davinci vs. v4l2: lots of conflicts in merge for linux-next

davinci vs. v4l2: lots of conflicts in merge for linux-next

[Kevin Hilman]

OK, this has gotten a bit out of control, to the point where I cannot
solve this myself.  This is partially due to me being on the road and
not keeping a close enough eye on the various video patches.

I've pushed a new 'davinci-next' branch to davinci git[1] which is
what I would like to make available for linux-next.  This includes all
the patches from davinci git master which touch
arch/arm/mach-davinci/*.

I then went to do a test a merge of the master branch of Mauro's
linux-next tree, and there are lots of conflicts.  Some are trivial to
resolve (the various I2C_BOARD_INFO() conflicts) but others are more
difficult, and someone more familar with the video drivers should sort
them out.

The two patches from davinci master that seem to be causing all the
problems are:

  ARM: DaVinci: DM646x Video: Platform and board specific setup
  davinci: video: restructuring to support vpif capture driver

These cause the conflicts with the v4l2 next tree.  So, in
davinci-next I've dropped these two patches.

I think the way to fix this is for someone to take all the board
changes from the v4l2 tree and rebase them on top of my davinci-next,
dropping them from v4l2 next. I'll then merge them into davinci-next,
and this should make the two trees merge properly in linux-next.

We need to get this sorted out soon so that they can be merged for the
next merge window.

Going forward, I would prefer that all changes to arch/arm/* stuff go
through davinci git and all drivers/* stuff goes through V4L2.  This
will avoid this kind of overlap/conflict in the future since DaVinci
core code is going through lots of changes.

Kevin

[1] git://git.kernel.org/pub/scm/linux/kernel/git/khilman/linux-davinci.git

RE: davinci vs. v4l2: lots of conflicts in merge for linux-next

[Karicheri, Muralidharan]

Kevin,

Ok, I see you have merged vpif capture architecture part to master branch
of davinci. 

So what you are suggesting is to remove all vpif/vpfe patches from arch/arm/davinci of v4l linux-next tree (So I guess this is what Mauro should do on linux-next). So architecture part of all future video patches are to be re-created and re-submitted based on davinci-next and will be merged only to davinci tree and Mauro will merge the v4l part.

Kevin & Mauro,

So only concern I have is that these patches may not compile (either architecture part or v4l part) until the counter part becomes available on the tree. Is this fine? 

Murali Karicheri
Software Design Engineer
Texas Instruments Inc.
Germantown, MD 20874
new phone: 301-407-9583
Old Phone : 301-515-3736 (will be deprecated)
email: m-karicheri2@ti.com

>-----Original Message-----
>From: Kevin Hilman [mailto:khilman@deeprootsystems.com]
>Sent: Wednesday, August 26, 2009 5:00 AM
>To: Karicheri, Muralidharan; Mauro Carvalho Chehab
>Cc: linux-media@vger.kernel.org; Hans Verkuil; DaVinci
>Subject: davinci vs. v4l2: lots of conflicts in merge for linux-next
>
>OK, this has gotten a bit out of control, to the point where I cannot
>solve this myself.  This is partially due to me being on the road and
>not keeping a close enough eye on the various video patches.
>
>I've pushed a new 'davinci-next' branch to davinci git[1] which is
>what I would like to make available for linux-next.  This includes all
>the patches from davinci git master which touch
>arch/arm/mach-davinci/*.
>
>I then went to do a test a merge of the master branch of Mauro's
>linux-next tree, and there are lots of conflicts.  Some are trivial to
>resolve (the various I2C_BOARD_INFO() conflicts) but others are more
>difficult, and someone more familar with the video drivers should sort
>them out.
>
>The two patches from davinci master that seem to be causing all the
>problems are:
>
>  ARM: DaVinci: DM646x Video: Platform and board specific setup
>  davinci: video: restructuring to support vpif capture driver
>
>These cause the conflicts with the v4l2 next tree.  So, in
>davinci-next I've dropped these two patches.
>
>I think the way to fix this is for someone to take all the board
>changes from the v4l2 tree and rebase them on top of my davinci-next,
>dropping them from v4l2 next. I'll then merge them into davinci-next,
>and this should make the two trees merge properly in linux-next.
>
>We need to get this sorted out soon so that they can be merged for the
>next merge window.
>
>Going forward, I would prefer that all changes to arch/arm/* stuff go
>through davinci git and all drivers/* stuff goes through V4L2.  This
>will avoid this kind of overlap/conflict in the future since DaVinci
>core code is going through lots of changes.
>
>Kevin
>
>[1] git://git.kernel.org/pub/scm/linux/kernel/git/khilman/linux-davinci.git

Re: davinci vs. v4l2: lots of conflicts in merge for linux-next

[Mauro Carvalho Chehab]

Em Wed, 26 Aug 2009 11:00:22 -0500
"Karicheri, Muralidharan" <m-karicheri2@ti.com> escreveu:

> Kevin,
> 
> Ok, I see you have merged vpif capture architecture part to master branch
> of davinci. 
> 
> So what you are suggesting is to remove all vpif/vpfe patches from arch/arm/davinci of v4l linux-next tree (So I guess this is what Mauro should do on linux-next). So architecture part of all future video patches are to be re-created and re-submitted based on davinci-next and will be merged only to davinci tree and Mauro will merge the v4l part.

