state of the cvs tree

state of the cvs tree

[Christoph Hellwig]

looks like the cvs tree is broken currently - fs/xfs/ is merged up to
2.6.23-rc, but everything else is still at 2.6.22-rc state leading to
various compile failures.

Re: state of the cvs tree

[Mark Goodwin]

Christoph Hellwig wrote:
> looks like the cvs tree is broken currently - fs/xfs/ is merged up to
> 2.6.23-rc, but everything else is still at 2.6.22-rc state leading to
> various compile failures.

I think Tim is in the middle of the .23 update and still has some more
to push in. Tim?

What else do you (or anyone for that matter) have in the pipeline for XFS?
Whilst we're taking huge patches and cleanups, let's get them all in asap.

Thanks

-- 

  Mark Goodwin                                  markgw@sgi.com
  Engineering Manager for XFS and PCP    Phone: +61-3-99631937
  SGI Australian Software Group           Cell: +61-4-18969583
-------------------------------------------------------------

Re: state of the cvs tree

[David Chinner]

On Thu, Sep 13, 2007 at 09:05:15AM +1000, Mark Goodwin wrote:
> 
> 
> Christoph Hellwig wrote:
> >looks like the cvs tree is broken currently - fs/xfs/ is merged up to
> >2.6.23-rc, but everything else is still at 2.6.22-rc state leading to
> >various compile failures.
> 
> I think Tim is in the middle of the .23 update and still has some more
> to push in. Tim?
> 
> What else do you (or anyone for that matter) have in the pipeline for XFS?
> Whilst we're taking huge patches and cleanups, let's get them all in asap.

Let's plan this a little better than "ASAP". We've already got a
full queue for .24 - I'm uncomfortable with pushing anything more
given the nature of the changes we've made in this cycle and we
really want some testing time on that code before the .24 merge
window opens. Given that we are at .23-rc6 already, it won't be long
before .24-rc1 merge window is open, so lets stop pushing large
changes into the tree until after the .24-rc1 merge is done.

IOWs, I consider stuff like Eric's spin lock clean to be .25 material
at this point, not .24, and we should only be taking bug fixes and small,
contained features (e.g. fallocate support) for .24. Everything else
can wait until .25....

Thoughts?

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group

Re: state of the cvs tree

[Lachlan McIlroy]

David Chinner wrote:
> On Thu, Sep 13, 2007 at 09:05:15AM +1000, Mark Goodwin wrote:
>>
>> Christoph Hellwig wrote:
>>> looks like the cvs tree is broken currently - fs/xfs/ is merged up to
>>> 2.6.23-rc, but everything else is still at 2.6.22-rc state leading to
>>> various compile failures.
>> I think Tim is in the middle of the .23 update and still has some more
>> to push in. Tim?
>>
>> What else do you (or anyone for that matter) have in the pipeline for XFS?
>> Whilst we're taking huge patches and cleanups, let's get them all in asap.
> 
> Let's plan this a little better than "ASAP". We've already got a
> full queue for .24 - I'm uncomfortable with pushing anything more
> given the nature of the changes we've made in this cycle and we
> really want some testing time on that code before the .24 merge
> window opens. Given that we are at .23-rc6 already, it won't be long
> before .24-rc1 merge window is open, so lets stop pushing large
> changes into the tree until after the .24-rc1 merge is done.
> 
> IOWs, I consider stuff like Eric's spin lock clean to be .25 material
> at this point, not .24, and we should only be taking bug fixes and small,
> contained features (e.g. fallocate support) for .24. Everything else
> can wait until .25....
> 
> Thoughts?
> 

Sounds fair to me, Dave.

Re: state of the cvs tree

[Timothy Shimmin]

Mark Goodwin wrote:
> 
> 
> Christoph Hellwig wrote:
>> looks like the cvs tree is broken currently - fs/xfs/ is merged up to
>> 2.6.23-rc, but everything else is still at 2.6.22-rc state leading to
>> various compile failures.
> 
> I think Tim is in the middle of the .23 update and still has some more
> to push in. Tim?
> 
No I don't :-) I did have some followup dmapi fixes but they got added
yesterday.
The sgi ptools 2.6.x-xfs tree seems just fine for me, is building and
does have the 23-rc4 (only rc4 b/c only had kdb for that at the time) AFAICS.
I guess there must be something wrong then with the cvs sync.
It must be sync'ing the embedded xfs-linux tree okay but not
the 2.6.x-xfs kernel tree.
I'll see what I can find out.

