[vlc-devel] [PATCH 1/2] d3d11_deinterlace: always fill the output picture->context

[Martin Storsjö]

On Tue, 8 Aug 2017, Steve Lhomme wrote:

> On Mon, Aug 7, 2017 at 8:04 PM, Rémi Denis-Courmont <remi at remlab.net> wrote:
>> Le maanantaina 7. elokuuta 2017, 12.46.22 EEST Steve Lhomme a écrit :
>>> It is necessary for snapshots which create a picture from scratch and then
>>> copy "pixels" using the planes or the context callback.
>>>
>>> Fixes #18533
>>>
>>> --
>>> replaces https://patches.videolan.org/patch/17586/
>>> - remove unnecessary extra indirection
>>
>> Is this notation supposed to be understood by git?
>>
>> My version (Debian 2.13.3) does not recognize it, which ostensibly annoys J-B.
>
> I know it's not welcome on ffmpeg/libav too.

What? Explicitly saying what has been changed since the previous patch is 
absolutely encouraged.

> But I think it's good for the reviewer to tell the differences with a 
> previous patch but should not be merged in the end. I wish there was a 
> formal/official way to do this.

If you separate it with three dashes instead of two as you did here, "git 
am" will not include it as part of the commit message, and libav 
absolutely welcomes that as well.

If the patch doesn't get merged via "git am" but directly some other way, 
and you wrote this annotation directly in the commit message instead of 
editing the output from git format-patch, I don't know of a way to 
automatically strip it though.

I do it that way myself, adding the annotations to commit messages, to 
incrementally work on them while working on the patches. Having some way 
to automatically strip it on e.g. "git cherrypick" would be great though.

// Martin