I'll drop those patches from my tree.

> Kevin & Mauro,
> 
> So only concern I have is that these patches may not compile (either architecture part or v4l part) until the counter part becomes available on the tree. Is this fine? 

The strategy we use for solving those troubles is to move the Kbuild patches
that adds the compilation for the driver to be merged at the end of the series.
If, without this patch, the kernel will keep compiling, everything is fine.

Cheers,
Mauro

Re: davinci vs. v4l2: lots of conflicts in merge for linux-next

[Mauro Carvalho Chehab]

Em Wed, 26 Aug 2009 12:00:11 +0300
Kevin Hilman <khilman@deeprootsystems.com> escreveu:

> OK, this has gotten a bit out of control, to the point where I cannot
> solve this myself.  This is partially due to me being on the road and
> not keeping a close enough eye on the various video patches.
> 
> I've pushed a new 'davinci-next' branch to davinci git[1] which is
> what I would like to make available for linux-next.  This includes all
> the patches from davinci git master which touch
> arch/arm/mach-davinci/*.
> 
> I then went to do a test a merge of the master branch of Mauro's
> linux-next tree, and there are lots of conflicts.  Some are trivial to
> resolve (the various I2C_BOARD_INFO() conflicts) but others are more
> difficult, and someone more familar with the video drivers should sort
> them out.
> 
> The two patches from davinci master that seem to be causing all the
> problems are:
> 
>   ARM: DaVinci: DM646x Video: Platform and board specific setup
>   davinci: video: restructuring to support vpif capture driver

> These cause the conflicts with the v4l2 next tree.  So, in
> davinci-next I've dropped these two patches.

Hmm... here, I got a different set of patches...

I've created a quilt tree with the DaVinci patches I have here:
	http://linuxtv.org/downloads/patches-davinci/

There are 3 patches touching arch/arm:

$ grep -l arch/arm *
arch_arm_mach_davinci_platform_and_board_specific_setup_for_dm646x_evm.patch
v4l_dvb_dm355_platform_changes_for_vpfe_capture_driver.patch
v4l_dvb_dm6446_platform_changes_for_vpfe_capture_driver.patch

All those three files touches just arch/arm files:

$ diffstat -p1 `grep -l arch/arm *`
 arch/arm/mach-davinci/board-dm355-evm.c     |   76 ++++++++++++++++-
 arch/arm/mach-davinci/board-dm644x-evm.c    |   72 ++++++++++++++++
 arch/arm/mach-davinci/board-dm646x-evm.c    |  122 ++++++++++++++++++++++++++++
 arch/arm/mach-davinci/dm355.c               |   83 +++++++++++++++++++
 arch/arm/mach-davinci/dm644x.c              |   56 ++++++++++++
 arch/arm/mach-davinci/dm646x.c              |   62 ++++++++++++++
 arch/arm/mach-davinci/include/mach/dm355.h  |    2
 arch/arm/mach-davinci/include/mach/dm644x.h |    2
 arch/arm/mach-davinci/include/mach/dm646x.h |   24 +++++
 arch/arm/mach-davinci/include/mach/mux.h    |    9 ++
 10 files changed, 503 insertions(+), 5 deletions(-)

If I move those tree files out, the remaining ones are touching only on
drivers/media and include/media:

$ diffstat -p1 *
 drivers/media/video/Kconfig                    |   71
 drivers/media/video/Makefile                   |    4
 drivers/media/video/davinci/Makefile           |   15
 drivers/media/video/davinci/ccdc_hw_device.h   |  110 +
 drivers/media/video/davinci/dm355_ccdc.c       |  978 +++++++++++
 drivers/media/video/davinci/dm355_ccdc_regs.h  |  310 +++
 drivers/media/video/davinci/dm644x_ccdc.c      |  878 ++++++++++
 drivers/media/video/davinci/dm644x_ccdc_regs.h |  145 +
 drivers/media/video/davinci/vpfe_capture.c     | 2124 +++++++++++++++++++++++++
 drivers/media/video/davinci/vpif.c             |  234 ++
 drivers/media/video/davinci/vpif.h             |  632 +++++++
 drivers/media/video/davinci/vpif_display.c     | 1697 +++++++++++++++++++
 drivers/media/video/davinci/vpif_display.h     |  175 ++
 drivers/media/video/davinci/vpss.c             |  301 +++
 drivers/media/video/tvp514x.c                  | 1150 +++++--------
 drivers/media/video/tvp514x_regs.h             |   10
 include/media/davinci/ccdc_types.h             |   43
 include/media/davinci/dm355_ccdc.h             |  321 +++
 include/media/davinci/dm644x_ccdc.h            |  184 ++
 include/media/davinci/vpfe_capture.h           |  198 ++
 include/media/davinci/vpfe_types.h             |   51
 include/media/davinci/vpss.h                   |   69
 include/media/tvp514x.h                        |    4

> 
> 
> I think the way to fix this is for someone to take all the board
> changes from the v4l2 tree and rebase them on top of my davinci-next,
> dropping them from v4l2 next. I'll then merge them into davinci-next,
> and this should make the two trees merge properly in linux-next.
> 
> We need to get this sorted out soon so that they can be merged for the
> next merge window.
> 
> Going forward, I would prefer that all changes to arch/arm/* stuff go
> through davinci git and all drivers/* stuff goes through V4L2.  This
> will avoid this kind of overlap/conflict in the future since DaVinci
> core code is going through lots of changes.

