[FFmpeg-devel] [RFC] Development model

[Stefano Sabatini]

On date Thursday 2011-03-17 15:46:12 +0100, Michael Niedermayer encoded:
> On Thu, Mar 17, 2011 at 03:24:20PM +0100, Stefano Sabatini wrote:
> > On date Thursday 2011-03-17 15:15:50 +0100, Michael Niedermayer encoded:
> > > Hi
> > > 
> > > A quick question
> > > 
> > > What do people prefer
> > > A. Reviewer applies patch if he is sure its ok
> > > B. Reviewer ok-es patch and author applies later
> > 
> > Whatever is more convenient for both the reviewer and/or the author
> 
> true, maybe authors should state in their patch submission what they would
> prefer ?
> and i prefer to do the apply+push for them if they feel uncertain about git usage
> 
> 
> 
> > (and I forgot how to apply patches ;-).
> 
> git am -s the mail or attachment
> make fate
> git push --dry-run
> look at output, make sure you dont push funny branches or other funny things
> git log (-p)123456..123456 (from git dry-run)
> look at output, make sure this is what you want to push
> git push

Thanks.

I was too used to the git svn dcommit workflow, I was used to work in
separate branches, and use:
git svn dcommit REF

for committing the patches in the series up to REF.

So I was trying to find the equivalent in git. After some hours spent
digging docs and asking on #git I came to the conclusion that this
can't be directly achieved, so I'm doing now:

git log --pretty=oneline --reverse co BRANCH

I take note of the REF hash, then:
git co master
git cherry-pick REF
make test...
git push --dry-run
...
git push

Surely this can be automated (but for the moment I'll be prudent until
I'll get more confident).
-- 
FFmpeg = Forgiving and Fancy Mega Philosofic Ermetic Guru