 * [new tag]         v2.6.30-rc8 -> v2.6.30-rc8
NOTE: Creating tarball of git repository
fatal: failed to unpack tree object 03022e38318bef016836dae71c70b9ea975723e7
NOTE: Task failed: Fetch failed: Fetch command export
HOME=/home/bjornfor; export
SSH_AUTH_SOCK=/tmp/keyring-f2gNsd/socket.ssh; export
PATH=/home/bjornfor/dev/oe-stuff/build/tmp/staging/i686-linux/usr/bin/dm6446-evm-angstrom-linux-gnueabi:/home/bjornfor/dev/oe-stuff/build/tmp/staging/i686-linux/usr/bin/armv5te-angstrom-linux-gnueabi:/home/bjornfor/dev/oe-stuff/build/tmp/staging/i686-linux/usr/sbin:/home/bjornfor/dev/oe-stuff/build/tmp/staging/i686-linux/usr/bin:/home/bjornfor/dev/oe-stuff/build/tmp/cross/armv5te/bin:/home/bjornfor/dev/oe-stuff/build/tmp/staging/i686-linux/sbin:/home/bjornfor/dev/oe-stuff/build/tmp/staging/i686-linux/bin:/home/bjornfor/dev/oe-stuff/bitbake/bin:/home/bjornfor/dev/sb2/bin:/home/bjornfor/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/home/bjornfor/dev/arm-2008q3/bin:/opt/mv_pro_5.0/montavista/pro/devkit/arm/v5t_le/bin:/home/bjornfor/bin;
git read-tree 03022e38318bef016836dae71c70b9ea975723e7 failed with
signal 128, output:
fatal: failed to unpack tree object 03022e38318bef016836dae71c70b9ea975723e7

ERROR: TaskFailed event exception, aborting
ERROR: Build of
/home/bjornfor/dev/oe-stuff/openembedded/recipes/linux/linux-davinci_git.bb
do_fetch failed
---------------------------------------------------------------

Regards,
Bjørn Forsman

Re: bitbake virtual/kernel fails for davinci

[Denys Dmytriyenko]

On Tue, Jun 30, 2009 at 01:44:43PM +0200, Bj??rn Forsman wrote:
> Hi,
> 
> I am a bit new to OpenEmbedded, but have successfully built angstrom
> for davinci and qemu in the past. Today I tried to "bitbake
> virtual/kernel" for a davinci (dm6446) system. The build fails during
> unpack of linux-davinci kernel because git cannot find a given object
> hash (see output below). Has anybody else experienced this? How come
> OE references a non-existent git object?
> 
> The non-existent git object hash is specified in
> "recipes/linux/linux-davinci_git.bb" and was lastly modified by commit
> 160616d5f2a57ba828ed5ec06e37fc1cd4713af7.

Since then linux-davinci tree was updated to 2.6.30 final and more 
specifically, staging/vpfe branch was rebased from v2 to v3 vpfe patches. 
Unfortunately, rebase rewrites the history and changes commit numbers, hence 
the old commit cannot be found any more...

I'll update the recipe with the latest state of the linux-davinci tree. But, 
it may brake in the future again, as this is a moving target. If you want 
something, which does not break that often, I'd suggest you don't use the 
"git" version of the recipe - try setting PREFFERED_PROVIDER/VERSION for 
2.6.28 or 2.6.29...

-- 
Denys

Re: bitbake virtual/kernel fails for davinci

[Bjørn Forsman]

2009/6/30 Denys Dmytriyenko <denis@denix.org>:
> On Tue, Jun 30, 2009 at 01:44:43PM +0200, Bj??rn Forsman wrote:
>> Hi,
>>
>> I am a bit new to OpenEmbedded, but have successfully built angstrom
>> for davinci and qemu in the past. Today I tried to "bitbake
>> virtual/kernel" for a davinci (dm6446) system. The build fails during
>> unpack of linux-davinci kernel because git cannot find a given object
>> hash (see output below). Has anybody else experienced this? How come
>> OE references a non-existent git object?
>>
>> The non-existent git object hash is specified in
>> "recipes/linux/linux-davinci_git.bb" and was lastly modified by commit
>> 160616d5f2a57ba828ed5ec06e37fc1cd4713af7.
>
> Since then linux-davinci tree was updated to 2.6.30 final and more
> specifically, staging/vpfe branch was rebased from v2 to v3 vpfe patches.
> Unfortunately, rebase rewrites the history and changes commit numbers, hence
> the old commit cannot be found any more...

Ok. Thanks for explaining.

> I'll update the recipe with the latest state of the linux-davinci tree. But,
> it may brake in the future again, as this is a moving target. If you want
> something, which does not break that often, I'd suggest you don't use the
> "git" version of the recipe - try setting PREFFERED_PROVIDER/VERSION for
> 2.6.28 or 2.6.29...

Ok. I'll try setting PREFERRED_PROVIDER/VERSION next thing tomorrow.

BTW, the currently missing git object was specified as a SHA1 hash in
a variable called SRCREV. Could this variable be used for specifying
git objects in other formats? For example "staging/vpfe" or "HEAD"? Or
is it unwise to have recipes which uses "floating" revisions?

Regards,
Bjørn Forsman

Re: bitbake virtual/kernel fails for davinci

[Denys Dmytriyenko]

On Tue, Jun 30, 2009 at 09:14:30PM +0200, Bj??rn Forsman wrote:
> 2009/6/30 Denys Dmytriyenko <denis@denix.org>:
> > Since then linux-davinci tree was updated to 2.6.30 final and more
> > specifically, staging/vpfe branch was rebased from v2 to v3 vpfe patches.
> > Unfortunately, rebase rewrites the history and changes commit numbers, hence
> > the old commit cannot be found any more...
> 
> Ok. Thanks for explaining.
> 
> > I'll update the recipe with the latest state of the linux-davinci tree. But,
> > it may brake in the future again, as this is a moving target. If you want
> > something, which does not break that often, I'd suggest you don't use the
> > "git" version of the recipe - try setting PREFFERED_PROVIDER/VERSION for
> > 2.6.28 or 2.6.29...
> 
> Ok. I'll try setting PREFERRED_PROVIDER/VERSION next thing tomorrow.

I guess I should also add 2.6.30 version of the recipe. Although I've seen 
Kevin updating/moving the "v2.6.30-davinci1" tag...

> BTW, the currently missing git object was specified as a SHA1 hash in
> a variable called SRCREV. Could this variable be used for specifying
> git objects in other formats? For example "staging/vpfe" or "HEAD"? Or
> is it unwise to have recipes which uses "floating" revisions?

The branch "staging/vpfe" is already specified in the SRC_URI. As of pointing 
to the HEAD, you can check the comments in the linux-davinci_2.6.27.bb, if it 
still works. But it is not recommended, as it would be hard to manage 
different builds and support the correct upgrade path. You can still do it 
locally, as long as you keep updating the MACHINE_KERNEL_PR in the 
conf/machine/include/davinci.inc

-- 
Denys