git tree updates....

git tree updates....

[Dave Chinner]

Lachlan,

I just noticed you've been updating the git tree. You might want to
change the order that certain patches have been committed - the
changes to the fs/inode.c needed to be committed before the changes
to fs/xfs/xfs_inode.c that use them. They've been committed the
wrong way around and about 20 commits apart, so there's significant
region in the commit history where the kernel will fail to build.
Given the distance between the commits, there's a fair chance that
a git bisect could land in this range of compiliation failures.

Any chance of redoing these commits before you push to Linus so
that they are ordered correctly in the git tree?

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

Re: git tree updates....

[Lachlan McIlroy]

Dave Chinner wrote:
> Lachlan,
> 
> I just noticed you've been updating the git tree. You might want to
> change the order that certain patches have been committed - the
> changes to the fs/inode.c needed to be committed before the changes
> to fs/xfs/xfs_inode.c that use them. They've been committed the
> wrong way around and about 20 commits apart, so there's significant
> region in the commit history where the kernel will fail to build.
> Given the distance between the commits, there's a fair chance that
> a git bisect could land in this range of compiliation failures.
> 
> Any chance of redoing these commits before you push to Linus so
> that they are ordered correctly in the git tree?

Yeah I'll try to get them back in the right order.  Our merge tools
conveniently missed anything outside fs/xfs so I didn't realise I had
missed them until after I merge all the XFS mods.  I just wanted to
get something out for linux-next and also since the cvs tree was busted.

Out of curiosity Dave, why didn't you send those changes directly to
Linus?

Re: git tree updates....

[Dave Chinner]

On Mon, Oct 20, 2008 at 01:17:21PM +1000, Lachlan McIlroy wrote:
> Dave Chinner wrote:
>> Lachlan,
>>
>> I just noticed you've been updating the git tree. You might want to
>> change the order that certain patches have been committed - the
>> changes to the fs/inode.c needed to be committed before the changes
>> to fs/xfs/xfs_inode.c that use them. They've been committed the
>> wrong way around and about 20 commits apart, so there's significant
>> region in the commit history where the kernel will fail to build.
>> Given the distance between the commits, there's a fair chance that
>> a git bisect could land in this range of compiliation failures.
>>
>> Any chance of redoing these commits before you push to Linus so
>> that they are ordered correctly in the git tree?
>
> Yeah I'll try to get them back in the right order.  Our merge tools
> conveniently missed anything outside fs/xfs so I didn't realise I had
> missed them until after I merge all the XFS mods.  I just wanted to
> get something out for linux-next and also since the cvs tree was busted.

Fair enough - I wasn't sure that you noticed it.

> Out of curiosity Dave, why didn't you send those changes directly to
> Linus?

Because trying to co-ordinate an dependent updates in different
trees is extremely painful. We'd have to delay the XFS update 
till the inode patches were in mainline anyway, so it makes little
sense to separate them....

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

Re: git tree updates....

[Christoph Hellwig]

While we at it, the b0rked m_vfsmount addition for DMAPI keeps on
leaking into the git tree for -next..

Re: git tree updates....

[Lachlan McIlroy]

Christoph Hellwig wrote:
> While we at it, the b0rked m_vfsmount addition for DMAPI keeps on
> leaking into the git tree for -next..

Thanks for pointing that out - that shouldn't have happened.  It wont
be going to mainline.

The OSS xfs trees are in a bit of a mess and have been for a while.
I've been wanting to clean them up but figured it would be easier to
wait until we've moved over to git internally and then clone new trees.