BTW, sorry for not mentioning the update on oss. I will do that
next time.

--Tim

Re: state of the cvs tree

[Timothy Shimmin]

Hi Christoph,

Christoph Hellwig wrote:
> looks like the cvs tree is broken currently - fs/xfs/ is merged up to
> 2.6.23-rc, but everything else is still at 2.6.22-rc state leading to
> various compile failures.
> 
Looking at cvs web it looks like the 2.6.x-xfs was updated 8 hours
ago. So I am guessing that you saw a state of the tree whilst it
was doing its sync up.
Let me know if things are not fine for you still?

--Tim

Re: state of the cvs tree

[Timothy Shimmin]

Timothy Shimmin wrote:
> Hi Christoph,
> 
> Christoph Hellwig wrote:
>> looks like the cvs tree is broken currently - fs/xfs/ is merged up to
>> 2.6.23-rc, but everything else is still at 2.6.22-rc state leading to
>> various compile failures.
>>
> Looking at cvs web it looks like the 2.6.x-xfs was updated 8 hours
> ago. So I am guessing that you saw a state of the tree whilst it
> was doing its sync up.
> Let me know if things are not fine for you still?
> 
> --Tim
> 

Looking at the oss scripts briefly it reminded me of all the trees
involved:
   $s[0]->{"fromdir"} = "$pushroot/p_src/slinx_2.4.x-xfs";
   $s[0]->{"todir"} =   "$pushroot/CVSROOT/slinx_2.4.x-xfs";

   $s[1]->{"fromdir"} = "$pushroot/p_src/slinx_2.6.x-xfs";
   $s[1]->{"todir"} =   "$pushroot/CVSROOT/slinx_2.6.x-xfs";

   $s[2]->{"fromdir"} = "$pushroot/p_src/xfs-cmds";
   $s[2]->{"todir"} =   "$pushroot/tmp/CVSROOT/xfs-cmds";

   $s[3]->{"fromdir"} = "$pushroot/p_src/xfs-linux";
   $s[3]->{"todir"} =   "$pushroot/tmp/CVSROOT/xfs-linux";

   $s[4]->{"fromdir"} = "$pushroot/p_src/dmapi-linux";
   $s[4]->{"todir"} =   "$pushroot/CVSROOT/dmapi";

   $s[5]->{"fromdir"} = "$pushroot/p_src/xfs-website";
   $s[5]->{"todir"} =   "$pushroot/CVSROOT/xfs-website";

The xfs-linux and xfs-dmapi trees are needed for 2.6.x-xfs and 2.4.x-xfs.
I modified xfs-linux, xfs-dmapi, and 2.6.x-xfs as part of the update.

BTW, Donald, we'll have to do something about the 2.4 ptools tree
and cvs sync up if/when 2.4 support is dropped.

--Tim

Re: state of the cvs tree

[Christoph Hellwig]

On Thu, Sep 13, 2007 at 11:41:22AM +1000, Timothy Shimmin wrote:
> Looking at cvs web it looks like the 2.6.x-xfs was updated 8 hours
> ago. So I am guessing that you saw a state of the tree whilst it
> was doing its sync up.
> Let me know if things are not fine for you still?

Everything is fine now, thanks a lot!

Re: state of the cvs tree

[Christoph Hellwig]

On Thu, Sep 13, 2007 at 09:05:15AM +1000, Mark Goodwin wrote:
> 
> 
> Christoph Hellwig wrote:
> >looks like the cvs tree is broken currently - fs/xfs/ is merged up to
> >2.6.23-rc, but everything else is still at 2.6.22-rc state leading to
> >various compile failures.
> 
> I think Tim is in the middle of the .23 update and still has some more
> to push in. Tim?
> 
> What else do you (or anyone for that matter) have in the pipeline for XFS?
> Whilst we're taking huge patches and cleanups, let's get them all in asap.

I have a long pipeline waiting, but as Dave said most of that really
shouldn't go into 2.6.24.

