Re: [FAILED] BLD phyCORE-i.MX35 + linux-next git kernel

Hi,

On Wed, Jul 07, 2010 at 11:48:58AM +1000, Stephen Rothwell wrote:

Hi Uwe,

On Tue, 6 Jul 2010 17:22:46 +0200 Uwe Kleine-König [off-list ref] wrote:

quoted

our -next autobuilder barfs:

On Mon, Jul 05, 2010 at 11:10:37PM +0200, Pengutronix Sisyphus Druid wrote:

quoted

...
                              Source URL .... git://git.kernel.org/pub/scm/linux/kernel/git/sfr/linux-next.git
                        Current git hash .... 288092896e2097eebee7d4bf1df9a0c7b550e225
                    Kernel configuration .... use config: mx3_defconfig
                            Kernel build .... [FAILED]
...
  CC      arch/arm/mm/fault-armv.o
  CC      arch/arm/mm/flush.o
  CC      arch/arm/mach-mx3/clock-imx35.o
  CC      arch/arm/mach-mx3/mach-mx31ads.o
  CC      arch/arm/mach-mx3/mach-mx31lilly.o
  CC      arch/arm/kernel/armksyms.o
  CC      arch/arm/mm/ioremap.o
arch/arm/mach-mx3/mach-mx31lilly.c:300: error: redefinition of 'devices'
arch/arm/mach-mx3/mach-mx31lilly.c:295: error: previous definition of 'devices' was here

The problem is a mis-merge of i..MX/for-next as it contains a patch
c0d1dac4a91a665fb5b5f1674acf57dc8cdc07ec that is already in Linus' tree
as 4d5d85906ad109987504033fa04991f1b0a2e552:

OK, this is only the third time I have seen git mismerge something
automaitcally. :-(

I can reproduce that here[1].  Is this a bug in git?  Can somebody explain?
I expected a merge conflict or only a single occurence of devices[]

quoted

Stephen: The obvious fix is to remove one of the two identical devices[].

I have fixed this up in today's linux-next (and will continue to do so
while necessary).

How do you remember this.  By "hand" or do you have some automatism for
such things?

Best regards and thanks
Uwe

[1] 
	git clone -n git://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git
	cd linux-next
	git checkout 815c4163b6c8ebf8152f42b0a5fd015cfdcedc78
	git merge c8adc84533d7e98487b77154eedf2eaf0ed6037e
	grep devices[[] arch/arm/mach-mx3/mach-mx31lilly.c
	

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |

Hi Uwe,

On Wed, 7 Jul 2010 08:47:35 +0200 Uwe Kleine-König [off-list ref] wrote:

quoted

OK, this is only the third time I have seen git mismerge something
automaitcally. :-(

I can reproduce that here[1].  Is this a bug in git?  Can somebody explain?
I expected a merge conflict or only a single occurence of devices[]

I suspect it is a matter of not enough context (or the context looks the
same after applying one of the updates) for git to realise that both
trees are trying to add the same code in the same place.  There is a
conflict, but the recursive merge decides that each tree is adding
something different and so include them both (or some such :-)).  Most of
the time, git gets this right.

quoted

I have fixed this up in today's linux-next (and will continue to do so
while necessary).

How do you remember this.  By "hand" or do you have some automatism for
such things?

In this case, my scripts automatically apply a patch to the merge commit.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/