Agreed. I dropped the 3 arch patches from my tree:
	http://linuxtv.org/downloads/patches-davinci/arch_arm_mach_davinci_platform_and_board_specific_setup_for_dm646x_evm.patch
	http://linuxtv.org/downloads/patches-davinci/v4l_dvb_dm355_platform_changes_for_vpfe_capture_driver.patch
	http://linuxtv.org/downloads/patches-davinci/v4l_dvb_dm6446_platform_changes_for_vpfe_capture_driver.patch

Could you please check those 3 patches and see what's missing on your tree?

That's said, we need to make sure that Stephen will merge both trees, otherwise
compilation will break for DaVinci. As the Makefiles are the last ones on my
tree, provided that I merge the DaVinci patches after you, we shouldn't have
troubles upstream.

P. S. I've removed DaVinci ML, as it is a subscription-only list and I'm not subscribed there.
> 
> Kevin
> 
> [1] git://git.kernel.org/pub/scm/linux/kernel/git/khilman/linux-davinci.git


-- 

Cheers,
Mauro

Re: davinci vs. v4l2: lots of conflicts in merge for linux-next

[Kevin Hilman]

"Karicheri, Muralidharan" <m-karicheri2@ti.com> writes:

> Kevin,
>
> Ok, I see you have merged vpif capture architecture part to master branch
> of davinci. 
>
> So what you are suggesting is to remove all vpif/vpfe patches from
> arch/arm/davinci of v4l linux-next tree (So I guess this is what
> Mauro should do on linux-next). So architecture part of all future
> video patches are to be re-created and re-submitted based on
> davinci-next and will be merged only to davinci tree and Mauro will
> merge the v4l part.

Yes.

Also note the two patches below that I dropped in davinci-next.  These
should be re-added as well.

Kevin

> Kevin & Mauro,
>
> So only concern I have is that these patches may not compile (either architecture part or v4l part) until the counter part becomes available on the tree. Is this fine? 
>
> Murali Karicheri
> Software Design Engineer
> Texas Instruments Inc.
> Germantown, MD 20874
> new phone: 301-407-9583
> Old Phone : 301-515-3736 (will be deprecated)
> email: m-karicheri2@ti.com
>
>>-----Original Message-----
>>From: Kevin Hilman [mailto:khilman@deeprootsystems.com]
>>Sent: Wednesday, August 26, 2009 5:00 AM
>>To: Karicheri, Muralidharan; Mauro Carvalho Chehab
>>Cc: linux-media@vger.kernel.org; Hans Verkuil; DaVinci
>>Subject: davinci vs. v4l2: lots of conflicts in merge for linux-next
>>
>>OK, this has gotten a bit out of control, to the point where I cannot
>>solve this myself.  This is partially due to me being on the road and
>>not keeping a close enough eye on the various video patches.
>>
>>I've pushed a new 'davinci-next' branch to davinci git[1] which is
>>what I would like to make available for linux-next.  This includes all
>>the patches from davinci git master which touch
>>arch/arm/mach-davinci/*.
>>
>>I then went to do a test a merge of the master branch of Mauro's
>>linux-next tree, and there are lots of conflicts.  Some are trivial to
>>resolve (the various I2C_BOARD_INFO() conflicts) but others are more
>>difficult, and someone more familar with the video drivers should sort
>>them out.
>>
>>The two patches from davinci master that seem to be causing all the
>>problems are:
>>
>>  ARM: DaVinci: DM646x Video: Platform and board specific setup
>>  davinci: video: restructuring to support vpif capture driver
>>
>>These cause the conflicts with the v4l2 next tree.  So, in
>>davinci-next I've dropped these two patches.
>>
>>I think the way to fix this is for someone to take all the board
>>changes from the v4l2 tree and rebase them on top of my davinci-next,
>>dropping them from v4l2 next. I'll then merge them into davinci-next,
>>and this should make the two trees merge properly in linux-next.
>>
>>We need to get this sorted out soon so that they can be merged for the
>>next merge window.
>>
>>Going forward, I would prefer that all changes to arch/arm/* stuff go
>>through davinci git and all drivers/* stuff goes through V4L2.  This
>>will avoid this kind of overlap/conflict in the future since DaVinci
>>core code is going through lots of changes.
>>
>>Kevin
>>
>>[1] git://git.kernel.org/pub/scm/linux/kernel/git/khilman/linux-davinci.git