There's one patch from me that I sent a long time ago that's a trivial
cleanup and should probably go into 2.6.24 still, that's

	"[PATCH] kill unused IOMAP_EOF flag"

One thing that is in my alreayd submitted queue that should go into CVS
ASAP after a small review is:

	"[PATCH] kill probe_* sysctl leftovers"

this is stuff that never was in mainline, so putting it in seems fine.

Then I have a patch from Eric sitting in the front of my queue,

	"[PATCH V2] refactor xfs_mountfs for clarity & stack savings"

which might be a little too big for 2.6.24, but should at least go into
CVS ASAP.  I think Eric would be really happy to see it in 2.6.24 aswell
because that means FC8 could actually mount xfs out of the box without
running out of stack or something.

Re: state of the cvs tree

[Christoph Hellwig]

And there's another patch I still have in my already sent out queue for
dmapi that needs applying:

	"[PATCH] xfs_dmapi: add MODULE_ tags"

of course dmapi doesn't go to mainline so no 2.6.24 considerations here,
but getting rid of that no license specified warning would be nice.

Re: state of the cvs tree

[Eric Sandeen]

Christoph Hellwig wrote:

> Then I have a patch from Eric sitting in the front of my queue,
> 
> 	"[PATCH V2] refactor xfs_mountfs for clarity & stack savings"
> 
> which might be a little too big for 2.6.24, but should at least go into
> CVS ASAP.  I think Eric would be really happy to see it in 2.6.24 aswell
> because that means FC8 could actually mount xfs out of the box without
> running out of stack or something.

Yes ;-)  I've got that patch in Fedora now, but it'd be great to have it
come via upstream.

-Eric

Re: state of the cvs tree

[David Chinner]

On Thu, Sep 13, 2007 at 12:40:00PM +0200, Christoph Hellwig wrote:
> On Thu, Sep 13, 2007 at 09:05:15AM +1000, Mark Goodwin wrote:
> > 
> > 
> > Christoph Hellwig wrote:
> > >looks like the cvs tree is broken currently - fs/xfs/ is merged up to
> > >2.6.23-rc, but everything else is still at 2.6.22-rc state leading to
> > >various compile failures.
> > 
> > I think Tim is in the middle of the .23 update and still has some more
> > to push in. Tim?
> > 
> > What else do you (or anyone for that matter) have in the pipeline for XFS?
> > Whilst we're taking huge patches and cleanups, let's get them all in asap.
> 
> I have a long pipeline waiting, but as Dave said most of that really
> shouldn't go into 2.6.24.
> 
> There's one patch from me that I sent a long time ago that's a trivial
> cleanup and should probably go into 2.6.24 still, that's
> 
> 	"[PATCH] kill unused IOMAP_EOF flag"

Ah, that's still sitting in my tree from a past life. It fell through
the cracks, I think. It should go in to .24

> One thing that is in my alreayd submitted queue that should go into CVS
> ASAP after a small review is:
> 
> 	"[PATCH] kill probe_* sysctl leftovers"

*nod*. yeah, that's pretty trivial so should go as well. 

> this is stuff that never was in mainline, so putting it in seems fine.
> 
> Then I have a patch from Eric sitting in the front of my queue,
> 
> 	"[PATCH V2] refactor xfs_mountfs for clarity & stack savings"
> 
> which might be a little too big for 2.6.24, but should at least go into
> CVS ASAP.  I think Eric would be really happy to see it in 2.6.24 aswell
> because that means FC8 could actually mount xfs out of the box without
> running out of stack or something.

Yeah, that's been floating about for a bit and has been tested in
FC8 so seems like a no-brainer for .24.

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group

Re: state of the cvs tree

[Eric Sandeen]

David Chinner wrote:

>> Then I have a patch from Eric sitting in the front of my queue,
>>
>> 	"[PATCH V2] refactor xfs_mountfs for clarity & stack savings"
>>
>> which might be a little too big for 2.6.24, but should at least go into
>> CVS ASAP.  I think Eric would be really happy to see it in 2.6.24 aswell
>> because that means FC8 could actually mount xfs out of the box without
>> running out of stack or something.
> 
> Yeah, that's been floating about for a bit and has been tested in
> FC8 so seems like a no-brainer for .24.

you're assuming anyone besides me tested xfs in F8TestX... *grin*

(F8Test2 was released today... hint hint...)

-Eric